ArticlePDF AvailableLiterature Review

The Proprioceptive Senses: Their Roles in Signaling Body Shape, Body Position and Movement, and Muscle Force


Abstract and Figures

This is a review of the proprioceptive senses generated as a result of our own actions. They include the senses of position and movement of our limbs and trunk, the sense of effort, the sense of force, and the sense of heaviness. Receptors involved in proprioception are located in skin, muscles, and joints. Information about limb position and movement is not generated by individual receptors, but by populations of afferents. Afferent signals generated during a movement are processed to code for endpoint position of a limb. The afferent input is referred to a central body map to determine the location of the limbs in space. Experimental phantom limbs, produced by blocking peripheral nerves, have shown that motor areas in the brain are able to generate conscious sensations of limb displacement and movement in the absence of any sensory input. In the normal limb tendon organs and possibly also muscle spindles contribute to the senses of force and heaviness. Exercise can disturb proprioception, and this has implications for musculoskeletal injuries. Proprioceptive senses, particularly of limb position and movement, deteriorate with age and are associated with an increased risk of falls in the elderly. The more recent information available on proprioception has given a better understanding of the mechanisms underlying these senses as well as providing new insight into a range of clinical conditions.
Content may be subject to copyright.
Uwe Proske and Simon C. Gandevia
Department of Physiology, Monash University, Victoria, Australia; and Neuroscience Research Australia
and University of New South Wales, Sydney, Australia
LProske U, Gandevia SC. The Proprioceptive Senses: Their Roles in Signaling Body
Shape, Body Position and Movement, and Muscle Force. Physiol Rev 92: 1651–1697,
2012; doi:10.1152/physrev.00048.2011.—This is a review of the proprioceptive
senses generated as a result of our own actions. They include the senses of position
and movement of our limbs and trunk, the sense of effort, the sense of force, and the
sense of heaviness. Receptors involved in proprioception are located in skin, muscles, and joints.
Information about limb position and movement is not generated by individual receptors, but by
populations of afferents. Afferent signals generated during a movement are processed to code for
endpoint position of a limb. The afferent input is referred to a central body map to determine the
location of the limbs in space. Experimental phantom limbs, produced by blocking peripheral
nerves, have shown that motor areas in the brain are able to generate conscious sensations of limb
displacement and movement in the absence of any sensory input. In the normal limb tendon organs
and possibly also muscle spindles contribute to the senses of force and heaviness. Exercise can
disturb proprioception, and this has implications for musculoskeletal injuries. Proprioceptive
senses, particularly of limb position and movement, deteriorate with age and are associated with
an increased risk of falls in the elderly. The more recent information available on proprioception has
given a better understanding of the mechanisms underlying these senses as well as providing new
insight into a range of clinical conditions.
It is useful and justifiable for level of investigation to have
its own language but we must expect that with a greater
advancement of our knowledge it will be easy to translate
one such language into another. Until such a time, each field
must develop along its own lines, unhindered by the many
possibilities for misinterpretation. E. von Holst (1954)
A. Introduction
In everyday activities we depend on signals coming from
our moving bodies to be able to respond to the space around
us and react rapidly in changing circumstances. Much of
this knowledge about position and movement of the limbs
and trunk is provided by sensations arising in propriocep-
tors. The information they provide allows us to maneuver
our way around obstacles in the dark and be able to manip-
ulate objects out of view.
The term proprioception was passed down to us by Sher-
rington (356). He stated, “In muscular receptivity we see
the body itself acting as a stimulus to its own receptors—the
proprioceptors.” In a strict interpretation of that definition,
our bodies are supplied by many types of proprioceptors,
not just those concerned with muscular sensibility. For ex-
ample, we have receptors signaling distension of arteries,
lungs, and the gut. Traditionally, however, the term propri-
oceptor has been restricted to receptors concerned with
conscious sensations, and these include the senses of limb
position and movement, the sense of tension or force, the
sense of effort, and the sense of balance. Kinesthesia, a term
introduced by Bastian (17), is used here to refer to sensa-
tions of limb position and movement. In this review we have
not discussed the sense of balance. However, like the pro-
prioceptive system, the vestibular system contributes to a
range of conscious sensations as well as the guidance of
movement and posture (for review, see Refs. 10, 87, 88).
Proprioceptive sensations are mysterious because we are
largely unaware of them. They are distinguishable from
Physiol Rev 92: 1651–1697, 2012
16510031-9333/12 Copyright © 2012 the American Physiological Society
exteroceptors such as the eye and the ear in that they are not
associated with specific, recognizable sensations. Yet, when
we are not actually looking at our limbs, we are able to
indicate with reasonable accuracy their positions and
whether they are moving. Part of the explanation for this
lack of identifiable sensations relates to the predictability of
proprioceptive signals. We are aware that we are making a
willed movement and so anticipate the sensory input that it
generates. A general concept in sensory physiology is that
what we feel commonly represents the difference between
what is expected and what has actually occurred (e.g., Refs.
19, 86, 423). On that basis, if a movement goes to plan and
there is no mismatch between the expected signals and those
actually generated, no definable sensation is produced, yet
the subject knows precisely the location of their limb. It is
possible to generate an artificial proprioceptive signal using
muscle vibration (159). Vibration produces sensations of
limb displacement and movement, leading the subject to
express astonishment at the unwilled nature of the sensa-
tions. This suggests that the will to move and the subse-
quent proprioceptive sensations are intimately linked.
The study of proprioception has always attracted wide-
spread interest. Part of the reason is the important role
played by proprioception in motor control. We are unable to
move towards a target without ongoing visual and proprio-
ceptive feedback. In addition, proprioception is appealing be-
cause it promises a better understanding of our everyday sen-
sory experiences. In recent years, the topic has received ad-
ditional impetus from developments in neuroimaging, in
particular, magnetic resonance imaging (MRI). It has al-
lowed study of the central activity patterns produced by
proprioceptive stimuli. That, in turn, has led to recognition
of the importance of integration of proprioceptive inputs
with inputs from other senses such as vision and touch and
identification of central areas likely to be involved in the
integration. We are beginning to understand how some pro-
prioceptive sensations arise and how they are used to create
a body image. In designing experiments to test these ideas, it
has also proven useful to resort to modern methods of gen-
erating virtual realities (341).
The subject of proprioception lies at the boundary between
neurophysiology and neuropsychology. In this review we
have taken a more physiological view and restricted our-
selves to a discussion of aspects of the physiology of pro-
prioceptors, their central projection patterns, and the gen-
erated sensations. An expanding field concerns the interac-
tions between proprioception, vision, and vestibular inputs.
While we discuss some of this, we have not reviewed the
area exhaustively. The same applies at the more psycholog-
ical end of the subject, for example, sensorimotor integra-
tion in the generation of concepts of “wellness,” emotions,
and social interactions. We have chosen not to review the
extensive literature on eye movements, although some ref-
erence will be made to it. For comprehensive reviews, see,
e.g., References 95, 216.
The topic of proprioception, as we have approached it, has
been reviewed before (136, 244, 300, 303, 304). However,
it is notable that some major textbooks claiming to cover all
aspects of neuroscience (e.g., Ref. 286) and a even multi-
volume “comprehensive” handbook on the senses (49) have
failed to include any detailed discussion of proprioception.
Such omissions plus the many new developments in the field
highlight the need for a reassessment of the topic and, hope-
fully, it will remind students, teachers, and others of the
importance of the material. In this review we have tried to
focus on recent developments, the role of motor commands
in the generation of proprioceptive sensations, the ensemble
signaling properties of proprioceptive afferents, and the in-
formation, based on MRI studies, of the central structures
identified as playing a role in the generation of propriocep-
tive sensations.
B. Historical Background
The history of proprioception has been the subject of dis-
cussion for hundreds of years, with ideas emerging, their
rejection, and subsequent reemergence as scientific progress
takes its tortuous path. In reading some of the 19th century
accounts, the sophistication of the ideas and clarity of ex-
pression are astonishing. An account of early speculations
and the rise of a proposal for a “sixth sense” is provided by
Wade (395). (Other articles with historical perspectives in-
clude References 95, 159, and 244.)
Aristotle firmly believed that there were only five senses:
sight, hearing, smell, taste, and touch. He specifically ex-
cluded the existence of a sixth sense (see 339). Speculations
about a muscle sense date back at least to the 17th century.
William Harvey (181) speculated about the fact that mus-
cles which move the fingers lie in the forearm. “Thus we
perceive and so feel the fingers to move, but truly we neither
perceive nor feel the movement of the muscles, which are in
the elbow”. Discovery of the sixth sense, the muscle sense, is
attributed to Bell (20). He posed the question, “(do) muscles
have any other purpose to serve than merely to contract
under the impulse of their motor nerves?” He concluded,
“We are sensible of the most minute changes of muscular
exertion, by which we know the position of the body and
limbs, when there is no other means of knowledge open to
us.” Bell also speculated about whether the signals were of
central or peripheral origin.
The idea of a muscle sense was debated repeatedly during
the 19th century. German physiologists talked about the
“Muskelsinn.” What was meant here was not a sensation
originating in the muscles themselves, but in the brain. It
was also referred to as a “sensation of innervation” (184,
268). The idea was that whenever we willed a movement,
1652 Physiol Rev VOL 92 OCTOBER 2012
this gave rise centrally to sensations of muscular activity
and movement. Sherrington (355) in his influential text-
book chapter on “The muscular sense” rejected these ideas,
largely based on the observation that in the absence of mo-
tor commands, when our limbs lie relaxed, we still know
where they are, even if we are not looking at them. Sher-
rington stated, “An objection to this hypothesis (sensation
of innervation) is that it sunders sharply the sensations of
passive from those of active movement, whereas there is
strong ground for believing the two intimately allied.” So
there were two schools of thought, one claiming that the
muscle sense had an entirely central origin, and the other
believing that a peripheral signal was principally responsi-
ble. Bastian (17), who coined the term kinesthesis, was the
only one at the time who contemplated a hybrid theory,
comprising both central and peripheral components (see
also Ref. 95). He later abandoned this idea in favor of a
purely peripheral mechanism [for details, see Jones (204)].
Sherrington’s views prevailed, and for the first half of the
20th century the subject of proprioception was largely
founded on muscular sensations, although some counter
views were proposed by the psychologist Lashley (222).
In his account of the muscle senses, Sherrington passed
without comment from the role of muscle receptors in “ac-
tive movement” to their contribution to spinal reflex action.
This brushes over the two remarkably different roles of
muscle receptors, their contribution to conscious sensation
as proprioceptors, and their unconscious, automatic reflex
action. This dichotomy of action led Merton (257) to de-
clare, “Everyone agrees that the muscle spindles are the
receptors for the stretch reflex and the mechanisms based on
it, but it is also held at the same time that the muscle spin-
dles give information about the length of the muscles to be
used in our conscious judgements of limb position. Until the
underlying incompatibility of these two notions is felt, one
cannot properly appreciate the character of the problems
that face us in this field.” Here it is worth reflecting on the
strategy of assigning to one class of receptors such diverse
roles and what this means for central integrative processes.
In the present review we have limited the meaning of “pro-
prioceptor” to a receptor that gives rise to conscious sensa-
tions (300). In doing so, we are aware that receptors such as
muscle spindles and tendon organs also play important
roles in the unconscious, reflex control of movements.
When we think about limb position sense, an obvious place
to look for receptors signaling position is in the joint about
which the limb moves. That, indeed, was the prevailing
view for much of the 20th century. Kinesthetic sensations
were thought to arise exclusively in the joints. Sherrington
himself was cautious, “The individual contributions to-
ward muscular sense of these different sets of end-organs
can only be conjectured” (355). In any case, as single affer-
ent recording techniques were introduced, the view arose
that the principal source of kinesthetic information lay in
the joints themselves. The experimental basis for this view
was provided by Boyd and Roberts (33) and Skoglund
(362) [see also Skoglund (363), Mountcastle and Powell
(267), and Jones (204)]. Today we know that this is not the
case and that muscle receptors are the principal kinesthetic
receptors. In an influential series of animal studies which
challenged the view of joint receptors as the principal kin-
esthetic sensors, Burgess and colleagues (42, 43) showed
that slowly adapting receptors in the knee joint of the cat
provided ambiguous positional information because they
increased their rate of discharge at both extremes of joint
position. Here there was an interesting twist to the story. It
was claimed that in the cat posterior articular nerve supply-
ing the knee joint there were afferents which discharged
across the mid range of knee position (33, 115, 362). These
afferents changed their discharge rate monotonically with
changes in joint angle, as would be expected from a receptor
signaling joint position. Subsequently, however, these dis-
charges were shown to come from muscle spindles in the
popliteus muscle whose afferents had taken an aberrant
course in the posterior articular nerve (250).
The modern view has muscle spindles as the principal pro-
prioceptors. Here it is interesting to reflect on the tortuous
path taken by progress in the field. Early, quite sophisti-
cated ideas on a sensation of innervations were replaced by
a view that focussed entirely on peripheral receptors. But
this was still wrong, since the main receptor type was con-
sidered to be joint receptors, not muscle receptors. Nowa-
days, while we are aware of the importance of peripheral
signals for passive proprioception, we have revived some of
the old ideas on a sensation of innervations, to give “out-
flow” signals a bigger role, especially in active propriocep-
tion, where the generation of peripheral signals is accom-
panied by voluntary motor activity.
A. Introduction
During limb movement and changes in position, the tissues
around the relevant joints will be deformed, including skin,
muscles, tendons, fascia, joint capsules, and ligaments (e.g.,
Refs. 1, 169). All these tissues are innervated by mechani-
cally sensitive receptors, and their density varies across
muscles and regions of the body (e.g., Ref. 16). The ques-
tion arises, which are the principal kinesthetic receptors? In
this review the view is argued that muscle spindles play the
major role in kinesthesia, with some skin receptors provid-
ing additional information. Emerging views suggest that
Golgi tendon organs contribute to proprioception, includ-
ing the senses of force and heaviness. Here the evidence
remains indirect, the problem being that it is difficult to
activate a population of tendon organs selectively. The ma-
jor features of the muscle spindle and tendon organ are
depicted in FIGURES 1 and 2, respectively. Joint receptors
1653Physiol Rev VOL 92 OCTOBER 2012
probably play only a minor role at most joints, acting as
limit detectors. However, there is evidence of a contribution
by joint receptors in the mid-range of movements at the
finger joints (69, 116). Here we have considered the evi-
dence for the contributions from each kind of mechanore-
ceptor and their relevant importance.
B. Evidence for Muscle Spindles as Major
Kinesthetic Sensors
This is a summary of the evidence in support of the view
that muscle spindles are major kinesthetic sensors. If kines-
thetic information can come equally well from skin and
joint receptors, what is the evidence that at many joints it is
the muscle spindles that play the dominant role? The first
pointer and perhaps the intuitively most appealing one
comes from observations of persistent senses of position
and movement after joint replacement.
1. Joint replacement surgery
In patients who have had a total hip replacement involving
removal of all capsular and ligamentous components, both
position and movement sense remained intact (170). The
ability to detect passive movements and to duplicate posi-
tions of the hip was retained shortly after recovery from the
operation and persisted unaltered over the subsequent pe-
riod of several months. The authors concluded that the
ability to detect joint position did not depend on stimuli
arising in the joint capsule or on the surfaces of the hip joint.
This view was supported by observations on local anesthe-
sia within the knee joint (70). The evidence therefore sug-
gests that, at least at some joints, joint receptors do not play
a significant role in kinesthesia. Further studies of joint
replacement and ligament repair have yielded somewhat
variable results, some even suggesting the performance is
improved post surgery (for review, see Ref. 316). This vari-
ation probably depends on the local pathology, its duration,
associated changes in contralateral joints, as well as the
methods of testing for deficits.
