Brain (1998), 121, 1603–1630
INVITED REVIEW
The perception of phantom limbs
The D. O. Hebb lecture
V. S. Ramachandran and William Hirstein
Center for Brain and Cognition, 0109, University of Correspondence to: V. S. Ramachandran, Center for Brain
California, San Diego, La Jolla, California, USA and Cognition, 0109, University of California, San Diego,
1610, LaJolla, CA 92093, USA. E-mail:
Summary
Almost everyone who has a limb amputated will humans, so that it is now possible to track perceptual
experience a phantom limb—the vivid impression that changes and changes in cortical topography in individual
the limb is not only still present, but in some cases, patients. We suggest, therefore, that these patients provide
painful. There is now a wealth of empirical evidence a valuable opportunity not only for exploring neural
demonstrating changes in cortical topography in plasticity in the adult human brain but also for
primates following deafferentation or amputation, and understanding the relationship between the activity of
this review will attempt to relate these in a systematic sensory neurons and conscious experience. We conclude
way to the clinical phenomenology of phantom limbs. with a theory of phantom limbs, some striking
With the advent of non-invasive imaging techniques such demonstrations of phantoms induced in normal subjects,
as MEG (magnetoencephalogram) and functional MRI, and some remarks about the relevance of these phenomena
topo- graphical reorganization can also be demonstrated to the question of how the brain constructs a ‘body
in image.’
Keywords: phantom pain; neural plasticity; somatosensory cortex; body image; synaesthesia
Abbreviations: fMRI = functional MRI; MEG = magnetoencephalogram
Contents Magnetoencephalogram correlates of cortical reorganization
Introduction Perceptual correlates of massive cortical reorganization
Part I: the phenomenology of phantom limbs Modality-specific effects
Incidence Mechanisms of reorganization
Onset Inter-subject variability
Duration Non-specific responses
Body part Extinction of referred sensations
Posture of the phantom Related studies by other groups
‘Telescoping’ Inter-manual referral of tactile sensations
Congenital phantoms Phantom limb pain
Factors enhancing or attenuating the phantom Reflecting on phantom limbs
Movement of the phantom Phantoms induced in normal individuals
Emergence of ‘repressed memories’ in the phantom Phantom limbs and sensory codes
A theory of phantom limbs
Part II: experimental findings The nature and nurture of phantom limbs
Plasticity in the somatosensory system Concluding remarks
© Oxford University Press 1998
,1604 V. S. Ramachandran and W. Hirstein
Introduction
A characteristic feature of the adult primate brain is the between the felt position of the limb and its actual position,
existence of a multiplicity of specialized areas, including e.g. as occurs during a spinal or brachial plexus block
distinct topographically organized ‘maps’ concerned with (Melzack and Bromage, 1973). It is important to note that
different sense modalities (Adrian and Zotterman, 1926; in all these cases the patient recognizes that the sensations
Mountcastle, 1957; Hubel and Wiesel, 1979; Lin and Kaas, are not veridical, i.e. what he/she experiences is an illusion,
1979; Sur et al., 1985; Kaas et al., 1990; Garraghty et al., not a delusion.
1991). In the visual domain alone, for example, over 30 Phantom limbs were probably known since antiquity and,
distinct areas have been described which contain either not surprisingly, there is an elaborate folklore surrounding
partial or complete maps of the visual field. A hundred years them. After Lord Nelson lost his right arm during an
of neurology, as well as three decades of single-unit unsuccessful attack on Santa Cruz de Tenerife, he experienced
recordings using microelectrodes, have established these compelling phantom limb pains, including the sensation of
basic ideas beyond any reasonable doubt (Allman and Kaas, fingers digging into his phantom palm. The emergence of
1971; Van Essen and Maunsell, 1980; Zeki, 1980). One of these ghostly sensations led the sea lord to proclaim that his
the most important early findings was that much of this phantom was a ‘direct proof of the existence of the soul’
intricate circuitry, especially in the primary visual cortex, is (Riddoch, 1941). If an arm can survive physical annihilation,
specified by the genome and remains largely stable why not the entire person?