2. Dorsal column lesions
If sensory receptors contribute to conscious sensation, their
afferents must project to the cerebral cortex. It is now well
established from studies on animals that afferents from
γ dynamic γ static Primary Ia Capsule Secondary II
Bag 1
Bag 2
ChainIntrafusal fibers
FIGURE 1. Diagrammatic representation of the mammalian muscle spindle. The intrafusal fibers include the
large nuclear bag 1 and bag 2 fibers together with the smaller nuclear chain fibers. Ends of the bag fibers
extend beyond the capsule while chain fibers lie within the limits of the capsule. Large, group Ia afferent fibers
terminate as primary endings, making spiral terminations around the nucleated portions of all three intrafusal
fiber types. Smaller, group II afferent fibers terminate as secondary endings, lying to one side of the primary
endings and supplying bag 2 and chain fibers. Gamma dynamic (
dynamic) fusimotor fibers innervate bag 1
fibers, while gamma static (
static) fusimotor fibers innervate bag 2 and chain fibers. [Rredrawn from Proske
Muscle fibers
Ib afferent
FIGURE 2. Diagrammatic representation of the mammalian Golgi tendon organ. The Group Ib axon pene-
trates the receptor capsule and branches, each branch terminating on a tendon strand that is attached to a
muscle fiber. A typical tendon organ has 10 or more muscle fibers attached to it, each fiber belonging to a
different motor unit. Contraction of a motor unit supplying a tendon organ stretches the tendon strand to which
its muscle fiber is attached, generating activity in the Ib axon. [Redrawn in part from Proske (302).]
1654 Physiol Rev VOL 92 OCTOBER 2012
muscle spindles and tendon organs, as well as those from
skin and joint project to the cortex (221, 247, 248, 278).
The pathway taken by skin and joint afferents is via the
dorsal (posterior) columns, gracile and cuneate nuclei, me-
dial lemniscus, and thalamus. Muscle afferents from the
forelimb also project centrally via the dorsal columns (336).
However, for the hindlimbs, muscle afferents, unlike skin
and joint afferents, leave the dorsal column in the upper
lumbar region, synapse in Clarke’s column, and project
centrally in the dorsolateral funiculus as the dorsal spino-
cerebellar tract (for more detail, see Refs. 27, 220).
There is some evidence that the central projection pathways
for skin and muscle afferents in humans are similar to that
in mammals. Posterior column section at a thoracic level in
humans leads to extensive loss of skin and joint sensation in
the legs, but the central projection pathway for muscle af-
ferents remains intact, indicated by preservation of the
senses of position and movement. The evidence comes from
a limited number of well-described patients with partial or
total sections of their dorsal columns (396; see also Ref.
338). In a patient with thoracic cord injury, including total
section of both dorsal columns, but only limited damage to
the underlying cord, sensation of joint movement and di-
rection of movement of the legs was close to normal. This
finding was corroborated in two other patients. Overall,
these data indicate that destruction of dorsal columns in the
thoracic region does not impair kinesthesia in the lower
limbs. By implication, skin and joint input is not essential
for qualitatively normal kinesthesia in the legs.
3. Muscle thixotropy
Discussions of muscle thixotropy have often been missing
from reviews of proprioception, yet the thixotropic behav-
ior of muscle spindles frequently leads to errors in the inter-
pretation of experimental data on kinesthesia. In addition,
intrafusal thixotropy provides important evidence in sup-
port of muscle spindles as the principal kinesthetic sensors.
Thixotropy is the dependence of a muscle’s passive mechan-
ical property on its previous history of contraction and
length changes. It arises from the presence of long-lasting
stable cross-bridges between actin and myosin in the sarco-
meres of resting muscle, including both extrafusal and in-
trafusal muscle fibers (308). The presence of these bridges is
indicated by a frictional stiffness of the muscle at the start of
a stretch, the short-range elastic component (SREC) (188).
Accompanying the SREC is a sustained rise in resting ten-
sion, called filament resting tension (FRT) by Hill. When a
muscle relaxes after a contraction, stable crossbridges form
in the fibers at that length to give them their SREC (219,
266). If the muscle is then shortened, the compressive forces
acting on sarcomeres which have been stiffened by the pres-
ence of the bridges are insufficient to detach the majority of
bridges so that the muscle fiber is unable to fully take up the
shorter length and falls slack. Such thixotropic states can
persist for long periods provided the muscle is left undis-
turbed (311).
At long lengths a resting muscle will lie taut, regardless of its
contraction history, whereas at short lengths it becomes
slack (158) and may even buckle (186). There are interme-
diate lengths where a muscle can be either slack or taut
depending on its history of contraction and length changes.
What is meant by slack is that the muscle is effectively
longer than the distance between its two points of attach-
ment, so it is obliged to lie slack. Slack can be introduced at
any particular length, by contracting the muscle at a longer
length, letting it relax for several seconds, and then short-
ening it to the original length. While the presence of slack in
the whole muscle may not always be apparent, the muscle
spindles with their compliant connections to adjacent ex-
trafusal fibers are particularly prone to it (for review, see
Ref. 310). The method used to condition muscle and its
effects on spindle discharge rates are shown in FIGURE 3.
For the soleus muscle of the anesthetized cat, in which slack
has been removed by a conditioning contraction, the spin-
dle resting discharge rate of 40 pulses/s falls to 10 pulses/s
when slack is introduced in the muscle (FIGURE 3). It means
that intrafusal tension falls, and therefore, the stress exerted
by the intrafusal fibers on the spindle sensory endings is
reduced. Thus spindle background discharge rate falls. If
now the muscle is contracted voluntarily, to include both
intrafusal and extrafusal contractions, the slack in spindles
will be removed and the background discharge rates will
rise. So simply by contracting and then relaxing a muscle,
without changing its length, background discharge rates in
spindles can be changed (164). This produces significant
changes in perceived limb position (FIGURE 4).
Thixotropic behavior is restricted to extrafusal muscle and
muscle spindles. Its effects are prominent at short and inter-
mediate muscle lengths, and it fades at long lengths, as a
result of spontaneous detachment of stable crossbridges in
the presence of the high passive tension (164). Because most
tendon organs have a high threshold to passive stretch, the
muscle has to be stretched to long lengths for tendon organs
to generate maintained activity. At those lengths thixo-
tropic effects are small, or absent (305). Skin and joint
afferents, because of the viscoelasticity of the underlying
tissue, will show some hysteresis in their responses to
lengthening and shortening movements. However, these
changes are independent of muscle contraction, and they do
not persist after the movement has stopped, so they are not
truly thixotropic, that is, dependent on the previous history.
If elbow flexors of one arm are conditioned to lie either
slack or taut, this can lead to perceived changes in arm
position of 5° (165). If the two arms are conditioned in
opposite ways so that elbow flexors on one side lie slack
while on the other side they are taut, it can lead to mean
1655Physiol Rev VOL 92 OCTOBER 2012
matching errors of 20°, representing a quarter of the total
range of movement of the arm (5). Despite such large dif-
ferences in arm position, on interrogation, the blindfolded
subjects insist that their arms are accurately aligned. The
fact that conditioning of the two arms in this way leads to
such large errors suggests that the brain is listening to both
arms and most probably is responding to the difference
signal from them (412).
Since much of the early data were acquired without knowl-
edge of thixotropy, the complication this introduces was over-
looked. It has meant that sometimes uncertainty has remained
about the significance of particular results. The frequently
cited report that position sense is more accurate with active
rather than passive movements (280) probably has as its basis
muscle thixotropy. Thixotropy is a property of passive muscle
so its influence on proprioception is especially important for
position and movement sense in the passive limb. However,
thixotropic influences can persist with voluntary contractions
of 5–10% of maximum (168, 200).
The failure of many experiments concerned with limb po-
sition sense to place the test muscle in a defined state, has led
to uncertainties in the interpretation of some of the data. So,
for example, in a recent study of the perceived position of a
passively moved arm (131), no conditioning contractions of
elbow muscles were carried out. It was assumed that as
subjects remained passive throughout the 0.5-h testing ses-
sion, no change in muscle spindle activity was likely over
this time. Nevertheless, any unintended movements would
risk changing spindle sensitivity and alter the outcome. In
addition, the thixotropic state of the subject’s arm would
depend on what the subject had been doing immediately
before the experimental trial, risking bigger intersubject dif-
ferences during the subsequent measurements. Similarly, if
the aim of an experiment is to compare position sense in an
unloaded limb, with that when the limb is supporting a
load, the voluntary contraction used to support the load
will remove any preexisting slack in spindles during the
unloaded measurements and lead to position errors that
have nothing to do with central effects of the load (419).
Thixotropy-dependent errors in position sense are typically
present only in the passive limb, so the motor output gen-
erated by the CNS is not directly involved. The direction
and distribution of the errors strongly imply that it is the
A Flexion conditioning C Extension conditioning
Test angle
40 imp/s
5 mm
5 sec
Test angle
40 imp/s
5 mm
5 sec
FIGURE 3. The method of conditioning muscle to put it into a defined state. In Aand B, at the top is shown
a diagrammatic human forearm with one flexor and one extensor muscle drawn in. During flexion conditioning
(A, red) the forearm is flexed, elbow flexors are briefly contracted, and once the muscle has relaxed, the
passive forearm is placed at the test angle. During extension conditioning (C, blue) the forearm is extended,
elbow extensors are contracted, and on relaxation the relaxed forearm is placed at the test angle. Traces B and
Dbelow the forearm diagrams give schematic representations of firing rates of muscle spindles in the soleus
muscle of the anesthetized cat after it had undergone two kinds of muscle conditioning comparable to those
used at the human elbow joint. A conditioning contraction of soleus when it was shortened by 5 mm (B, red
trace), on return to the test length led to a spindle resting discharge of 40 impulses/s. A conditioning
contraction after soleus had been stretched by 5 mm (D, blue trace) led on return to the test length of a spindle
rate of 10 impulses/s. [Redrawn in part from Wood et al. (425).]
1656 Physiol Rev VOL 92 OCTOBER 2012
resting activity of muscle spindles which generates a signal
of limb position and that in a matching task the parameter
measured by the brain is the difference in signals between
the limbs. Such behavior raises the question of what is the
reference point used to determine the starting position of
the limb. Another issue is what happens in a contracting
muscle (e.g., Ref. 6). During a voluntary contraction,
some spindle impulses will be generated by fusimotor
activity (e.g., Refs. 3, 47, 386, 416). It remains uncertain
how position sense is generated under these conditions
(e.g., Refs. 93, 242, 401). A speculative proposal about
how the brain might distinguish between spindle im-
pulses generated by stretch and fusimotor activity is put
forward below (FIGURE 16).
Thixotropy provides the only known method for manipu-
lating position sense in the passive limb without the use of
external stimuli such as vibration. No attempts have been
made so far to grade thixotropic conditioning to see
whether this produces progressive changes in position
sense. What change in the ensemble rate of spindle dis-
charge in a muscle is needed per degree of position sense
change? Here there are potentially confounding effects
from extrafusal fibers on muscle spindle discharge (e.g.,
Ref. 44). In addition, thixotropy represents a useful tool for
the study of subjects who are suspected to have propriocep-
tive disturbances, such as Parkinsonian patients (240, 428)
and the elderly (195).
4. Muscle vibration
Arguably the most influential evidence in support of muscle
spindles as the principal kinesthetic receptors is the illusion
of limb movement and displaced position produced by vi-
bration over the tendon or muscle (111, 159). These obser-
vations, more than any others, swung prevailing opinion in
the 1970s away from joint receptors, in favor of muscle
receptors. Vibration of the tendon of biceps or triceps
brachii at 100 Hz produced an illusion of movement and of
changed position at the elbow in a direction that would
elongate the vibrated muscle. Vibration over the elbow joint
produced no illusion.
Simultaneous vibration of antagonist muscles at similar fre-
quencies and amplitudes produced no movement sensation
at all, suggesting that the difference in signal between ago-
nists and antagonists is what is perceived centrally (325).
Further details of the frequency dependence of the vibration
illusion were provided by McCloskey (243). He showed
that the velocity of the vibration illusion slowed in direct
proportion to the load being supported by the vibrated
muscle. When the muscle generated half-maximal contrac-
tions or more, vibration no longer produced any illusion, a
result that has recently been confirmed (13). Furthermore,
vibration at lower frequencies with larger amplitudes pro-
duced illusions of displaced position only.
The studies with vibration have since been repeated and
extended many times under both passive conditions and
during voluntary movements (e.g., Refs. 57, 82, 217, 391).
Importantly, microneurography has revealed that the pri-
mary endings of spindles are largely responsible for the
illusion and that the vibration frequency for an optimal
response in human muscles is 80 Hz (e.g., Refs. 205, 332,
333). While most observations have confirmed that it is
principally the primary endings of spindles that are respon-
sive to vibration in the relaxed muscle, it is well known that
some Golgi tendon organs respond to tendon or muscle
vibration (46), particularly if the subject is slightly tensing
their muscles (113). This is a reflection of the fact that tendon
organs are tension sensors and relatively insensitive to me-
chanical stimuli, unless the tendon strands on which they are
sited are under tension (FIGURE 2).
The vibration illusion can be demonstrated in both arms. In
a position matching task, vibration of elbow flexors or ex-
tensors of the indicator arm produced errors in a direction
opposite to that from vibrating the same muscles in the
reference arm (412). This is to be expected in a task where,
to achieve a match, the vibrated indicator must move in a
direction that reduces the activity in its muscle spindles
sufficiently to bring their responses closer to those in the
nonvibrated reference arm. Vision of the indicator arm (or
Position sense: arm supported
Flexion conditioning
Extension conditioning
Position error (°)
Trial number
FIGURE 4. The effects of muscle conditioning on human position
sense at the forearm measured in a two-arm matching task. Shown
are 10 pairs of successive trials from one subject. Following a
conditioning contraction of the reference arm, the relaxed arm was
placed by the experimenter on a support at the test angle. To
indicate its location, the subject then moved their other arm to a
matching position. Errors were calculated as the difference in posi-
tion of the two arms and were displayed using the convention, errors
in the direction of extension as positive, errors in the direction of
flexion as negative. Symbols in red, position errors after flexion
conditioning; symbols in blue, errors after extension conditioning
(see FIGURE 3). Trials were alternated between the two forms of
conditioning, and each pair of measurements has been joined by a
dotted line. The dashed line indicates zero error. Flexion conditioning
leads errors to lie systematically in the direction of extension relative
to the errors after extension conditioning. [Data from Winter et al.
1657Physiol Rev VOL 92 OCTOBER 2012
its mirror image) can reduce the size of the vibration illusion
(199) (see also Refs. 193, 218). Furthermore, the perceived
speed of extension of the reference arm from vibration of its
elbow flexors can be altered by flexion or extension move-
ments of the indicator arm (199) (see also Refs. 82, 359).
These experiments demonstrate that during tasks such as
placement of the two hands close together, proprioceptive
inputs from both arms are used. This is presumably part of
a motor control strategy to allow use of the two hands as a
single instrument in skilled tasks (199).
The vibration illusion can be manipulated by means of thixot-
ropy. In animal experiments, deliberately introducing slack in
muscle spindles by contraction of the muscle at a long length
and then shortening it left all spindles insensitive to vibration
applied longitudinally to the tendon (306). The effect of vibra-
tion on human position sense can be abolished if sufficient
slack is introduced in the muscle (412). On the other hand, the
speed of the movement illusion generated by vibration can be
doubled by removing slack in the muscle with a conditioning
contraction (160). These and other observations on thixot-
ropy (e.g., Refs. 104, 417) point strongly to muscle spindles as
the prime candidates for providing position and velocity infor-
mation in proprioception.
C. The Senses of Position and Movement
Originally the senses of position and movement were con-
sidered a single sense: kinesthesis (18). Part of the reason for
combining them is that both share inputs from the same
receptor, the primary endings of muscle spindles. Second,
muscle vibration elicits illusions of both movement and dis-
placed position (159). Similar illusions can be produced
with electrical stimulation of peripheral nerve, which prob-
ably excited spindle primary endings (135). An example is
given in FIGURE 5. Primary endings respond both to the
length change and rate of length change of the muscle (241).
During stretches at increasing velocities, the response of the
primary ending increases in direct proportion to the rate of
length change. Movement sense is therefore signaled by the
velocity component of the primary ending’s response to the
length change. As the muscle is stretched to longer lengths,
the background rate of spindle discharge increases in direct
proportion to the length change. Position sense can there-
fore be envisaged as signaled by the mean rate of back-
ground discharge in muscle spindles, including that gener-
ated by both primary and secondary endings (see also sect.
IIB3). Secondary endings are also likely to contribute to
position sense (243). When the frequency of vibration was
reduced from 100 to 20 Hz, the sensation of movement
blended into one of position only. McCloskey (243) argued
that the two senses were generated by separate lines of input
and that position sense was not derived from an integration
of the velocity signal (see also Ref. 360).
Other evidence supports the existence of two senses, both
generated by muscle spindles. It is possible to manipulate
the sense of position using thixotropy. The background rate
of both primary and secondary endings can be increased by
means of a conditioning contraction (165; see also Ref.
417). This led to a significant change in the perceived posi-
tion of the limb, without any accompanying sensation of
FIGURE 5. Photographs of illusory positions of the hand during trains of electrical stimulation of the ulnar
nerve at the wrist using a stimulus strength below motor threshold. Top panels: rest positions before
commencement of stimulation. Bottom panels: posture adopted during stimulation. Left: illusory positions
adopted for all fingers, perceived flexion of the interphalangeal joints and extension at the metacarpophalangeal
joints. Right: for the little finger, perceived flexion of the interphalangeal joints and extension at the metacar-
pophalangeal joint. For further details see text. [Photographs based on Gandevia (135).]
1658 Physiol Rev VOL 92 OCTOBER 2012
movement (165). In an experiment on the effect of muscle
fatigue on position and movement sense (see sect. V), fa-
tiguing exercise of elbow flexors produced errors in the
perceived position of the exercised limb but without a
change in passive movement sense (8). The result again
suggested two senses, each responding differently to a po-
tential disturbance.
In thinking about the two senses, it is easy to see how the
spindle firing rate during a length change might be con-
verted into a movement sensation. It is less clear how posi-
tion sense is generated. Some spindles generate background
activity at all muscle lengths, no matter how short the mus-
cle; others fall silent at short lengths (167). Perhaps such
differences become submerged within the population sig-
nal. How are the limits of the range of limb position estab-
lished? Vibration and electrical stimulation of muscle spin-
dles can produce sensations of position at anatomically im-
possible joint angles (85, 135, 218). This indicates that the
brain can be misled to perceive an anatomically impossible
position. Perhaps under normal circumstances it is the joint
receptors, acting as limit detectors, that help to define the
limits of limb movement (see sect. IID).
In experiments measuring limb position sense using a bilat-
eral matching task, typically subjects are consistent in their
matching performance, yet they can make substantial bias
errors. Depending on how the reference arm has been con-
ditioned, they often perceive its position as more extended
or more flexed than it really is, and this error can persist in
repeated trials on different days. So, for example, in a study
of the effects of exercise on position sense, control errors
after conditioning elbow flexors lay 2° or more in the direc-
tion of flexion (7). Such observations suggest that the cali-
bration of absolute arm position is not very accurate, yet
subjects are aware of the position of one arm relative to the
A comment on the processing of position and movement
information is provided by studies of dorsal spinocerebellar
tract (DSCT) cells in the anesthetized cat (29). Cells are
described whose discharges are modulated by both position
and movement of the foot. The two inputs summate non-
linearly, and the amplitude of the modulation depends on
the position of the foot. In other words, a neuronal process-
ing network is available that can simultaneously process
position and movement information, making it unnecessary
to postulate separate pathways. Furthermore, this conclu-
sion is consistent with observations made on motor cortical
neurones in monkeys (390).
D. Joint Receptors
Joint rotation will stretch the joint capsule on one side, as
well as the overlying skin, while tissue on the other side is
unloaded. Mechanoreceptors signaling tissue stress within
the capsule are Ruffini-like endings, comparable to the cu-
taneous SA2 endings, while Paciniform corpuscles respond
to local compression (for review, see Ref. 169).
In their study of joint receptor properties in the posterior
articular nerve of the knee of the cat, Burgess and Clark (42)
found 70% of the receptors were slowly adapting. The
vast majority of these responded only to marked flexion and
marked extension of the joint, that is, they responded at
both limits of the range of joint movement, making their
position signal ambiguous. A true position sensor would
change its discharge rate monotonically with joint angle in
one direction only.
Within the rapidly adapting group of receptors, some were
Pacinian corpuscle-like, responding briefly to joint move-
ment in any direction. Other phasic receptors produced a
rapidly adapting response at most joint angles but gave a
sustained response to maximal extension of the joint if this
was combined with a twisting force (42).
There are limited data on joint receptor properties in human
subjects. Recordings were made from joint afferents supply-
ing the metacarpophalangeal joint and the interphalangeal
joints of the digits. These afferents responded to local pres-
sure over the joint capsule (45) and generated a slowly
adapting discharge at extreme angular displacements. They
commonly discharged to more than one axis of movement
and at both ends of an angular range. Only 1 of 18 joint
afferents responded to passive joint movement across the
physiological range. A limited number of joint receptors
responding to joint flexion (but not extension) were found
in the superficial radial nerve (99). Intraneural microstimu-
lation of some joint afferents generated small movement
sensations at the joint in response to trains of electrical
stimuli (234). The perceived movements corresponded to
the pattern of discharge observed when moving the passive
joint in a similar way. Thus the brain can obtain some
information about the movement of a joint from these kinds
of receptors, although their signaling capacity was rather
limited (234). There are no data for stimulation of single
joint afferents at joints outside the hand.
Concerning the psychophysical evidence for a role for
joint receptors in kinesthesia, the work of Gandevia and
McCloskey (138, 143) showed that at the distal interpha-
langeal joint of the middle finger, in the absence of muscle
afferent signals, subjects were poor at detecting small
slowly applied angular displacements. When the experi-
ment was done the other way around, with skin and joint
input blocked but muscle afferent input intact, detection at
low angular velocities was lower than with the full propri-
oceptive machinery intact. In other words, contributions
from all sources were required for full proprioceptive acu-
ity. When only skin and joint input were available, position
sense deteriorated slightly if the joint capsule was infiltrated
1659Physiol Rev VOL 92 OCTOBER 2012
with anesthetic, directly pointing to a role for joint recep-
tors (116), although this result could not be reproduced by
others (69). In similar experiments on the proximal inter-
phalangeal joint, block of the digital nerves did not impair
position sense (340). When the experiments on the hand
were repeated over a wider range of joint angles, position
errors occurred (117), particularly near the extremes of the
range of joint movement. This finding led Ferrell and Smith
(118) to conclude that joint receptors provided positional
information principally at the extremes of the normal range
of joint movements, perhaps acting as “limit detectors.” In
contrast, muscle spindles increased their discharge mono-
tonically across the full angular range with some differences
between synergists (43), but they did not behave like limit
detectors. Hence, it is not surprising that illusions attributed
to them generate perceptions of anatomically impossible
joint angles (e.g., Refs. 85, 135, 159, 218).
To conclude, there is clear evidence in support of a role for
joint receptors in signaling joint movement, but it appears
that they are unable to signal movement direction or joint
position within the normal range. This issue has recently
been brought up again (131). The authors propose that as
joints approach the limits of their movement range, joint
receptor information may bias perception of joint angles. In
considering proprioception at the finger joints, it should be
kept in mind that for movements at these joints muscle afferent
signals are coming both from short intrinsic muscles in the
hand as well as from proximal muscles in the forearm that are
connected to the fingers by long, compliant tendons. Under
these potentially ambiguous circumstances, it may be that skin
and joint input is more important than muscle afferent input.
At more proximal joints it appears that muscle afferents pro-
vide the major proprioceptive signal (68).
E. Skin Receptors
Joint rotation causes skin on one side of the joint to be
stretched and to be slackened or even folded on the other
side. Such deformations will stimulate skin mechanorecep-
tors. There are four kinds of specialized mechanoreceptors
in glabrous skin: rapidly adapting type I, the Meissner cor-
puscles; rapidly adapting type II, Pacinian corpuscles; slowly
adapting type I, Merkel endings; and slowly adapting type II,
Ruffini endings (e.g., Ref. 203). While all four receptor types
are likely to contribute to movement sensations, slowly adapt-
ing type II, the skin stretch receptors, are potentially able to
signal limb position (e.g., Refs. 64, 100, 101).
Illusions of finger joint movement are produced by strain of
the adjacent skin, without any actual movement of the joint
(78, 79, 103). In a recent study of the contribution of skin
receptors to kinesthesia (80), skin input to position and
movement sense was examined at the index finger, the el-
bow, and the knee. An example is shown for the illusions
generated by muscle vibration and skin stretch at the elbow
(FIGURE 6). Skin receptors were activated by skin stretch using
adhesive tape, and muscle receptors were activated by vibra-
tion. Graded skin stretch in a direction in line with muscle
stretch applied during vibration significantly increased per-
ceived movement sensation above that from skin stretch alone
or vibration alone. This was not just a matter of skin input
facilitating the muscle input because just skin stretch alone
commonly produced illusory movements. Therefore, input
from skin stretch is able to contribute to kinesthesia in its own
right. The sensitivity of human skin stretch receptors when
expressed as impulses per degree of joint motion is similar to
that of muscle spindle afferents (101, 171).
Signals from skin receptors can also have an occluding ac-
tion on kinesthetic sensations. Stimulating rapidly adapting
receptors, presumably Pacinian type Is, with high-fre-
quency vibration at very low amplitude interferes with
movement detection (406, 408). Similarly, focal pain in-
duced either in the skin around a joint or in the muscles
which move it impairs movement detection at that joint
(405). The neural basis for this is presumably convergence
between cutaneous and muscle afferents at spinal cord and
thalamic levels along the projection paths to the cortex.
A situation where the contribution to kinesthesia from skin
receptors becomes indispensable is in the skin adjacent to
the fingertips. The muscles that move the fingers lie in the
forearm and hand and their tendons must cross three or
more joints. Here signals from muscle spindles are poten-
tially ambiguous (see also Refs. 347, 373). The presence of
skin receptors adjacent to each finger joint allows them to
provide joint-specific information (e.g., Refs. 79, 103).
However, the skin input itself can be ambiguous. Slowly
adapting type II afferents from hairy skin of the back of the
hand show sustained responses to flexion of the finger joints at
intermediate joint angles (102). However, most of these affer-
ents responded to movements of more than one finger, and
responses to flexion or extension of one finger depended on the
posture adopted by another finger. Hence, it is only by consid-
ering the spatial array of particular inputs that the sizes and
directions of imposed movements can be computed.
Electrical stimulation of single identified cutaneous afferents
has provided further information about the contribution of
different inputs from the hand, but with only limited insight
into proprioceptive coding (234, 276, 387). Rapidly adapting
afferents, when stimulated, evoke a sense of vibration or tap-
ping, but there is no sense of movement. Stimulation of single
afferents of slowly adapting type I receptors produces sensa-
tions of local indentation or pressure, while stimulation of
slowly adapting type II receptors leads to no sensation at all.
Furthermore, in a study which also assessed muscle afferents
(234), stimulation of single spindle afferents produced no sen-
sation. Thus, for the digits and probably all other joints, it is
necessary to stimulate a population of stretch receptors to
generate detectable sensations (see sect. IIIC).
1660 Physiol Rev VOL 92 OCTOBER 2012
In a population study of receptors in the skin over the ankle
joint, responses were recorded to movements in different
directions (2). The majority of movement sensitive afferents
were from slowly adapting type II or rapidly adapting type
II receptors. The response patterns of the skin afferents,
when considered in terms of a population vector model,
matched, in their directional sensitivity, those of muscle
receptors in the underlying muscles, making both kinds of
receptors suitable candidates for generating kinesthetic sen-
sations. In conclusion, skin afferents play a significant role
in kinesthesia, and they are likely to contribute to move-
ment sensation at most joints. However, their contribution
to position sense at the more proximal joints is likely to be
less important than the input from muscle spindles. Perhaps
a special case should be made for skin receptors in signaling
facial expressions, since facial muscles are believed not to
contain any muscle spindles (385).
A. Spindles as Length and Velocity Sensors
An important question is what aspects of a movement are
signaled by spindles? The work carried out largely in the
1960s and 1970s established that the primary endings of
spindles respond to ramp-and-hold stretches with a dis-
charge rate that is proportional to the size of the stretch and
to the rate of stretch. Secondary endings of spindles have a
lower dynamic sensitivity, and their response is largely pro-
portional to the size of the stretch. This led to the view that
primary endings were muscle length and velocity sensors
and secondary endings were length sensors (for a detailed
account, see, e.g., Ref. 241).
These findings have led to the widely accepted view that
spindles provide information about length changes in mus-
cles, and this is represented as changes in joint angles. The
question is then posed, Does the brain compile muscle
movement and joint angular information from each muscle
group to compute the sizes and directions of movements of
the whole limb? When a vibrator is strapped to elbow flex-
ors of one arm, vibration at 100 Hz produces an illusion of
forearm movement into extension, as signaled by placement
of the other arm. Vibration of elbow extensors produces
movement illusions in the opposite direction, elbow flexion
(159). So vibration of one muscle group is able to generate
a muscle-specific sensation. It means that the brain has ac-
cess to information specific to individual muscles. While
that is so, it does not preclude the possibility that input from
the whole limb is used to calculate movement related pro-
prioceptive signals.
elbow flexion
vibration weak
Stimuli on right
skin stretch
perceived movements
vibration only
vibration and strong
skin stretch
FIGURE 6. Illusory motion at the right elbow produced by muscle vibration, skin stretch, and combined
vibration and stretch. A: stimuli delivered to the right arm, perceived as movements and indicated by the left
arm. B: mean amplitude of perceived movements for all trials (vibration, n15 cycles; all others, n10
cycles). Asterisk indicates significant differences from vibration alone. Diagrams in Cshow the position of
the right elbow during the experiment (gray silhouette) and the average perceived position when vibration
was applied by itself (top) and simultaneously with strong skin stretch near the elbow (bottom). [From
Collins et al. (80).]
1661Physiol Rev VOL 92 OCTOBER 2012
B. Movement Detection Thresholds
The dynamic sensitivity of muscle spindles and skin stretch
receptors provides the basis for our ability to detect small
movements of our limbs. A simple test of proprioception,
widely used in the clinic, is to impose movements at a joint,
usually a finger joint or big toe joint, and to ask the subject
to declare when they feel the movement and to indicate its
direction. It is an old observation that movement detection
thresholds for proximal joints are lower than for more dis-
tal joints (157). This point has been reexamined more re-
cently (178).
Measurements are imposed on the passive limb by a servo-
motor (FIGURE 7). Threshold is usually measured as the
movement amplitude required for generating 70% correct
responses. Here subjects are asked to indicate both that a
movement has occurred and its direction. Detection thresh-
old at the forearm depended on the velocity of movement.
At a speed of 1°/s threshold was 0.2°, while at 0.1°/s it was
eight times higher, at 1.6° (178). Lower thresholds can be
achieved if the muscle is appropriately conditioned before-
hand (300, 421). Thresholds at the finger joints were several
times higher than at the elbow and shoulder joints. In think-
ing about their data, and assuming that muscle spindles
were responsible for the threshold sensations, Hall and Mc-
Closkey (178) proposed that what mattered was not the
angular range through which a joint was moved but the
proportional length change the movement imposed on the
muscles that operated at that joint. To check the point,
measurements were made on muscle fascicles in human ca-
davers, and changes in fascicle length per degree of joint
rotation were calculated. When these values were used to
compare detection thresholds, differences in thresholds be-
tween the finger, elbow, and shoulder largely disappeared.
A similar result was achieved for detection thresholds at
joints in the leg (317, 319). This outcome was considered
evidence in support of muscle spindles as the principal pro-
prioceptor responsible for the detection of movements.