throughout life, under ordinary circumstances (Wiesel and
Since the time of Mitchell’s (1872) original description,
Hubel, 1963). This finding provided a vindication of what
there have been literally hundreds of fascinating clinical case
neurology has always believed: that no new neural
reports of phantom limbs. However, there has been a tendency
connections can be formed in the adult mammalian brain.
to regard the syndrome as a clinical curiosity, and very little
Once connections have been laid down in foetal life, or
experimental work has been done on it. Contrary to this
during certain critical periods in early infancy, it was
view, we will argue that a study of phantom limbs can
assumed that they remain fixed throughout life. Indeed, it is
provide fundamental insights into the functional organization
this stability of connections in the adult brain that is often
of the normal human brain and that they can serve as
used to explain why there is usually very little functional
perceptual markers for tracking neural plasticity in the
recovery after damage to the nervous system and why
adult brain.
neurological diseases are so notoriously difficult to treat.
This article is divided into two parts. First, we will discuss
In the last two decades, however, several experiments on the
the phenomenology of phantom limbs. Secondly, we will
effects of deafferentation (or amputation) on somatosensory
describe some new experiments that have been done on
maps in adult primates and other mammals suggest that we
animals and on human patients, and we will attempt to link
may need to revise this conception of the nervous system
these experiments to the perceptual phenomena of phantoms.
(Wall, 1977; Kaas et al., 1983; Merzenich et al., 1984; Wall,
The first part will be brief since there are already a number
1984; Byrne and Calford, 1991; Calford and Tweedale,
of lucid reviews that deal specifically with the clinical
1991a, b; Pons et al., 1991; Florence and Kaas, 1995;
manifestations (e.g. Henderson and Smyth, 1948; Cronholm,
Florence et al., 1996; Kaas and Florence, 1997). These
1951; Sunderland, 1978; Melzack, 1992). The elegant work
animal experiments have shown that sensory maps can
of Melzack (1992) deserves special mention here, since he
indeed change in the adult brain, and they have been largely
has emphasized, quite correctly, that although stump
responsible for the current resurgence of interest in the
neuromas can contribute to phantom sensations, they are
clinical phenomenon of phantom limbs (Ramachandran,
merely part of a much more complicated picture. In
1993b). Taken collectively, the work on animals and human
particular, the occurrence of phantoms in patients born
patients provides a valuable experimental opportunity to
without limbs (La Croix et al., 1992; Ramachandran, 1993b;
investigate not only how new connections emerge in the
Saadah and Melzack, 1994) obviously cannot be due to
adult human brain, but also how information from different
neuromas, and it suggests that a central representation of the
sensory modules, e.g. touch, proprioception and vision,
limb survives after amputation and is largely responsible for
interact. The study of phantom limbs also provides an
the illusion of a phantom. The nature and origin of this
opportunity to understand exactly how the brain constructs
representation (and its neural basis) and the extent to which
a body image, and how this image is continuously updated
it can be modified by sensory experience will be the main
in response to changing sensory inputs.
The phrase ‘phantom limb’ was introduced by Silas Weir concern of Part II.
Patients also frequently complain that the phantom is
Mitchell (1871, 1872), who also provided their first clear
painful. The incidence of severe pain is such that it poses a
clinical description. Patients with this syndrome experience
major clinical problem; as many as 70% of phantoms remain
an amputated extremity as still present, and in some cases
painful even 25 years after loss of the limb (Sherman et al.,
also experience pain or cramping in the missing limb. The
1984). The origin of phantom pain is no less mysterious than
term is sometimes also used to designate a dissociation
the origin of phantoms themselves, although there is
certainly no shortage of speculation (see for example,
Sunderland,
, The perception of phantom 16
1978; Postone, 1987; Katz, 1992). Since very little is known opinion that neuromas are the primary cause of phantom
about the physiology of pain—especially its central limbs.
mechanisms—we will touch on this topic only briefly in this
review, despite its clinical importance (see Postone, 1987).