If muscle spindles signal changes in fascicle length, this
raises a problem. Fascicle lengths are very different for mus-
cles at different joints (285), yet muscle spindle lengths are
about the same from one muscle to another (32). As fasci-
cles in distal muscles are shorter than in more proximal
muscles, it means that in distal muscles the spindles run the
full length of the fascicle while in proximal muscles they
span only a fraction of the fascicle length. However, as
shown above, this difference in arrangement is not accom-
panied by any difference in movement detection threshold
between proximal and distal muscles. How can a spindle,
much shorter than the adjacent muscle fascicle, accurately
signal length changes in the fascicle? To solve this problem,
it is first necessary to assume that a muscle fascicle behaves
0.2 deg/sec
Threshold (normalized)
0.0 Relaxed Co-contract
FIGURE 7. Measurement of movement detection thresholds. A: blindfolded subjects were required to
indicate the direction of small movements (0.2°/s) applied to the right forearm with a servomotor. Detection
thresholds were measured for elbow extension and flexion movements under relaxed and cocontraction
conditions (15% MVC cocontraction of elbow flexors and extensors, monitored as EMG). B: average thresholds
measured for 7 subjects. There were no differences between thresholds for flexion and extension movements,
and values have been pooled. Thresholds have been normalized with respect to the average threshold
measured for the relaxed condition. Thresholds measured during cocontractions of elbow muscles (blue bar)
were significantly higher than when the arm was relaxed (orange bar). [Redrawn from Wise et al. (420).]
1662 Physiol Rev VOL 92 OCTOBER 2012
in a mechanically uniform manner, that is, an imposed
stretch is distributed uniformly along all parts of the fasci-
cle. If in proximal muscles the ends of the spindle make
lateral attachments to the perimysium of the fascicle, the
spindle will in effect signal length changes in only a part of
the fascicle. With such an arrangement, the absolute length
change a longer fascicle must undergo to produce the same
spindle signal will have to be greater than for a shorter
fascicle. Such a trend is evident in the relationship between
fascicle length and the length change required to reach
movement detection threshold (313).
Detection thresholds are higher for slow movements. If the
movement is made slowly enough, it will go undetected, but
the subject will eventually realize that the limb being moved
is no longer where it was previously but be unaware of how
it got to its new position (68). This is the sense of position,
and its threshold is 2.5° at the metacarpophalangeal joint
(377). The much lower detection threshold for faster move-
ments presumably reflects the dynamic sensitivity of spindle
primary endings and skin stretch receptors.
Since in everyday life limb movements are invariably ac-
companied by muscle contraction, it raises the question of
movement detection thresholds during a contraction. This
has been measured at the elbow joint during flexion con-
tractions (378) or isometric cocontractions (420). At the
lower velocity end of the range of imposed movements,
Taylor and McCloskey (378) have reported a 10-fold fall in
detection threshold during contraction, compared with
threshold measured with the passive arm. A similar result
has been reported for finger movements when the input
from joint and skin had been blocked (39, 143). In contrast,
Wise et al. (420) obtained a fivefold increase in detection
threshold at the elbow when subjects were generating a
15–20% of maximum cocontraction (FIGURE 7). One pos-
sible explanation for this difference in results is the influence
of thixotropy on threshold measurements in the passive
limb (421). During a voluntary contraction there is coacti-
vation of fusimotor neurons (see sect. IIB3). So, does
fusimotor coactivation increase or decrease the ability of
spindles to detect movements? Experiments on animals
have yielded the unexpected result that responses of pri-
mary endings of spindles to small movements are larger in
the passive spindle than when the spindle’s fusimotor sup-
ply was stimulated (422). After a conditioning contraction,
responses of passive spindles to a stretch were larger than
when the stretch was applied during static fusimotor stim-
ulation, dynamic fusimotor stimulation, or their combina-
tion. Thus the passive spindle is more sensitive to move-
ments than when its intrafusal fibers are contracting. This
finding is at odds with the common view that propriocep-
tion is more accurate under active than passive conditions.
This misconception has probably arisen due to thixotropic
effects (see sect. IIB3).
C. Signaling Properties
of Spindle Ensembles
Does the brain access information from individual spindles
or does it rely on the population response from the whole
muscle? Here pertinent observations have been made using
microneurography and stimulation of single afferents in
human subjects. Stimulation of single muscle afferents in-
nervating intrinsic hand muscles did not produce any sen-
sations (234). However, electrical stimulation of a pre-
sumed population of muscle afferents did produce the ex-
pected position and movement illusions (135). Hence, for
hand muscles, generation of a sensation required input from
more than one muscle afferent. Since each spindle has its
own location in the muscle and is therefore exposed to a
unique set of mechanical conditions during muscle move-
ments, it implied that these differences were combined and
the signal arising from the muscle, relevant to propriocep-
tion, was the population response of its afferents.
An answer to the question of how direction of a movement
may be signaled, while incorporating differences between
individual neurons, has been proposed by Georgopoulos
and colleagues (e.g., Refs. 151, 152). They recorded the
discharges of movement-sensitive neurons in the motor cor-
tex of conscious monkeys during arm reaching movements.
Most neurons discharged maximally for movements in one
preferred direction. However, the directional sensitivity of a
particular cell was broadly tuned, with weaker responses
for movements not in the preferred direction. Tuning curves
of different cells overlapped. It was proposed that each cell
made a vector contribution to the population response, with
the size of the vector depending on the relation between the
imposed movement and the cell’s preferred direction of dis-
charge. Vector contributions from individual cells summed to
give the population vector which corresponded closely to the
direction of the movement made by the monkey (345). This
kind of proposition is relevant to how populations of spindles
might signal limb movements (see sect. IIID).
A further clue about the kinematic information provided by
spindles and how it is processed centrally has come from
recordings of second-order neurons in the central projec-
tion pathway for spindles (30). In anesthetized cats, the
discharges of some DSCT cells were consistent with a limb-
based rather than a joint-based reference frame. The au-
thors demonstrated this by fixing a rigid splint between the
thigh and shank of one leg, thereby reducing movement
about the knee. In the splinted leg, about half of the DSCT
neurons continued to signal the limb end-point representa-
tion, that is, position of the foot, as distinct from the specific
limb geometry associated with the end-point. The findings
suggested a wide convergence of muscle afferent input from
the hindlimb to allow the DSCT circuitry to compute an
estimate of foot position that was independent of overall
limb geometry. This could be achieved by combining and
1663Physiol Rev VOL 92 OCTOBER 2012
redistributing the relative weights of the inputs from the
different limb segments.
In summary, we are now confronted by the realization that
inputs are combined, not just of individual afferent re-
sponses from one muscle but of pooled responses from com-
binations of muscles acting at different joints. This concept
is not restricted to muscle spindles, as it is likely to apply to
other proprioceptive inputs as well (e.g., Refs. 2, 98, 101).
Finally, recent observations of activity in cortical motoneu-
rons during multijoint movements of the arm have shown
that firing rates of individual neurons are modulated by the
kinematics of multiple joints and that only a small pool of
motoneurons is necessary for the generation of a wide range
of movements (390). Such a conclusion adds support to the
idea of limb-wide convergence of afferent information and
its processing at spinal and cortical levels.
D. Generating Predicted Sensations
With Vibration
The concepts of population coding and vector summation
for motor cortical output (e.g., Refs. 151, 152) were used to
model the generation of proprioceptive sensations (e.g.,
Refs. 4, 21, 332, 333, 381). Roll and colleagues used mi-
croneurography and muscle vibration to study the relevant
features of a movement signaled by muscle spindles. The
first point they made was that, based on the sensory percept
reported by subjects, integration of afferent information
from more than one muscle was possible. Vibration of ag-
onist and antagonist muscles at similar frequencies and am-
plitudes produced little or no movement sensation (e.g.,
Refs. 153, 325). Similarly vibration of hand muscles at two
different sites when carried out separately produced illusory
hand movements in orthogonal directions. Simultaneous
stimulation of the two sites produced a single percept in a
direction with an oblique trajectory (330).
Population vector coding is based on the requirement that
the directional sensitivities differ from one receptor to an-
other in the muscle and that these properties overlap be-
tween receptors. Such requirements were met by muscle
spindles studied in a group of ankle muscles (21). Example
records are shown in FIGURE 8.FIGURE 8Ashows microneu-
rographic recordings of responses of an identified primary
spindle ending of the extensor digitorum longus muscle during
ankle movements in two dimensions, vertical and horizontal.
Responses of the spindle are shown during movements in eight
directions. Such responses, recorded for each of four different
ankle muscles, were used to calculate the preferred sensory
directions for the muscles (FIGURE 8B). These corresponded
well with the perceived directions of movements during vibra-
tion of individual muscles (FIGURE 8C).
An estimate of the population responses of spindle afferents
for each of a number of muscles involved in a movement was
used to calculate the required patterns of vibration necessary
to achieve such a response. Simultaneous vibration of various
muscles, using the computed vibration patterns, allowed the
generation of sensations in two (e.g., Ref. 331) and three di-
mensions (382). Subjects correctly reported illusory move-
ment patterns including images of graphic symbols and com-
plex three-dimensional figures.
To summarize, two important principles for proprioception
emerge from this work. One is that it is not the activity of
individual afferents but the combined response of the pop-
ulation of afferents that provides useful information about
a movement. Second, proprioceptive signals generated dur-
ing a movement by muscle afferents from a group of mus-
cles are typically not interpreted in terms of muscle length
or joint angle percepts, but in terms of the dynamic dis-
placement of the limb’s end-point. Similar conclusions have
been reached by others using different techniques (e.g.,
Refs. 297, 370). Studies on the hindlimb of the cat indicate
that the input from as few as 10 proprioceptive afferents
can provide reasonably accurate information about the po-
sition of the limb (370). Issues that remain to be resolved are
how muscle receptor responses combine to signal limb po-
sition as against limb movement, and how movement and
position information is handled during active movements
involving muscle contraction when the fusimotor system is
engaged. Here an important issue is the process of central
integration of afferent signals with signals of motor com-
mand or corollary discharge.
E. Motor Equivalence
It is a common experience that one’s signature in a book is
identifiably similar to the same signature written on a large
blackboard. In terms of the direct motor output for the two
tasks the requirements are very different, involving different
muscles and forces. Yet the signatures closely resemble one
another. This is referred to as the motor equivalence prin-
ciple. It implies that actions are encoded within the CNS in
terms that are more general than the commands to specific
muscles. Details of motor implementation such as stroke
speed and size are left unspecified until the effector is known
(418). The principle of motor equivalence can be applied in
proprioception to the signaled property for movements,
that is, movement of the limb segment end-point (30).
Some psychophysical observations are consistent with such a
view. In a two-arm position matching task, subjects were more
accurate when they held their arms by their side, hands in front
(160) and subjects aligned their arms by matching hand posi-
tion rather than elbow angle. Subjects are better at matching
arm orientation than elbow angle (368, 426). When subjects
were asked to indicate the position of their unseen forearm,
they were less accurate in determining elbow angle than when
asked to locate fingertip position. This result was interpreted
in terms of optimizing estimates of limb end-point position
1664 Physiol Rev VOL 92 OCTOBER 2012
(131). Similarly, observations on monkeys suggest that the
desired end-point of a movement is one variable coded by the
motor cortex (162, 293).
There are at least two reasons for including body schemas
and images in a discussion of proprioception. First, while
proprioceptors provide information about position and
movement of the limb, they are unable to signal the length
of limb segments and therefore the absolute location of the
limb in space. So when we vibrate biceps brachii, it always
leads to a sensation of elbow extension, but extension from
where? There must be a reference point to which the vibra-
tion evoked proprioceptive information is referred. This
requires knowledge of a body map. Second, there is the
issue of body ownership. When we move an unseen limb,
while proprioceptive feedback tells us about the movement,
we need to be able to identify the moving limb as our own.
So the body image tells us about the shape and location of
different parts of our body and allows us to distinguish be-
tween what parts are our own and, in some circumstances,
what parts are foreign. The subject of body images is a large
one, and we have taken a somewhat narrow view, focusing on
proprioceptors and their central actions. For more detail, the
reader is directed to recent work (22, 89, 229).
A. The Deafferented Subject
The rare condition of deafferentation due to a large-fiber
sensory neuropathy deserves particular attention because it
2 s
Preferred sensory direction
Direction of vibration illusion
FIGURE 8. Coding of movement directions by human muscle spindles recorded by microneurography.
A: responses of an extensor digitorum longus muscle spindle to ankle movement, shown for 8 of the 16 tested
directions. The diagram in the middle indicates the directions: the solid lines, the directions for which
responses have been shown, and the dashed lines, where the responses have not been shown. In the response
record alongside each direction line, the top trace shows instantaneous frequency of spindle discharge, the
middle trace the recorded impulse train, and the bottom two traces the X and Y coordinates (vertical and
horizontal) of the movement. B: responses, like those in A, were used to determine the preferred sensory
direction for each afferent and, using a population vector model, the mean preferred sensory direction for the
population of afferents in the muscle was calculated. Different-colored lines indicate the mean preferred
sensory directions for each of four muscles (PL, peroneus lateralis; EDL, extensor digitorum longus; EHL,
extensor hallucis longus; TA, tibialis anterior). C: mean directions of vibration-evoked movement illusions in the
four muscles, pooled from 10 subjects. Directions of the perceived illusions were indicated by subjects on a
report sheet. They corresponded reasonably well with the calculated mean preferred sensory directions of the
muscles. [Data modified from Bergenheim et al. (21), with kind permission from Springer Science and
Business Media.]
1665Physiol Rev VOL 92 OCTOBER 2012
provides unique insight into aspects of movement control
and the body image in the absence of proprioceptive feed-
back. These patients typically have sensory nerve fibers be-
low 7
m diameter and all motor nerves intact (e.g., Refs.
71, 75, 81). It is notable that reports of these large-fiber
sensory neuropathies have involved adult subjects who
have already acquired their full repertoire of movements.
The subject loses proprioceptive and tactile inputs (from
large-diameter afferents) from the neck down, while motor
nerves (and small-diameter afferents) are left intact. On first
experiencing the sensory loss, the prone subject is unable to
move and feels as though they are disembodied, floating in
air (72). This sensation receded as the subject taught them-
selves to move. Learning to move was a lengthy process, in-
volving extensive retraining, and even then movements re-
mained slow with some persistent ataxia. Recovery of a sense
of embodiment in deafferented subjects suggests that in nor-
mal subjects the intention to move may be as important as
afferent feedback in acquiring a sense of embodiment (72).
In the deafferented subject, control of movement is based on
visual attention to the target and on cognitive effort (the
will to move). The subject is required to focus on the act of
the movement itself, such as making the arm move and the
fingers flex to grasp the object. In the dark, the subject does
not know where their limbs are and is unable to make
controlled movements. The subject has to visualize external
space and their own body to move one within the other.
Here topokinetic movements, movements to target posi-
tions, are distinguished from morphokinetic movements
used to shape the body part involved in the motor task, for
example, shaping the hand to pick up an object (74). While
the deafferented subject initially loses the ability to make
gestures, they are able to recover them remarkably well. It
has been suggested that in the deafferented subject some
aspects of gestures remain normal, and this has led to the
proposal that gesture is a linguistic phenomenon and not
related to instrumental movements (74).
B. Representations of the Body
The concept of body images in the brain was introduced by
Head and Holmes (182). They proposed the concept of
body representations based on neurological patients with
selective loss of particular sensations. These pioneering
ideas have evolved over the years (for review, see Refs. 89,
133). One proposal is that there are two distinct body rep-
resentations. The body image is a cognitive representation
of the body that is based on stored knowledge and experi-
ence and is thought to underlie perceptual judgements. In
addition, there is the body schema that is dependent on ongo-
ing proprioceptive input, operates largely unconsciously, and
is concerned with body movements (133, 279). Areas of cere-
bral cortex attributed to these functions are the parietal cortex
for immediate guidance of action while conscious perception
and memory may be associated with the insula (92).