Body part
Although phantoms are most commonly reported after
Part I: the phenomenology of phantom limbs amputation of an arm or leg, they have also been reported
When one first encounters a patient with a phantom limb the following amputation of the breast (Scholz, 1993; Aglioti,
following questions arise. We shall try to provide answers to 1994a, b), parts of the face (Hoffman, 1955; Sacks, 1992) or,
them based partly on our own experience and partly on a sometimes, even internal viscera, e.g. one can have
number of earlier clinical case reports and review articles. sensations of bowel movement and flatus after a complete
removal of sigmoid colon and rectum (Ovesen et al., 1991),
and phantom ‘ulcer pains’ after partial gastrectomy (Szasz,
Incidence 1949). It has also been noted that phantom erections and
Almost immediately after the loss of a limb, between 90 and ejaculation can occur in paraplegics as well as in patients
98% of all patients experience a vivid phantom. There are who have had the penis removed (Sunderland, 1978), and we
hints that the incidence may be higher following a traumatic have personally seen patients with phantom menstrual
loss, or if there has been a pre-existing painful condition in cramps after hysterectomy, or even the acute pain of
the limb, than after a planned surgical amputation of a non- appendicitis following removal of the inflamed appendix.
painful limb. These findings suggest that very elaborate sensory memories
Phantoms are seen far less often in early childhood. Perhaps can re-emerge in the phantom in spite of, or perhaps as a
in young children there has not yet been enough time for the result of, deafferentation (see below under section headed
body image to ‘consolidate’. In the Simmel (1962) study, Emergence of ‘repressed memories’ in phantoms). The
phantoms were reported to occur in 20% of child amputees vividness of phantoms appears to depend on both cortical
<2 years old, in 25% of children between 2 and 4 years old, magnification (hence the vividness of perception of the
in 61% between 4 and 6 years old, in 75% between 6 and 8 phantom hand) as well as the subjective vividness of that part
years old and in 100% of children >8 years old. in one’s body image prior to amputation (which would explain
why phantoms occur more often following a traumatic loss,
or after a painful appendage has been removed, than after a
planned amputation of a non-painful limb). This might imply
Onset
that factors such as pre-amputation attention to a body part
Phantoms appear immediately in 75% of cases, as soon as
can modulate the subsequent vividness of the phantom—an
the anaesthetic wears off and the patient is conscious, but
observation that would have important clinical implications.
their appearance may be delayed by a few days or weeks in
the remaining 25% of patients (Moser, 1948). Carlen et al.
(1978) found that among male Israeli soldiers who
underwent amputations during the 1973 Yom Kippur War,
Posture of the phantom
Patients often comment that the phantom occupies a ‘habitual’
33% experienced phantom limb sensations immediately after
posture, e.g. partially flexed at the elbow, with the forearm
amputation, 32% within 24 h and 34% within a few weeks.
pronated. Spontaneous changes in posture are also common,
Onset is not affected by the limb amputated or the place
however. For instance, soon after the patient wakes up in
where the amputation is made (Sunderland, 1978).
the morning it may assume an unusual and sometimes
uncomfortable posture, only to return to the habitual posture
a few minutes later. Sometimes the phantom will also
Duration temporarily assume, or even become more permanently fixed
In many cases the phantom is present initially for a few days in, an awkward and painful posture (e.g. the arm twisted
or weeks, then gradually fades from consciousness. In back behind the head). Intriguingly, ‘memories’ of the limb’s
others, it may persist for years, even decades (30% of posture and form prior to amputation often survive in the
patients, according to Sunderland, 1978). There are case phantom (Jackson, 1889; Katz and Melzack, 1990); there is
reports of phantoms which persisted for 44 years even an anecdotal report of soldier who had a grenade
(Livingston, 1945) and 57 years (Abbatucci, 1894). explode in his hand, leaving behind a phantom hand stuck
Some patients are able to recall a phantom limb at in a permanently clenched and painful posture. We have
will after its disappearance with intense concentration or seen a patient whose arm was in a vertical wooden splint,
sometimes merely by rubbing the stump. Mitchell (1872) flexed at the elbow, with the fingers hooked over the end of
was able to resurrect a long-lost phantom by faradic the splint, gripping it tightly. Two days later his arm was
stimulation applied to the stump of an above-knee amputee. amputated, and when we saw him several weeks later, his
It is perhaps findings such as this that have led to the phantom was in exactly the same position that his real arm
widespread clinical had occupied, with the fingers hooked over an imaginary
INVITED REVIEW
The perception of phantom limbs
The D. O. Hebb lecture
V. S. Ramachandran and William Hirstein
Center for Brain and Cognition, 0109, University of Correspondence to: V. S. Ramachandran, Center for Brain
California, San Diego, La Jolla, California, USA and Cognition, 0109, University of California, San Diego,
1610, LaJolla, CA 92093, USA. E-mail:
Summary
Almost everyone who has a limb amputated will humans, so that it is now possible to track perceptual
experience a phantom limb—the vivid impression that changes and changes in cortical topography in individual
the limb is not only still present, but in some cases, patients. We suggest, therefore, that these patients provide
painful. There is now a wealth of empirical evidence a valuable opportunity not only for exploring neural
demonstrating changes in cortical topography in plasticity in the adult human brain but also for
primates following deafferentation or amputation, and understanding the relationship between the activity of
this review will attempt to relate these in a systematic sensory neurons and conscious experience. We conclude
way to the clinical phenomenology of phantom limbs. with a theory of phantom limbs, some striking
With the advent of non-invasive imaging techniques such demonstrations of phantoms induced in normal subjects,
as MEG (magnetoencephalogram) and functional MRI, and some remarks about the relevance of these phenomena
topo- graphical reorganization can also be demonstrated to the question of how the brain constructs a ‘body
in image.’
Keywords: phantom pain; neural plasticity; somatosensory cortex; body image; synaesthesia
Abbreviations: fMRI = functional MRI; MEG = magnetoencephalogram
Contents Magnetoencephalogram correlates of cortical reorganization
Introduction Perceptual correlates of massive cortical reorganization
Part I: the phenomenology of phantom limbs Modality-specific effects
Incidence Mechanisms of reorganization
Onset Inter-subject variability
Duration Non-specific responses
Body part Extinction of referred sensations
Posture of the phantom Related studies by other groups
‘Telescoping’ Inter-manual referral of tactile sensations
Congenital phantoms Phantom limb pain
Factors enhancing or attenuating the phantom Reflecting on phantom limbs
Movement of the phantom Phantoms induced in normal individuals
Emergence of ‘repressed memories’ in the phantom Phantom limbs and sensory codes
A theory of phantom limbs
Part II: experimental findings The nature and nurture of phantom limbs
Plasticity in the somatosensory system Concluding remarks
© Oxford University Press 1998
,1604 V. S. Ramachandran and W. Hirstein
Introduction
A characteristic feature of the adult primate brain is the between the felt position of the limb and its actual position,
existence of a multiplicity of specialized areas, including e.g. as occurs during a spinal or brachial plexus block
distinct topographically organized ‘maps’ concerned with (Melzack and Bromage, 1973). It is important to note that
different sense modalities (Adrian and Zotterman, 1926; in all these cases the patient recognizes that the sensations
Mountcastle, 1957; Hubel and Wiesel, 1979; Lin and Kaas, are not veridical, i.e. what he/she experiences is an illusion,
1979; Sur et al., 1985; Kaas et al., 1990; Garraghty et al., not a delusion.
1991). In the visual domain alone, for example, over 30 Phantom limbs were probably known since antiquity and,
distinct areas have been described which contain either not surprisingly, there is an elaborate folklore surrounding
partial or complete maps of the visual field. A hundred years them. After Lord Nelson lost his right arm during an
of neurology, as well as three decades of single-unit unsuccessful attack on Santa Cruz de Tenerife, he experienced
recordings using microelectrodes, have established these compelling phantom limb pains, including the sensation of
basic ideas beyond any reasonable doubt (Allman and Kaas, fingers digging into his phantom palm. The emergence of
1971; Van Essen and Maunsell, 1980; Zeki, 1980). One of these ghostly sensations led the sea lord to proclaim that his
the most important early findings was that much of this phantom was a ‘direct proof of the existence of the soul’
intricate circuitry, especially in the primary visual cortex, is (Riddoch, 1941). If an arm can survive physical annihilation,
specified by the genome and remains largely stable why not the entire person?