In support of dual systems of body representation, Paillard
(279) described two patients with specific neurological def-
icits. One subject had a stroke with damage to the parietal
cortex on the left side. She was left with a complete cutane-
ous anesthesia in her right hand. Following touch of the
anesthetic area by the experimenter, she was able to localize
accurately the site of touch, yet she insisted that she had not
felt anything. Such behavior was interpreted as an example
of blind touch (by analogy with “blind sight,” Refs. 409,
410). In other words, it was an example of being able to
determine “where” without knowing “what.” The second
subject had a peripheral sensory neuropathy. She was
able to detect pain and thermal stimuli applied to the
hand but could not localize them to her own hand. Yet
she could accurately locate the touched spot on a drawing
of the hand. Here she appeared to resort to stored knowl-
edge of the spatial configuration of the hand. Similar
evidence for a dual system of body representation comes
from cases reported by Anema et al. (9). They described
two stroke patients, one who was able to accurately lo-
cate a touched spot on the hand, but poor at identifying
where on a line drawing of the hand they had been
touched. This was considered an example of a distur-
bance of the body image. The other patient exhibited the
reverse behavior; they did not know where they had been
touched, yet they could identify the spot on a line draw-
ing of the hand. This was considered a disturbance of the
body schema.
Patients with large-fiber sensory neuropathies typically
have sensory nerve fibers below 7
m diameter intact and
all motor nerves intact (e.g., Refs. 71, 75, 81). In subjects
with large-fiber sensory neuropathies, if itch was induced in
the skin by local topical application of histamine, they
could accurately indicate the site of the itch (22). This find-
ing suggests that senses such as itch and pain, as well as
warm and cold (73, 279), can, to a limited extent at least,
act as de facto proprioceptors to guide movement. How-
ever, a small-fiber system, capable of only slow, sluggish
responses, cannot replace the full proprioceptive feedback
available to normal subjects. In experimentally deaffer-
ented subjects, the small-fiber system does seem to maintain
the sense of limb ownership (282).
In the deafferented subject, partial motor control is achieved
using a consciously maintained body image (134). Mental
control of movements is limited to simple movements, is
slower than normal, and the movements are relatively
short, as otherwise they become too exhausting for the sub-
ject. Complex movements require a lot more energy to ex-
ecute than simpler ones. By comparison, a normal subject
can forget about their body in daily routines. It takes care of
itself. In simplistic terms, this is because the body schema
functions to control posture and movement unconsciously,
without the intervention of a body image.
1666 Physiol Rev VOL 92 OCTOBER 2012
Difficulties in providing a precise definition for concepts
such as body schema and body image have led to a variety of
related but distinct proposals. Carruthers has proposed that
all representations of the body are available to conscious-
ness (60, 61). “On-line,” newly constructed body represen-
tations, provided by inputs such as vision, touch, and pro-
prioception, generate a perception of the body as it actually
is at any moment in time, an image which is able to change
from moment to moment. It is distinguished from an “off-
line” representation constructed, in part, from current sen-
sory inputs, in part, from stored memories and is available
to consciousness both immediately and after retrieval of
The idea of an off-line representation is useful as it helps to
account for phenomena such as inclusion of hand-held tools
in the body schema (58) and the development of phantom
limbs after limb amputation (e.g., Refs. 154, 185, 252,
315). The failure to remove the amputated limb from the
off-line representation leads to development of a phantom.
A similar situation arises when a subject becomes unaware
of paralysis of a limb following hemiplegia (anosognosia).
Again, the paralyzed limb is not removed from the off-line
image (125). One way to eliminate these phenomena at least
temporarily is by use of mirror box therapy (125, 314).
An interesting comment on the sensory processing involved
in the generation of a body image is provided by patients
suffering from anorexia nervosa. They experience pro-
nounced body image distortions in combination with the
desire to maintain a low body weight. Anorexic patients,
when looking into a mirror, do not see their actual body im-
age, but an overweight one. In support of the idea that the
integration of visual and proprioceptive information is abnor-
mal in these patients, it was shown that anorexic individuals
do not exhibit the normal size:weight illusion (63).
C. Phantom Limbs
A phantom limb can persist for months and years after
amputation (252). Indeed, it even arises in some subjects
with congenitally absent limbs (254). In response to motor
commands, the phantom can be made to move in two ways.
Component parts of the phantom may change position rel-
ative to one another. Here the sensation is thought to arise
centrally. Second, when the body part bearing the stump of
the limb is moved, the whole phantom may move with it
while retaining its relative position to other body parts.
Voluntary movements made in this way by a phantom limb
are often accompanied by contraction of the appropriate
muscles in the stump. If the remnant muscles are dener-
vated, the ability to move the phantom is lost (185, 321,
374). However, observations with experimental phantom
limbs (see below) suggest that movement sensations can
also arise centrally.
Putting the observations together, the phenomenon of a
phantom limb is largely of a static limb. The sensation does
not appear to depend on peripheral input and can be gen-
erated by central neural activity, but which can be modified
by peripheral sensory input. The phenomenon emphasizes
the uniqueness of the sense of limb position. The sense of a
moving phantom may be dependent on stump muscle activ-
ity, but it can also arise centrally (205, 256, 322).
The more distal parts of a phantom limb such as the hands
or feet, especially the digits, are more strongly perceived
than proximal segments. This may relate to the denser pe-
ripheral innervation and, consequently, more extensive cor-
tical representation of distal segments. Perception of the
phantom limb changes with time, the more weakly per-
ceived parts fading from awareness, often leading to a tele-
scoping of the limb (e.g., Ref. 327). Such an effect presum-
ably relates to a reorganization of the cerebral cortex where
areas deprived of peripheral input are invaded by adjacent
areas (e.g., Refs. 123, 294; for review, see Refs. 41, 207).
D. Experimental Phantom Limbs
The idea of producing an acute block of sensory and motor
nerves to a limb to produce an experimental phantom was
pioneered by Melzack and Bromage (253). With their eyes
closed, prone subjects described the location of a phantom
arm after an anesthetic nerve block at the brachial plexus.
The phantom lay at the side of the body, above the lower
abdomen or above the lower chest. During the block, or-
derly spontaneous changes in posture occurred between
these three positions, positions which bore no direct rela-
tionship to the position of the arm at the time or before the
onset of the block. The reported postures are reminiscent of
postures of amputation phantoms (185).
Recent observations provide insight into how an on-line
representation of the body might be constructed (198). An
experimental phantom hand and forearm can be produced
by means of an ischemic nerve block in the upper arm; the
hand becomes both anesthetic and paralyzed (e.g., Refs.
148, 398). The method is depicted in FIGURE 9A. To signal
the position of the experimental hand, subjects manipulated
an articulated model hand with their other hand. When the
nerve block was established, subjects signaled the perceived
position of the unseen hand every 3 min by adjustments of
the model. As the block took effect, perceived posture of the
hand gradually changed. If the fingers were held straight
throughout the block, the perceived posture of the “dead”
hand had the fingers flexed. When the fingers were held
flexed during the block, the phantom adopted a more ex-
tended posture at the finger joints. The results are shown in
In the simplest interpretation of the observations, posture of
the hand is signaled largely by afferents from the muscles
1667Physiol Rev VOL 92 OCTOBER 2012
and skin stretched by that posture. Perhaps the key param-
eter is the difference in discharge rates between afferents in
hand flexor and extensor muscles. When the hand is held
extended, flexors are stretched and extensors shortened. As
the nerve block takes effect, the discharge from the
stretched flexors falls, and this is interpreted by the subject
as the hand moving into a more flexed posture. When the
starting position is a flexed hand, the declining discharge
coming from stretched extensors leads to perception of a
more extended posture. Such observations emphasize the
lability of the body schema and its dependence on the mo-
ment-to-moment input of activity coming from the body
In relation to the question of movement sensations by a
phantom limb, under the conditions of the above experi-
ment, when subjects attempted to move their paralyzed,
anesthetized hand, they perceived a distinct movement of
the hand as well as displacement of its position (e.g., Refs.
148, 398). The velocity of the movements and total angular
displacement of the hand graded with the level of effort
while total displacement also graded with duration of ef-
fort. Increasing the effort 10-fold increased the size of the
movement 2- to 3-fold and the perceived movement speed
about 2-fold (398) (see FIGURE 9C). There are interesting
differences with movement sensations in the oculomotor
system: here recent studies during full paralysis of the eye
muscles have shown that in darkness commands to move
–30 –20 –10 0 10 20
B Position sense
C Movement sense
Target wrist angle (°)
5% effort
Slow Slow
Medium Medium
5 s 5 s 5 s 5 s
1 s 1 s
1 s 1 s
50% effort
Perceived wrist angle (°)
Fast Fast
Fastest Fastest
FIGURE 9. The experimental phantom hand: the effect of motor
commands on perceived position and movement. A: method used to
produce ischemic anesthesia and paralysis of the right forearm and
hand using a pressure cuff on the upper arm. To test position sense,
the wrist and hand can be placed in specific postures by the exper-
imenter, and the subject uses their left hand to move the pointer to
indicate its perceived position. B: when the hand had been anesthe-
tized and paralyzed, the perceived position of the relaxed hand (),
with the subject making flexion efforts (), and with them making
extension efforts () plotted against test angles from 30 to 20°.
The neutral position of the phantom hand is perceived as slightly
flexed (). Attempted flexion or extension movements at 30% max-
imum effort causes large perceived displacements in the direction of
flexion for flexion efforts and in the direction of extension for exten-
sion efforts. [Bfrom Gandevia et al. (148).] C: perceived motion of
the phantom hand. The filled hands (thumb, index finger, and middle
finger are shown) represent the actual position of the hand. Left
figurine: subjects made effort of 5% maximum (for 1 or 5 s); right
figurine: subjects made effort of 50% maximum (for 1 or 5 s). When
subjects made voluntary efforts to move their paralyzed and anes-
thetized hand, the blue-outlined hands show the mean size of the
phantom wrist movements reported during a voluntary effort that
lasted 1 s. The red-outlined hand shows the mean size of the phan-
tom movements reported duringa5svoluntary effort. Subjects
indicated that they perceived movements of their phantom that were
bigger when they made stronger efforts (compare left and right
figurine). Movements with the greatest speed over the duration of
the efforts occurred for efforts at 50% maximum lasting 1 s (blue
hand in right figurine). The speeds of the movements produced by
the four voluntary efforts are ranked by the terms slow, medium,
fast, and fastest. [Credrawn from data in Walsh et al. (398).]
1668 Physiol Rev VOL 92 OCTOBER 2012
the eye are accompanied by a rapid shift of the visual world
(413). In contrast, the illusory movements with paralyzed
limb muscles are slow, as if the hand is moving through
treacle (148, 398).
E. Local Anesthesia
In the above experiments, as the nerve block began to take
effect, subjects reported an increase in the perceived size of
their hand (198). It is a common experience that dental
anesthesia leads to perceived increases in size of the lips,
tongue, and parts of the face (384). Similarly, anesthesia of
the thumb increases its size by 60–70% (147), and a bra-
chial plexus block produces illusory swelling of the arm
(281). These phenomena, while relevant to the broader sub-
ject of body image, probably relate to acute changes within
cortical areas responsible for their representation. Interac-
tion between excitatory and inhibitory influences at cortical
and subcortical sites produce an expansion of cortical areas
representing the body part and that expansion is presum-
ably responsible for the perceived increase in size (54–56).
It suggests that the cortical representation of each body part
has an inhibitory surround maintained by small-fiber input
which, if removed, leads to an expansion of the representa-
tion. Illusory increases in size may not appear to be so
dramatic when a large part of the body is anesthetized or
removed (227, 258). While these immediate distortions of
body size with local anesthesia have been linked to loss of
input from small-diameter afferents (e.g., Refs. 54, 281,
282), recent work using ischemic anesthesia indicates that
the loss of large-diameter afferents also plays a role (198).
F. The Body Form
It has been known for some time that the consciously per-
ceived body image is distorted. When normal subjects are
asked to point with their left index finger to different parts
of the hidden right arm, the resting arm is systematically
perceived to lie closer to the midline as well as closer to the
body than its true position. Errors increased the longer the
arm remained hidden. The observed postures resembled
those seen with phantom limbs (e.g., Ref. 173). More de-
tailed, information about what we perceive to be the size
and shape of our body has recently been provided (229).
The authors make the point that although information
about body posture is specified by ongoing afferent input,
no sensory signals are available to detail body size and
shape. The incoming positional information must be re-
ferred to a stored model of the body’s metric properties. To
reveal aspects of the model, maps of the mental representa-
tion of the unseen hand were drawn. The size of a finger was
determined from landmark locations, the finger tip and
knuckles, indicated by the subject. They revealed a distorted
map with the fingers shorter than their actual size and the
hand broader than it really was. These distortions resem-
bled cortical somatosensory representations of the hand as
drawn in the sensory homunculus (290). In contrast to these
distortions, explicit judgements of hand shape assessed in a
template-matching task revealed that the mental image of
the hand lay close to reality, and it did not reveal the dis-
tortions measured by pointing. It was concluded that in
addition to the body schema and body image, a new repre-
sentation had to be included, which was called the “body
27 min 27 min
13 min
13 min
FIGURE 10. Perceived changes in posture of the relaxed hand during paralysis and anesthesia. When the
starting position was an extended hand (Start, left, blue), as the nerve block took effect, the perceived posture
became progressively more flexed over the 40 min the block was in place (Final). When the starting position
was a flexed hand (Start, right, red), the perceived posture became progressively more extended. [Redrawn
from Inui et al. (198).]
1669Physiol Rev VOL 92 OCTOBER 2012
form.” In answer to the question, how was the distorted
body form reconciled with the body schema/image, one
suggestion was that the motor system simply avoided ex-
plicit representation of initial limb position by coding the
desired end-point of a movement (162, 293), although such
end-point coding has some limitations (65). As the motor
system typically relies on vision as well as proprioception,
correction of the distorted body form may occur from mo-
tor learning using these additional inputs.
G. Development
There is some information on developmental aspects of a
body image/schema. Soon after birth humans show a strong
innate tendency to mimic sounds and motor acts, copying
orofacial movements performed by adults before them
(251). Such behavior supports the existence of mirror neu-
rons (for review, see Ref. 180), with imitation providing the
basis for an understanding of the relationship between one-
self and others. This is postulated to lead to the sense of self
as an agent and as the target of social interaction. Babies a
few weeks old vocalize in response to language and attend
to their mother’s orofacial area, much more than her body,
as a source of emotional support. They see movements and
then copy them. Visual recognition of spatial configurations
of the body below the face matures after one year of age,
followed by the ability to recognize and name body parts, as
well as acquisition of language skills (364). About this time
the child begins to recognize his or her own face and body in
a mirror, an ability considered an objective index of emerg-
ing self-consciousness (211).
A broader question concerns the role of proprioception in
development of motor skills in children. There must be
regular recalibration of the body schema in response to
growth of the body. Pioneering observations by Laszlo and
Bairstow (223) have led to the view that the ability of chil-
dren to use proprioceptive feedback has matured by the age
of 7 years. Other reports suggest that there is ongoing learn-
ing of movement control which continues through adoles-
cence (156), and into adulthood (183). So, development of
proprioceptively controlled movements continues beyond
the period required for acquisition of motor skills, thought
to be complete by the age of 10 years (348).
H. Body Ownership
Proprioception, by definition, is associated with actions oc-
curring to the self. This begs the question, how does the
brain know what is part of the body and what is not?
Exteroceptive sensations such as vision and hearing cannot
distinguish between foreign body parts and our own. Belief
in what we think belongs to our body can be surprisingly
easily manipulated by experiment. This has been revealed
by many recent studies in which subjects become convinced
that an artificial limb is their own, using what is known as
the “rubber hand” illusion (e.g., Ref. 31). In extensions of
this illusion, a subject can, for example, have multiple limbs
(106, 274) or even appear to be located outside their own
body (105, 291). In the rubber hand illusion, the subject’s
unseen hand is stroked at the same time as a rubber hand
lying near it. Within seconds, subjects become convinced
that the rubber hand is their own to the extent that it is
used in motor planning (e.g., Refs. 31, 208) and they
respond defensively when the artificial hand is threatened
(e.g., Refs. 14, 110).
Adoption of “ownership” of the artificial limb does not
require vision and can be equally effectively achieved in a
blindfolded subject if the experimenter guides the subject’s
left hand to touch the rubber hand while at the same time
touching the subject’s own right arm (107). Cutaneous
stimulation by brushing the skin has been the method used
to induce the illusion. Recent studies have shown that own-
ership illusions can be evoked by both passive and active
movements (97) even of just one finger (400). The illusion is
just as strong after local anesthesia of cutaneous and joint
afferents from the finger and thus the input provided by
muscle spindle afferents with vision of a false finger is suf-
ficient to generate sensations of hand ownership (400). Im-
portantly, in unpublished work (Walsh, Heroux, Butler,
and Gandevia), this role of muscle spindles has been estab-
lished even when there is no vision of the hand (using a
complimentary approach to Ehrsson et al., Ref. 107).
I. Brain Areas Associated With
the Body Image
Neuroimaging studies have revealed regions that are acti-
vated by different aspects of proprioception and its interac-
tion with the body representation. These include multisen-
sory regions in the parietal cortex (e.g., Refs. 26, 83, 107,
109, 126, 176, 226, 289), as well as parts of more distrib-
uted networks including the frontal cortex and insula (e.g.,
Refs. 346, 383). There are brain areas selectively responsive
to viewing human bodies, except the face, while others are
preferentially responsive to viewing the face. The extrastri-
ate body area is concerned with individual body parts like
the fingers and recognizing that they belong to oneself (e.g.,
Refs. 96, 190, 269). Adjacent to it is the fusiform body area
which also shows a selective responsiveness to viewing the
body, but it seems to be more concerned with movement of
large parts of the body like the torso. The fusiform body
area overlaps in part with the face selective area in the
middle of the fusiform gyrus. A second face area is the
occipital area (150, 288). Even the primary somatosensory
cortex, which is traditionally viewed as a map of the phys-
ical body, participates in illusory proprioceptive distortions
of the limb (343, 344).
1670 Physiol Rev VOL 92 OCTOBER 2012
Left parietal lesions are implicated in an alteration of body
awareness in which patients are unable to point to parts of
their own body on verbal instruction, in spite of being able
to respond to verbal commands to move that part towards
an external object (e.g., “touch the pedal of that bicycle”).
These patients have lost the sense of relations between body
parts, knowing for example, that the mouth is between the
nose and the eyes (114).
To conclude, there are a number of distinct brain areas respon-
sible for recognition of body and face, arguing for more than
just two “bodies in the brain.” The posterior parietal cortex,
anterior insula, and extrastriate body area are important ar-
eas. They are present in both hemispheres, but most studies
have provided evidence for a right-sided dominance in many
aspects of body awareness (22), and this may involve parts of
the primary somatosensory cortex (271).
When we carry out a motor task we are aware of the muscle
contraction accompanying the action, how much effort is
required and, if it is an object to be moved, we have a sense
of its heaviness. The senses of effort, force, and heaviness
are traditionally believed to be generated, at least in part, by
signals of central origin associated with motor commands.
Peripheral sensory receptors are also available to contribute
to the sensation, including tendon organs, muscle spindles,
and pressure-sensitive skin receptors.
As proprioceptive sensations, the senses of effort, force, and
heaviness are distinct in that they are always associated with
motor commands, while kinesthetic sensations can arise in a
passive limb. It implies that for the senses of force and heavi-
ness the peripheral input is always reafferent in origin. An-
other issue to consider is the meanings of the terms. What do
we mean by a sense of effort as distinct from a sense of tension?
A. The Sense of Effort
As mentioned in section I, the prevailing view during much
of the 19th century was that our muscle senses arose en-
tirely centrally, and this included the sense of effort experi-
enced during execution of a motor task. However, during
the middle of last century, two influential reports argued for
a combined central-peripheral mechanism for propriocep-
tion. The impact of these ideas is still evident today. In a
study of eye movements in fish, Sperry (369) introduced the
term corollary discharge meaning, “a corollary discharge of
motor patterns into the sensorium may play an important
adjuster role. . . . along with non-retinal and postural influ-
ences from the periphery”. So here was the suggestion of a
central sensory area acted upon by both motor command
signals as well as signals originating in the periphery. In the
same year, 1950, von Holst and Mittelstaedt (393) put for-
ward a similar proposition based on experiments in insects.
They introduced one additional important concept that
continues to be influential. They distinguished between af-
ferent signals generated by the animal’s own actions and
signals generated by external sources as the animal moved
about in its environment. The term exafference was used for
afferent signals generated by stimuli of an external origin
and the term reafference was used for afferent activity aris-
ing from the body’s own actions. The reafference was cal-
culated using an efference copy of the motor command. It
remains a matter of conjecture whether the efference copy is
an accurate copy of the motor outflow, or a computation,
based on past experience, of the expected afferent signal for
a given motor outflow (95). The total afferent signal gener-
ated during a motor act is made up of two components: the
predicted reafferent signal, determined by the motor com-
mand, and afferent activity generated independently of the
motor command, as a result of external influences. While
the corollary discharge hypothesis was put forward to deal
with the problem of visual stability during eye movements,
the efference copy hypothesis has implications for all move-
ments (194). These ideas have recently been formalized in a
model for motor control (19, 423). Some aspects of such a
model are considered in FIGURE 11.
The underlying idea is that not all of the signals generated
by a movement reach consciousness, and we perceive only
the exafferent component. So, for example, the sensation
when we are tickled by someone else differs in quality and is
much more intense than when we attempt to tickle our-
selves (23, 24). Part of the mystery of the proprioceptive
senses is that much of the afferent information generated by
self-initiated movements does not reach consciousness. As a
consequence, the detection and perceived intensity of affer-
ent signals are diminished during movement (for references,
see Refs. 260, 414). Such suppression can occur along so-
matosensory pathways (e.g., Ref. 67) as well as from the
actions of the motor command signals themselves (415).
If a compression force, generated externally by a motor, is
applied to the hand and the subject is then asked to repro-
duce the felt force using their other hand, they invariably
apply a greater force than was applied externally (353, 354,
394). This is illustrated in FIGURE 12A. The overestimation
is presumably due to the partial suppression of the afferent
activity associated with self-stimulation. In everyday activ-
ities, such sensory suppression would allow us to focus
attention on external stimuli and be less distracted by the
sensations arising from our own movements. A model has
been proposed for this attenuation of self-generated move-
ments (19). A recent reexamination of the phenomenon
showed that the overestimation of forces applied externally
to the index finger of one hand when they were reproduced
by self-generated force only occurs for low applied force
levels but not higher forces (402). The overestimation per-
1671Physiol Rev VOL 92 OCTOBER 2012
sisted when digital nerve anesthesia was used to block cu-
taneous and joint sensation, leaving only combined inputs
from muscle and central command sources available to sig-
nal the force. In addition, these new data revealed two com-
ponents to the overestimation at low forces: an offset as
well as an increase in the gradient between the matching
and target forces. The latter was present only when the
finger was actively held in position. That is, as the target
force was increased, the discrepancy between target and
matching forces increased. This is shown in FIGURE 12C.
The operation of forward models should not be seen as static
events, operating in isolation. So, for example, in learned mo-
tor tasks (overlearning), we carry out the action without think-
ing about it, unaware of its predictability. This is an example
where the predicted and observed afferences match, and there
is no accompanying sensation. This is very different from an
attention demanding task which, incidentally, can be carried
out simultaneously with the overlearned task (130). So, in our
everyday behaviors, we should imagine constantly shifting
strategies, from the operation of internally generated actions
to actions dominated by feedback from the periphery. The
example of overlearning also implies that there is a learning
stage to the task where proprioceptive feedback is used to
fine-tune its execution (130).
B. Paralysis and Fatigue
It has been known for a long time from the clinical literature
that partial paralysis comes with abnormal sensations of
increased heaviness (e.g., Ref. 192). So, if there is a discrep-
ancy between the reafference generated by a movement and
the efference copy derived from the command signal, that
the subject is typically unaware of, sensory illusions may
occur. Patients with partial paralysis from an upper mo-
toneuron lesion feel weights to be heavier on the affected
side (144). When paralysis is induced experimentally by
infusing a neuromuscular blocker, the effort required to lift
a weight and its perceived heaviness was increased, com-
pared with the unparalyzed condition (e.g., Refs. 141, 144).
The authors made the additional observation that during a
lifting task using thumb flexors, if skin and joint input from
the thumb was removed by anesthetic block, the perceived
heaviness increased further (see also Ref. 145). This was
interpreted as the removal of a reflex facilitation generated
by the cutaneous input, leading to an increase in central
command and therefore perceived heaviness. Such a result
fits with abolition of the long-latency stretch reflex in the
lifting muscles by thumb anesthesia (238). The digital an-
esthesia will also have removed the cutaneous contribution
to a peripheral force signal. Perceived heaviness can be ma-
nipulated by altering reflex excitability of hand motoneu-
rons (12). Inhibition of the motoneuron pool led to an in-
crease and facilitation led to a decrease in perceived heavi-
ness of weights. Presumably, the contribution from a
central command signal is adjusted based on the level of
other excitatory or inhibitory drives reaching the motoneu-
rons being used in the task. In other words, what we feel is
automatically adjusted according to any reflex contribution
of which we remain unaware. Given that the strength of
Body in
model of body
and sensory system
discrepancy Perception
Corollary discharge or
Efference copy Predicted sensory feedback
Actual sensory feedback
FIGURE 11. Model for a central comparison between sensory and motor signals, based on the proposal of
Bays and Wolpert (19). The process begins with the intention to move (Plan), leading to generation of a motor
command and its efference copy/corollary discharge. The Forward model uses the efference copy to calculate
the expected outcome, and this is compared with the input (Body in Environment) by means of a difference
calculator to quantify the sensory discrepancy that determines what is actually perceived (Perception). Three
points for consideration are as follows: 1) if the predicted and actual sensory feedback match, the anomalous
situation arises where there is no perception at all; 2) the forward model needs to be regularly updated over
both short and long time scales; and 3) the “difference calculator” is more complicated than depicted here
because, at least for force sensation, its output is not determined by a simple subtraction process (for details,
see Ref. 402).
1672 Physiol Rev VOL 92 OCTOBER 2012
cutaneous and muscle reflex effects on the human fusimotor
system is low even during contractions (e.g., Refs. 11, 146,
149), it is unlikely that the changes in perceived heaviness
that are evoked by reflex inputs (at least from skin and joint
receptors) can be explained by a change in muscle spindle
A method commonly used to disturb the sense of effort is
through fatigue from exercise (e.g., Refs. 62, 142, 206, 245,
404). A contralateral force-matching task is used, in which the
isometric forces generated in one arm are matched by the
corresponding muscle group of the other arm (FIGURE 13).
Under control conditions subjects are quite accurate at
matching forces, producing matching errors of 5% maxi-
mum voluntary contraction (MVC) or less between the two
sides. However, when one arm is exercised to fatigue, the
level of force generated in the fatigued muscle is usually
overestimated. The overestimation was considered to arise
from efferent signals of similar magnitudes being sent to the
muscles of the two arms, the unfatigued muscle responding
with a higher force for a given level of activation. Similar
conclusions have been arrived at by others (142, 245). In
the first of these studies, weights supported by fatigued
muscles felt heavier than when supported before fatigue. In
discussing the origin of the sense of effort, McCloskey et al.
(245) declared, “Whether this sense of effort should be
regarded as simply a centrifugal mechanism or as involving
some inherent comparison of outflow with afferent inflow
remains an open question.”
In support of an “effort” hypothesis, Cafarelli and Bigland-
Ritchie (51) carried out a force-matching task with muscles
1 2
5 100
Locking mechanism
Load cell
Target force
Presented force (N)
Matched force (N)
Presented force (N)
Matched force (N)
FIGURE 12. Comparison of perceived forces generated by self
stimulation compared with stimulation by an external source. A: sub-
jects were instructed to reproduce a target force, applied to the left
index finger by a torque motor, either directly by voluntary pressing
with the right index finger (red) or indirectly by controlling the output
of a torque motor with a remote joystick (blue). Average matching
force (SE) is shown as a function of target force. Dashed line
represents perfect performance. [Redrawn from Bays and Wolpert
(19).] B: the subject’s forearm rested on a table. When the subject’s
index finger was required to be passive, it rested on a support, but
when subjects were required to actively hold the index finger in
position, the support was removed. The shaft of the load cell could
be locked in place or free to move up and down. The subject was
asked to generate a matching force either by pushing down on the
top of the platform with the other hand while the shaft was free to
move or by pushing up isometrically against the load cell with their
index finger while the shaft was locked. [From Walsh et al. (402).]
C: diagrammatic representation of results obtained for forces below
10 N. Subjects overestimate an externally generated target force to
a passive finger by 2–3 N (offset) when matching it with a voluntary
contraction (e.g., flex the test finger or push down with the other
hand). This occurs at low forces and even when cutaneous and joint
inputs were blocked in the test finger. A central factor increases the
gradient (gain) of the relationship between matching and target
forces by 20 if the target force is received on an active finger.
[Data from Walsh et al. (402).]
1673Physiol Rev VOL 92 OCTOBER 2012
of different strengths. In thumb adductors and elbow mus-
cles, using the muscle’s length-tension relation to alter the
level of force produced, it was concluded that matches were
being made on the basis of equal sensations from the mus-
cles being compared, not their forces. A similar conclusion
was reached in a related experiment on elbow flexors. When
the reference arm was held at angles representing the as-
cending or descending limb of its elbow flexor torque-angle
relation, the matching torque generated by the indicator,
held at its optimum length, was always larger (404).
Another potential source of errors in a force-matching task
comes from activation of the small myelinated group III and
unmyelinated group IV afferents in skeletal muscle. It is
known that these afferents become active during a muscle
contraction (209). Muscle afferents mediating pain, pre-
sumably involving both group III and IV (255), can be stim-
ulated by injecting hypertonic (5%) saline into a muscle
(161). In a two-arm force-matching task, when biceps of
one arm was made sore by injecting hypertonic saline, the
force in that arm was overestimated. That is, the subject
thought the painful arm was generating more force than it
actually did (407). This kind of mechanism is likely to ac-
count for the force-matching errors observed after eccentric
exercise (312). Eccentric exercise is known to produce mus-
cle damage and soreness (309). It remains uncertain how
such an effect is mediated. Perhaps it is the result of influ-
ences on central sensory rather than motor pathways (213).
C. Two Senses
In an experiment in which they used muscle vibration,
McCloskey et al. (245) claimed that if appropriately in-
structed, subjects had the ability to select between a sense
of effort arising centrally and a sense of muscle tension,
presumably deriving from intramuscular receptors. The
same conclusion was reached by Roland and Ladegaard-
Pedersen (329). Subjects compressed springs of different
stiffness with one hand and matched the extent of com-
pression and the force required with the other hand. Dur-
ing partial paralysis, the observed errors were below the
expected levels if there had been a proportional relation-
ship between the exerted effort and the generated force.
When subjects were instructed not to match forces but
efforts, this led to the expected increases in errors during
neuromuscular blockade.
Put together, these various observations suggest that we
may have two senses, a sense of effort and a sense of muscle
tension. The sense of tension can only be revealed if the
subject is appropriately instructed. This is a difficulty for
studies in proprioception as it raises the problem of how to
instruct subjects to perform a task. In experiments con-
cerned with the effects of fatigue on force sense, subjects
were specifically instructed to match the force generated in
the reference arm, not the effort. Nevertheless, fatigue-re-
lated matching errors were observed that were interpreted
as the subject relying predominantly on a sense of effort
(407). Comparable observations had been made earlier (62,
206; see also Ref. 214). Such observations left open the
question of the contribution from peripheral receptors to
the sense of force.
0 20 40 60 80
Strain gauges
Time (hour)
Torque (% Pre)
Error (% MVC)
–50 0
FIGURE 13. Isometric force matching before and after fatigue.
A: the subject is seated with their forearms in the vertical position,
strapped to paddles locked at 90°. The torque generated by elbow
flexor muscles is recorded by strain gauges, and the output of the
reference arm is displayed on a computer screen visible to the
subject. The subject is required to match its torque level with their
other arm (output not visible). B: the control error (red, right hand
ordinate) is close to zero. Then the subject was required to carry out
a series of fatiguing eccentric contractions that led to a 40% drop in
MVC torque (blue, left hand ordinate). The fall in torque led to an
overestimation of 14% by the unexercised arm of the torque gener-
ated by the fatigued reference arm. Over the subsequent 100 h, the
matching errors gradually fell to a residual 4% while torque recov-
ered to control levels. [Redrawn from Weerakkody et al. (404), with
kind permission from Springer Science and Business Media.]