throughout life, under ordinary circumstances (Wiesel and
Since the time of Mitchell’s (1872) original description,
Hubel, 1963). This finding provided a vindication of what
there have been literally hundreds of fascinating clinical case
neurology has always believed: that no new neural
reports of phantom limbs. However, there has been a tendency
connections can be formed in the adult mammalian brain.
to regard the syndrome as a clinical curiosity, and very little
Once connections have been laid down in foetal life, or
experimental work has been done on it. Contrary to this
during certain critical periods in early infancy, it was
view, we will argue that a study of phantom limbs can
assumed that they remain fixed throughout life. Indeed, it is
provide fundamental insights into the functional organization
this stability of connections in the adult brain that is often
of the normal human brain and that they can serve as
used to explain why there is usually very little functional
perceptual markers for tracking neural plasticity in the
recovery after damage to the nervous system and why
adult brain.
neurological diseases are so notoriously difficult to treat.
This article is divided into two parts. First, we will discuss
In the last two decades, however, several experiments on the
the phenomenology of phantom limbs. Secondly, we will
effects of deafferentation (or amputation) on somatosensory
describe some new experiments that have been done on
maps in adult primates and other mammals suggest that we
animals and on human patients, and we will attempt to link
may need to revise this conception of the nervous system
these experiments to the perceptual phenomena of phantoms.
(Wall, 1977; Kaas et al., 1983; Merzenich et al., 1984; Wall,
The first part will be brief since there are already a number
1984; Byrne and Calford, 1991; Calford and Tweedale,
of lucid reviews that deal specifically with the clinical
1991a, b; Pons et al., 1991; Florence and Kaas, 1995;
manifestations (e.g. Henderson and Smyth, 1948; Cronholm,
Florence et al., 1996; Kaas and Florence, 1997). These
1951; Sunderland, 1978; Melzack, 1992). The elegant work
animal experiments have shown that sensory maps can
of Melzack (1992) deserves special mention here, since he
indeed change in the adult brain, and they have been largely
has emphasized, quite correctly, that although stump
responsible for the current resurgence of interest in the
neuromas can contribute to phantom sensations, they are
clinical phenomenon of phantom limbs (Ramachandran,
merely part of a much more complicated picture. In
1993b). Taken collectively, the work on animals and human
particular, the occurrence of phantoms in patients born
patients provides a valuable experimental opportunity to
without limbs (La Croix et al., 1992; Ramachandran, 1993b;
investigate not only how new connections emerge in the
Saadah and Melzack, 1994) obviously cannot be due to
adult human brain, but also how information from different
neuromas, and it suggests that a central representation of the
sensory modules, e.g. touch, proprioception and vision,
limb survives after amputation and is largely responsible for
interact. The study of phantom limbs also provides an
the illusion of a phantom. The nature and origin of this
opportunity to understand exactly how the brain constructs
representation (and its neural basis) and the extent to which
a body image, and how this image is continuously updated
it can be modified by sensory experience will be the main
in response to changing sensory inputs.
The phrase ‘phantom limb’ was introduced by Silas Weir concern of Part II.
Patients also frequently complain that the phantom is
Mitchell (1871, 1872), who also provided their first clear
painful. The incidence of severe pain is such that it poses a
clinical description. Patients with this syndrome experience
major clinical problem; as many as 70% of phantoms remain
an amputated extremity as still present, and in some cases
painful even 25 years after loss of the limb (Sherman et al.,
also experience pain or cramping in the missing limb. The
1984). The origin of phantom pain is no less mysterious than
term is sometimes also used to designate a dissociation
the origin of phantoms themselves, although there is
certainly no shortage of speculation (see for example,
Sunderland,
, The perception of phantom 16
1978; Postone, 1987; Katz, 1992). Since very little is known opinion that neuromas are the primary cause of phantom
about the physiology of pain—especially its central limbs.
mechanisms—we will touch on this topic only briefly in this
review, despite its clinical importance (see Postone, 1987).