1674 Physiol Rev VOL 92 OCTOBER 2012
D. Muscle Vibration
Another way to manipulate muscle force sense is to use
muscle vibration during the matching trials. However, the
effects of vibration are not straightforward to interpret. In a
passive muscle, vibration will stimulate predominantly pri-
mary endings of muscle spindles. The Ia input facilitates
spinal motoneurons and may produce a tonic vibration re-
flex (TVR) in the vibrated muscle (175). Once the muscle
begins to contract, the contribution of tendon organs to the
afferent activity will start to increase (113), adding segmen-
tal inhibitory reflex effects while generating force sensations
centrally. After a period of vibration, muscle receptors are
likely to become desensitized, and there may be reduced
motoneuron excitability from postactivation depression
(196, 425). Because of these multiple influences, psycho-
physical results from vibration experiments must be inter-
preted with caution. Not surprisingly, they have produced
contradictory findings.
McCloskey et al. (245) reported that in 9 of 15 subjects
carrying out a force matching task, vibration of biceps of
one arm led subjects to match the reference force in the
vibrated arm with smaller forces in their other arm. In ex-
periments on force matching in quadriceps, the opposite
result was obtained (52). When a vibrated muscle’s force
was matched with a nonvibrated muscle, subjects system-
atically chose higher forces with the nonvibrated muscle. To
explain their result, the authors proposed that force gener-
ated in the TVR was sentient and the extra central input
from it led the nonvibrated muscle to match with a larger
force. Alternatively, vibration of the contracting muscle
acted as a powerful stimulus for tendon organs, and the
additional tendon organ input was interpreted as a higher
level of muscle force. This is an interesting interpretation in
view of the more recent observations by Luu et al. (233)
who interpreted their observations on the senses of force
and heaviness after prolonged muscle vibration as the result
of a vibration-evoked desensitisation of muscle receptors.
The findings of Cafarelli and Kostka were extended by Ca-
farelli and Layton-Wood (53) who showed that the increase
in perceived force produced by vibrating a muscle gradually
faded if the muscle was subjected to fatiguing exercise. One
interpretation is that muscle receptors have become desen-
sitized by the fatigue process so that their response to vibra-
tion diminishes.
E. The Sense of Force and the Tendon Organ
The most eligible candidate for a tension receptor is the
tendon organ (for a review, see Refs. 136, 201, 302). Ten-
don organs, located at the ends of muscle fibers, are ideally
sited to monitor muscle tension (FIGURE 2). Like muscle
spindles, tendon organs are stretch receptors. However,
they are principally contraction receptors. A tendon organ
will respond powerfully to contractions of motor units that
contribute muscle fibers that insert into the receptor cap-
sule. It will largely ignore contraction of fibers outside the
capsule. Most of the muscle fibers inserting into the capsule
of a tendon organ come from different motor units, so one
tendon organ can be excited by 10–20 motor units. Calcu-
lations indicate that every motor unit in a muscle is linked to
at least one tendon organ (302). Most tendon organs signal
tension in a select group of motor units and can be consid-
ered to be regional tension sensors. They are therefore able
to monitor submaximal contractions involving the motor
units to which they are linked. Populations of tendon or-
gans can signal whole muscle force (e.g., Refs. 84, 136).
Group I afferents, including those of tendon organs, project
to the cerebral cortex (248), with projection to the cortex
being a prerequisite for conscious perception of the recep-
tor’s activity. If impulses from tendon organs evoke con-
scious sensations, it raises the possibility that they are di-
rectly involved in the sense of tension. Here a consideration
is that some central processing is necessary to convert sig-
nals of intramuscular tension into the required torque levels
for moving a limb about a joint (140).
In an attempt to link tendon organ activity to propriocep-
tive sensations, Thompson et al. (380) asked subjects to
reproduce a learned low force (4% MVC). After subjects
could reproduce the reference force reliably, they repeated
the task immediately after the muscle had undergone a
MVC. The force reproduced after the MVC was twice as
large as after a rest period. It was suggested that desensiti-
zation of tendon organ discharges during the MVC led sub-
jects to overestimate the level of force. This suggestion was
supported by observations in animal experiments that
showed responses of tendon organs to a low-force contrac-
tion were much smaller when this was preceded by a max-
imal contraction. The time course of this desensitization of
tendon organ responses matched the decline in matching
errors observed in the human experiments.
F. Force-Movement Illusions
Perceived limb position is influenced by the forces required
to move the limb to that location (for review, see Ref. 205).
In a position matching task limbs can be accurately aligned
when they remain passive or are supporting a static load
(13, 340). However, when the loaded arm is moved, errors
emerge (13). If a subject is required to generate an isometric
force against the device imposing a flexion movement, the
limb is perceived to be more flexed than it really is (340). A
similar result was obtained when positions were matched
with one arm undergoing elastic loading (403). The magni-
tude of the errors increases with greater applied force. Sim-
ilar conclusions were arrived at in experiments involving
compression of springs of different stiffnesses (329).
There is the old observation that when voluntary movement
of an arm is unexpectedly blocked by an obstacle, the dis-
1675Physiol Rev VOL 92 OCTOBER 2012
tance the subject thinks the arm had moved through, as
indicated by the other arm, is consistently overestimated
(191). The size of the illusion was influenced by the impact
force and by the time after movement onset at which the
impact was encountered; the later the impact point, the
smaller the illusion.
Finally, blindfolded subjects were asked to match the posi-
tions of their forearms in the horizontal plane, under con-
ditions where there were no direct effects of gravity. When
one arm was loaded and the subject moved the loaded arm
to the test angle, matching errors were seen that were not
present when subjects simply supported a static load at the
test angle (13).
All the experiments in this section have the common ele-
ments of position errors from the combined effects of load
and movement. Modern theory proposes that during a
movement, predictions about future states of the limb based
on efference copy combine with sensory feedback signals,
delayed by their conduction and processing times, to pro-
vide an estimate of the current state of the limb, allowing for
a continuous updating of the estimate during the movement
(see, for example, Ref. 423). If an arm movement is stopped
shortly after its onset, the subject’s estimate of hand posi-
tion is dominated by the efference copy based predictive
model, and the subject perceives the hand to be further from
its starting position than is actually the case. During passive
movements, in the absence of any efference copy, because of
transmission delays in the afferent input, such an overesti-
mate should convert to an underestimate. In a recent study
that tested this prediction (172), with passive movements
subjects still overestimated hand position, with errors that
were as large as or larger than with active movements. The
preferred explanation put forward by the authors was that
the kinesthetic system used a Bayesian inference process in
which afferent input and efference copy were both used
continuously during the movement to determine the current
position. This included an estimate of the expected state of
the arm, based on past experience. The greater the uncer-
tainty associated with the afferent input, the more the sys-
tem relied on past experience estimates.
G. Emerging Ideas
When a muscle is weakened by fatigue or paralysis, a given
level of force generated by the weakened muscle is accom-
panied by a less than proportional increase in the perceived
effort. This has been noted by many investigators for both
perceived heaviness and force (141, 242, 407; see also Ref.
328). More recently, Luu et al. (233), in a reexamination of
the problem, began by pointing out that during fatigue or
paralysis, the increase in perceived heaviness of objects, or
perceived level of force generated, was less than expected,
had a central command signal been entirely responsible for
the sensations. If a peripheral force sensor had contributed,
such sensations of increased heaviness or increased force
should not have occurred, since force would have been ac-
curately reported. It appears that the real situation lies
somewhere between these two extremes, that is, force and
heaviness sensations are generated by signals of both central
and peripheral origin.
A new proposal put forward by Luu et al. (233) was that
muscle spindles contributed to force sensations. In a heavi-
ness matching task after deep paralysis of one arm, lifted
objects felt lighter rather than heavier compared with the
control arm. They proposed that this was due to a weak-
ened spindle signal as a result of an intrafusal motor termi-
nal blockade. It is known that when a muscle is deeply
paralyzed, sufficient to block motor terminals on both ex-
trafusal and intrafusal fibers, during the recovery period the
extrafusal fibers unblock first, with recovery of intrafusal
terminals being delayed (365, 427). During recovery, in the
face of rapidly rising muscle force, the persisting intrafusal
paralysis weakens the coactivation of spindles, and the
lower level of afferent discharge leads to a reduction in the
sense of force. In any case, the observations provided by this
experiment were the opposite of what would have been
expected if a signal of purely central origin had been respon-
sible. An explanation based on a spindle signal can also
account for the increase in perceived heaviness at the onset
of paralysis. Extrafusal terminals block first, leaving intra-
fusal terminals functioning to evoke a strong spindle signal
during a voluntary contraction, as the spindle signal is not
suppressed by any unloading effects from extrafusal con-
traction. As a result of the increase in spindle signal, per-
ceived heaviness increases.
In a weight-matching experiment on thumb flexor muscles
fatigued with a sustained MVC, after fatigue of the refer-
ence muscle it was matched by a lighter load than before
fatigue (233). This was again contrary to what would have
been expected if a central command signal had been respon-
sible for the sense of force. The load should have felt
heavier! A similar reduction in perceived heaviness was not
seen if instead of using an MVC to fatigue the muscle, a
smaller contraction of 40% MVC was used. However, for
the exercise at 40% MVC, if the muscle was vibrated during
the fatigue process, the reference load was again matched
afterwards by a lighter load. Here the interpretation was
that the recruitment of muscle spindles and tendon organs
by the vibration led to their desensitization. As a conse-
quence, during test matching, the afferent component of the
force signal decreased, leading to perception of a lighter
In a further experiment, weight matching was carried out in
two subjects who had a large-fiber sensory neuropathy (76,
371). In these subjects who lacked input from muscle spin-
dles and tendon organs in limb muscles, there was no op-
portunity for a peripheral contribution to the sense of heavi-
1676 Physiol Rev VOL 92 OCTOBER 2012
ness. While before fatigue both subjects were reasonably
accurate in weight matching, when thumb flexors of one
hand were fatigued to 50% of their control force, the ref-
erence load was reported to double in weight, as expected if
the sense of heaviness was generated by a proportional sig-
nal of purely central origin.
These observations (233) throw new light on the senses of
force and heaviness and point to further experiments. It is
proposed that in normal subjects a reafference is generated
in peripheral receptors in response to a motor command
during weight matching. The signal includes input from
skin receptors, muscle spindles, and tendon organs. When
the sense of force is disturbed by fatigue or weakness, this
can be accounted for by an altered peripheral signal. The
new results suggest that signals of purely central origin are
available but do not play a dominant role in normal sub-
jects. It remains uncertain what the exact mix is of signals of
peripheral and central origin for the sense of force. The
sense of heaviness is closely related but includes a compo-
nent of movement that will alter the responses of muscle
spindles and tendon organs. We typically move our hands
up and down during the estimation of the heaviness of
objects. Subjects are less accurate if such movements are not
allowed (245). However, only small movements are needed
to achieve accurate judgements (142).
In a force-matching task during fatigue, can a spindle signal
account for the amount by which the unexercised muscle
overestimates the level of force in the fatigued muscle? It has
been calculated (416) that a 10-fold increase in force in
ankle dorsiflexor muscles produced only a 2- to 3-fold in-
crease in spindle firing. Such a nonlinear relationship be-
tween spindle firing and tension could explain why the loss
of force in the fatigued muscle is not fully expressed in the
overestimate indicated by the unexercised muscle.
There is one further recent development. It has been possi-
ble to study some proprioceptive effects of the sense of
effort in isolation by using an experimental phantom of the
hand. The sense of position at the wrist was studied before
and after a total nerve block at the upper arm using a pres-
sure cuff. In the anesthetized and paralyzed arm, attempted
flexion and extension movements at 30% of maximum ef-
fort produced perceived displacements of the phantom
hand by 20° in the direction of the attempted movement
(148) (FIGURE 9B). Smaller displacements were perceived
when the arm was paralyzed but not anesthetized, presum-
ably because the available afferent signals influenced the
final perception (367).
The sense of movement was studied while subjects at-
tempted to move their paralyzed, anesthetized hand (398).
Contrary to earlier observations (246), sustained efforts
produced illusory movements. The position change accom-
panying an effort increased monotonically with both the
duration and size of the effort. Perceived movement velocity
only increased with effort, not its duration: a 10-fold in-
crease in effort producing a doubling in speed (398). It is
concluded that a sense of effort, arising centrally in associ-
ation with motor commands, is able to contribute to a num-
ber of proprioceptive sensations including the sense of limb
position and the sense of limb movement. Similarly, motor
command signals may contribute to the senses of force and
heaviness. We will have to wait for future experiments to
reveal the precise mix of signals of different sources that are
responsible for these sensations.
There is one further role of the sense of effort which is
almost outside the scope of this review. In looking at a hill,
subjects consistently overestimate its steepness. Judgement
of the slope is influenced by the subject’s level of fatigue and
whether or not they are wearing a heavy backpack (298).
Similarly, perceptions of distance are influenced by the sub-
ject carrying a load or after visuomotor adaptation that
reduces optic flow during walking (299; see also Ref. 90).
The proposal is that the perceived distance increases as the
effort required to walk it increases. These observations sug-
gest that the sense of effort has wider perceptual conse-
quences than just signaling muscle fatigue and the informa-
tion can contribute to signal our current physiological po-
tential to perform intended actions.
A. Introduction
In recent years there has been increased interest in the phys-
iology of exercise. One reason is that exercise has become a
part of many people’s lives as they strive to maintain fitness
in the face of largely sedentary life-styles. This is in the face
of the global obesity epidemic (the majority of the United
States population is overweight) with everyone focusing
attention on strategies to reduce adiposity and improve
health (e.g., Refs. 34, 189, 350). Questions arise about the
effectiveness of different kinds of exercise and the risks of
exercise-related injury. Finally, television has led to a rise in
popularity of spectator sports. The elite athletes who par-
ticipate in such sports are constantly in search of exercise
strategies that will give them a competitive edge. All of this
has meant that today there is growing interest in exercise
B. Exercise Disturbs Proprioception
It is common to feel awkward and clumsy after intense
exercise. It is not just muscle weakness from fatigue; we are
less sure about placement of our fatigued limbs if we are not
looking at them. This has led to the realization that exercise
can disturb proprioception, probably as a result of the ac-
companying fatigue. Muscle fatigue is not just a matter of
1677Physiol Rev VOL 92 OCTOBER 2012
peripheral mechanisms accompanying depletion of muscle
energy supplies, but includes activation processes at spinal
and cortical levels (for review, see Ref. 137).
It has long been known that limb position sense can be
disturbed by exercise (e.g., Refs. 35, 323, 342, 361). To
explore underlying mechanisms, it is first necessary to dis-
tinguish between different kinds of exercise. There are three
types of muscle contractions: concentric, isometric, and ec-
centric. In concentric exercise, the contracting muscle short-
ens, typically while under load. In isometric exercise, the
load is too heavy to allow any shortening and the muscle
contracts at constant muscle length. In eccentric exercise the
load is sufficient to forcibly lengthen the contracting mus-
cle. All three kinds of exercise, if carried out sufficiently
intensively, cause muscle fatigue and disturb propriocep-
tion. The loss of force from concentric exercise is largely
due to depletion of metabolic factors and is therefore short-
lasting, with recovery complete within 2 h (366). Eccentric
exercise is distinguishable from other forms of exercise
because it is accompanied, not only by metabolic fatigue,
but a component of the loss in force is due to muscle
damage. Force recovery after eccentric exercise is only
partial, and a significant deficit persists for up to a week,
the time taken to repair the damage. Accompanying the
muscle damage is delayed-onset muscle soreness (DOMS)
that also persists for about a week (for review, see Ref.
309). The damage from eccentric exercise raises the pos-
sibility that it may damage muscle proprioceptors and
that this was responsible for the disturbance to proprio-
ception. The soreness, too, can alter proprioception
(407), an effect presumably mediated by group III and IV
muscle afferents (e.g., Ref. 255).
Position sense, movement sense, and force sense have all
been studied before and after exercise. For position sense an
important consideration is the method of measurement.