Body part
Although phantoms are most commonly reported after
Part I: the phenomenology of phantom limbs amputation of an arm or leg, they have also been reported
When one first encounters a patient with a phantom limb the following amputation of the breast (Scholz, 1993; Aglioti,
following questions arise. We shall try to provide answers to 1994a, b), parts of the face (Hoffman, 1955; Sacks, 1992) or,
them based partly on our own experience and partly on a sometimes, even internal viscera, e.g. one can have
number of earlier clinical case reports and review articles. sensations of bowel movement and flatus after a complete
removal of sigmoid colon and rectum (Ovesen et al., 1991),
and phantom ‘ulcer pains’ after partial gastrectomy (Szasz,
Incidence 1949). It has also been noted that phantom erections and
Almost immediately after the loss of a limb, between 90 and ejaculation can occur in paraplegics as well as in patients
98% of all patients experience a vivid phantom. There are who have had the penis removed (Sunderland, 1978), and we
hints that the incidence may be higher following a traumatic have personally seen patients with phantom menstrual
loss, or if there has been a pre-existing painful condition in cramps after hysterectomy, or even the acute pain of
the limb, than after a planned surgical amputation of a non- appendicitis following removal of the inflamed appendix.
painful limb. These findings suggest that very elaborate sensory memories
Phantoms are seen far less often in early childhood. Perhaps can re-emerge in the phantom in spite of, or perhaps as a
in young children there has not yet been enough time for the result of, deafferentation (see below under section headed
body image to ‘consolidate’. In the Simmel (1962) study, Emergence of ‘repressed memories’ in phantoms). The
phantoms were reported to occur in 20% of child amputees vividness of phantoms appears to depend on both cortical
<2 years old, in 25% of children between 2 and 4 years old, magnification (hence the vividness of perception of the
in 61% between 4 and 6 years old, in 75% between 6 and 8 phantom hand) as well as the subjective vividness of that part
years old and in 100% of children >8 years old. in one’s body image prior to amputation (which would explain
why phantoms occur more often following a traumatic loss,
or after a painful appendage has been removed, than after a
planned amputation of a non-painful limb). This might imply
Onset
that factors such as pre-amputation attention to a body part
Phantoms appear immediately in 75% of cases, as soon as
can modulate the subsequent vividness of the phantom—an
the anaesthetic wears off and the patient is conscious, but
observation that would have important clinical implications.
their appearance may be delayed by a few days or weeks in
the remaining 25% of patients (Moser, 1948). Carlen et al.
(1978) found that among male Israeli soldiers who
underwent amputations during the 1973 Yom Kippur War,
Posture of the phantom
Patients often comment that the phantom occupies a ‘habitual’
33% experienced phantom limb sensations immediately after
posture, e.g. partially flexed at the elbow, with the forearm
amputation, 32% within 24 h and 34% within a few weeks.
pronated. Spontaneous changes in posture are also common,
Onset is not affected by the limb amputated or the place
however. For instance, soon after the patient wakes up in
where the amputation is made (Sunderland, 1978).
the morning it may assume an unusual and sometimes
uncomfortable posture, only to return to the habitual posture
a few minutes later. Sometimes the phantom will also
Duration temporarily assume, or even become more permanently fixed
In many cases the phantom is present initially for a few days in, an awkward and painful posture (e.g. the arm twisted
or weeks, then gradually fades from consciousness. In back behind the head). Intriguingly, ‘memories’ of the limb’s
others, it may persist for years, even decades (30% of posture and form prior to amputation often survive in the
patients, according to Sunderland, 1978). There are case phantom (Jackson, 1889; Katz and Melzack, 1990); there is
reports of phantoms which persisted for 44 years even an anecdotal report of soldier who had a grenade
(Livingston, 1945) and 57 years (Abbatucci, 1894). explode in his hand, leaving behind a phantom hand stuck
Some patients are able to recall a phantom limb at in a permanently clenched and painful posture. We have
will after its disappearance with intense concentration or seen a patient whose arm was in a vertical wooden splint,
sometimes merely by rubbing the stump. Mitchell (1872) flexed at the elbow, with the fingers hooked over the end of
was able to resurrect a long-lost phantom by faradic the splint, gripping it tightly. Two days later his arm was
stimulation applied to the stump of an above-knee amputee. amputated, and when we saw him several weeks later, his
It is perhaps findings such as this that have led to the phantom was in exactly the same position that his real arm
widespread clinical had occupied, with the fingers hooked over an imaginary