Commonly a two-limb matching task is used (FIGURE 14).
Other methods include displacing a limb and asking the
subject to indicate its position with a pointer or asking the
subject to reproduce a previously remembered position.
With the two-arm or two-leg matching task, both limbs
contribute to matching accuracy (199, 412). As we use our
limbs in combination in everyday activities there is some
logic in measuring position sense in this way. However, it
makes the interpretation of results more difficult. For single
limb tasks, where the subject is asked to indicate a remem-
bered position, it introduces the extra variable of memory.
Measuring position sense in one limb with a pointer avoids
the problem of influences from the other limb but subjects
are less accurate. For all experiments aimed at measuring
position sense after exercise, an important consideration is
the way the test muscles are conditioned at the start of the
experiments, particularly if any measurements are made on
the relaxed limb (see sect. IIB3).
C. Concentric and Isometric Exercise
Skinner et al. (361) reported a decrease in the ability of subjects
to reproduce the position of the knee after flexion and exten-
sion movements. They attributed this decrease to changes in
muscle receptors (see also Ref. 66). The subject of propriocep-
tion at the knee and clinical aspects have been reviewed (e.g.,
Ref. 187; see also Ref. 224). In an experiment on elbow flex-
ors, fatigue of one arm using maximum voluntary contrac-
tions led to position matching errors (352). However, the er-
rors were not consistent between subjects, and there were no
“reciprocal” changes when the fatigued arm was the indicator
arm, compared with when it was the reference. It was con-
cluded that central as well as peripheral influences were re-
sponsible for the exercise effects. In related experiments,
Walsh et al. (399) found that both concentric and eccentric
exercise of elbow flexors produced position errors, although
here, too, the effects were not reciprocal. To examine the issue
of reciprocity, Allen and Proske (8) showed that if the errors
after concentric exercise were large enough, reciprocal effects
could be demonstrated reliably. However, while concentric
exercise altered position sense, it did not appear to impair the
ability to track the movement of a passively displaced arm. So
concentric exercise disturbs position sense but leaves move-
ment sense unaffected. In that context, it has been claimed that
movement detection thresholds at the shoulder joint are sig-
nificantly raised by fatigue from exercise (59), while the acuity
of movement sense is reduced (287). Finally, it has been
claimed that concentric exercise disturbed position sense to a
greater degree than isometric or eccentric exercise (124).
D. Eccentric Exercise and Damage to
Muscle Receptors
The first study to use eccentric exercise to examine its effects
on position and force sense was by Saxton et al. (342). Both
joint position sense and force sense were impaired after the
exercise. To account for their observations, the authors con-
sidered the possibility of disturbance to muscle proprioceptors
by the exercise. In a related experiment, Brockett et al. (35)
similarly suggested that eccentric exercise damaged muscle re-
ceptors. This was against a background of evidence indicating
that eccentric exercise damaged muscle fibers. These proposals
prompted animal experiments in which responses of tendon
organs (163) and muscle spindles (166) were tested before and
after a series of eccentric contractions. After exercise which
reduced muscle force to 50% of its control value, responses of
both receptor types to length and tension changes remained
essentially unaltered. This is shown in FIGURE 15. There was
no evidence that the intrafusal fibers of muscle spindles were
susceptible to the damaging effects of eccentric contractions,
unlike their extrafusal neighbors. Presumably the compliant
ends of spindle attachments to the adjacent endomysium pro-
tected them from damage.
These considerations led to the conclusion that answers to
the question of the source of the disturbance to propriocep-
1678 Physiol Rev VOL 92 OCTOBER 2012
tion by exercise had to be sought outside the muscle. How-
ever, one report was not consistent with that conclusion
(320). Measurements were made of the vibration illusion
(159) in ankle extensor muscles before and after a series of
stretch-shortening cycles. Two days after the exercise, vi-
bration of the exercised muscle led to a reduced vibration
illusion for the frequencies of 80–100 Hz, but an increased
illusion at 40 Hz. It was suggested that the delayed recovery
from the exercise was associated with a decreased dynamic
sensitivity of spindle primary endings.
E. Exercise and the Sense of Effort
The question of whether the sense of effort could contribute
to position sense was tested using both concentric and ec-
Indicator arm
Actual position
Perceived position
Reference arm
Reference leg
Indicator leg
Actual position
Perceived position
Position Matching
FIGURE 14. Position matching at the elbow and knee before and after fatiguing exercise. Top left: position
matching at the elbow. Blindfolded subjects sat at a table with their forearms strapped to lightweight paddles. The
hinges of the paddles were aligned with the elbow joint and had potentiometers attached to them, providing a
voltage signal proportional to elbow angle. Top right: position matching at the knee. Subjects were seated in a chair
mounted on a steel frame. The height of the chair was adjusted so that the medial and lateral epicondyles of the
knee were in line with the pivot point of the position matching apparatus. This consisted of a pair of paddles lying on
the shin with potentiometers at their hinge points giving voltage outputs proportional to knee angle. Bottom panels:
perceived positions of the elbow (left) and knee (right) after a series of fatiguing concentric contractions that led to
a 30% fall in torque in elbow flexors and knee extensors, respectively. The exercised arm felt more extended than
it really was, the exercised knee more flexed. [Redrawn in part from Allen et al. (7).]
1679Physiol Rev VOL 92 OCTOBER 2012
centric exercise (5, 397, 399). After eccentric exercise of
elbow muscles, subjects matched the position of the exer-
cised arm by placement of the unexercised arm in a more
extended position (399). The size of the errors correlated
with the drop in force. A similar pattern was seen after
concentric exercise. It was proposed that to hold the fa-
tigued arm at the test angle against gravity required more
effort, and this was interpreted as the arm having adopted a
more horizontal position than was actually the case. This
view was supported by observations in subsequent experi-
ments (397) in which, in one condition, arm position
matching was done using a counterweight to balance the
weight of the arm. No effort was required to hold the arm at
the test angle. In a second experiment, position matching
was done in the horizontal plane to remove the effect of
gravity. After eccentric exercise, position matching errors
increased significantly for unsupported matching, but not
for counterweighted matching or matching in the horizon-
tal plane.
Despite these apparently unambiguous results, some uncer-
tainty remained about the possible contribution from the
sense of effort to position sense, because of uncertainties
over the state of conditioning of muscles at the start of each
trial. The point was addressed in knee extensor muscles,
with care being taken to ensure that for each measurement
the muscle was in a defined state (155). Significant position
errors were produced by both eccentric and concentric ex-
ercise, but they were in the opposite direction to that pre-
dicted if gravity had played a role. That is, the position of
the exercised leg was matched by the unexercised leg adopt-
ing a more flexed posture at the knee, not a more extended
posture. In contrast, for exercise of elbow flexors, the errors
were in the direction of arm extension (5) (see FIGURE 14).
The result with knee extensor muscles (155) has been con-
firmed (283). In a repositioning task of a remembered po-
sition after eccentric exercise, subjects placed their exer-
cised leg in a more extended position, representing a shorter
knee extensor muscle. The result suggested that subjects
believed their exercised muscle was longer than it really
was, so they placed it in a position that shortened the mus-
cle. Finally, another recent study on the same muscle group
found errors in the same direction, but these were signifi-
cant only in non-weight bearing tasks (392).
F. Mechanism of Exercise Effects on
Position Sense
A possible explanation for the above observations was that
fatigue of a muscle had increased the discharge rate of its
muscle spindles and that led to a shift in perceived limb
position representing a more stretched muscle. To obtain
further information about the distribution of the errors,
rather than exercising elbow flexors, elbow extensors were
exercised (7). The unexpected result was that, as for elbow
flexors, the arm still felt more extended than it had before
the exercise. A similar result was obtained at the knee. Ex-
ercise of knee flexors led the leg to be perceived as more
flexed than its actual position, the same result as after ex-
ercise of knee extensors. Any direct involvement of muscle
receptors in the observed position errors was ruled out by
this result because, if the errors had been generated within
the muscles, they should have been in opposite directions
for exercise of antagonist muscles. Finally, exercise of finger
flexors did not lead to any position errors at the elbow, and
Tendon organ
Spindle primary
Before exercise
Before exercise
5 s
10 s
20 mm
20 N
20 mm
20 N
FIGURE 15. Responses of muscle spindles and tendon organs
before and after a series of eccentric contractions. A: response of a
single spindle primary ending to a slow stretch of the soleus muscle
of the anesthetized cat. Responses are shown as instantaneous
rates. Response in blue is before exercise, and responses in red are
after exercise. The trace at the bottom (black) represents muscle
length, and the two traces above represent muscle passive tension
before exercise (dashed blue line) and after exercise (continuous red
line). The eccentric contractions led to a rise in whole muscle pas-
sive tension. [Redrawn from Gregory et al. (166), with kind permis-
sion from Springer Science and Business Media.] B: responses of a
single tendon organ of the soleus muscle of the anesthetized cat to
a slow stretch before and after exercise. Traces are as in A. The
tendon organ remains a true tension sensor, responding to the rise
in passive tension from the eccentric contractions. [Redrawn from
Gregory et al. (163).]
1680 Physiol Rev VOL 92 OCTOBER 2012
thus the exercise effect was joint specific and did not spread
across several joints (7).
Additional control experiments have recently been carried
out (Leung, Allen, and Proske, unpublished observations).
If, instead of exercising elbow flexors, the passive elbow
was moved through the same angle range, over the same
period of time as during the exercise, but without any active
contractions, there were no effects on position sense. So it
was not the repeated movements that were responsible. Al-
ternatively, if elbow muscles were exercised isometrically,
without any accompanying overt movement at the elbow
joint, provided the test muscle was sufficiently fatigued, it led
to position errors. There was one final consideration: Was the
fatigue effect a consequence of the reduced force generating
capacity of the muscle, or was it the result of the accumulation
of metabolites that accompanies fatigue from exercise? The
influence of metabolites could arise peripherally, presumably
mediated by small-fiber afferents or, alternatively, it could ex-
ert effects within the brain (e.g., Ref. 275).
Concentric exercise produces both a fall in force and the
accumulation of metabolites. Typically fatigue from con-
centric exercise is fully reversed within 2 h after the exercise
(366). When measurements of force are made immediately
after eccentric exercise, there will be a component of the
force loss due to metabolic factors. However, a second com-
ponent will be present from damage to muscle fibers, and
this takes days to resolve. In most previous experiments
testing the effects of eccentric exercise on position sense, the
exercise was not made severe enough for a significant force
drop to persist at 24 h post exercise. In a new study, sub-
jects’ elbow flexors were eccentrically exercised, the exer-
cise being made sufficiently severe (50 60% fall in force
immediately after exercise) to ensure that a significant force
drop persisted at 24 h (25–35%) (Tsay, Allen, Leung, and
Proske, Exp Brain Res. In press). After 24 h, significant
position errors accompanied the fall in force. In other
words, at a time long after the exercise was completed, the
effects on position sense were still present. This made it
unlikely that local effects of metabolites were responsible.
One possibility was that the fall in muscle force had altered
the body schema. This change in the representation would
persist for as long as there were any sensory cues indicating
that recovery from fatigue was not complete.
After exercise of elbow muscles, the forearm is perceived as
more extended, and after exercise of knee muscles, the
lower leg is perceived as more flexed (FIGURE 14). The di-
rections of these errors can be incorporated in the hypoth-
esis by suggesting that the changes in the central map are
given by the direction in which gravity is exerting its effect
on the limbs. It is as though the brain is compensating for
the weaker limb muscles by shifting the position of the map
into a more nearly gravity-neutral posture. For the arm it is
in a direction towards a posture where the arm is hanging
by the side of the body, and for the leg, it is in a direction
where the lower leg is hanging vertically.
How can a change in the motor command:force relation
provoked by fatigue alter a central map? The work on
phantom limbs (314) suggests that motor commands, in the
absence of any afferent feedback from the arm, can lead to
sensation of a distorted, contracted phantom arm with
phantom limb pain. This pain can be relieved with mirror
box therapy. The mirror provides the missing afferent feed-
back (vision of the intact arm) convincing the subject that
their phantom arm can be relaxed and straightened. So,
motor commands can alter the body map, at least for the
phenomenon of phantom limbs. For the effects of exercise
there is, of course, no loss of afferent feedback. So in that
sense, it is unique.
Why place such importance on the disturbance to position
sense after exercise? Interest is driven by the possibility of
fatigue-related proprioceptive errors leading to injury. The
marathon runner who staggers into the stadium at the end
of the run is unsteady on his feet not just because his leg
muscles are fatigued but because he is unsure of where his
legs are. The finding that exercise of knee muscles leads to
perception of a more flexed knee (155) may mean that tiring
runners will overextend their knee, raising the possibility of
strain injuries in knee flexors (277). In the elderly (see sect.
VII), the progression of muscle frailty due to sarcopenia
implies reduced proprioceptive control (50) so that any dis-
turbance to proprioception in leg muscles from unaccus-
tomed exercise may increase the risk of falls. Given the
adaptability of skeletal muscle, selected training programs
may help to reduce the disturbance to proprioception from
unaccustomed exercise and thus reduce the risk of injury.
G. Exercise and Force Sense
The effect of eccentric exercise on force sense has been
studied by a number of groups (35, 62, 342, 404). The main
finding is that in a bilateral force matching task in elbow
flexors, if subjects are instructed “to make both arms feel
the same,” after the exercise they will overestimate the force
in the exercised arm (see sect. V, FIGURE 13). Errors attrib-
utable to the drop in force after exercise can be taken into
account by expressing them, not in terms of the preexercise
MVC, but the postexercise MVC. For example, a mismatch
of 14% expressed in terms of the preexercise MVC reduces
to 4% when fatigue is taken into account (404). This rem-
nant difference was attributed to a change in muscle EMG
(404). It is known that after eccentric exercise the EMG for
a given level of force increases (62, 351, 404). Thus a larger
neural drive was required to achieve a given level of force. In
addition, Carson et al. (62) observed a larger MEP in re-
sponse to magnetic brain stimulation of motor cortex after
eccentric exercise. That led to the suggestion that after ec-
centric exercise there was an increase in gain of the effort:
1681Physiol Rev VOL 92 OCTOBER 2012
motor command relation such that a given level of effort
resulted in greater excitability of motor cortex and greater
muscle activation. In that context, it has been shown that
single stimuli to the motor cortex are not associated with
any sensations of effort (e.g., Refs. 112, 139), but this may
simply reflect the unphysiological nature of the electromag-
netic stimulation. A similar observation was made on a
deafferented subject, and here the possibility was consid-
ered that the transcranial magnetic stimulus might itself
have inhibited any sensation (76).
Another factor to consider after eccentric exercise is DOMS
which begins 24 h after the exercise. Does the muscle pain
from eccentric exercise produce force matching errors? The
pain is mediated by group III and IV afferents (see sects. VB
and VIC). Noxious muscle input generated by injection of
hypertonic saline into muscle reduced excitability at the
level of the motor cortex (225, 239). Therefore, in the pres-
ence of pain, it should require a larger effort to generate a
given level of force. As a consequence, in an isometric force
matching task, the sore muscle would generate less force
(e.g., Ref. 307). The question of the roles of peripheral and
central factors, as well as pain in producing the force loss
after eccentric exercise, was addressed by Prasartwuth and
colleagues (295, 296). They concluded that eccentric exer-
cise produced a length-dependent impairment of voluntary
activation of the muscle, and this contributed to the loss in
maximal voluntary force. However, soreness did not con-
tribute to the early force loss (407). Note that because of the
recent evidence for a peripheral contribution to force sen-
sation (233), studies of changes in perceived force with pain
and fatigue may need to be reevaluated.
A. Introduction
“Even a long human life adds up to only 650,000 hours.
And when that modest milestone flashes into view. . . for
reasons unknown, your atoms will close you down, then
silently disassemble and go off to be other things. And that’s
it for you.” (40). With the effectiveness of modern medicine,
most human populations are ageing, and age-related prob-
lems are becoming a progressively larger proportion of na-
tional health budgets. One of the most important problems
for the elderly is injury from falls (e.g., Refs. 15, 358, 372).
At least one-third of people aged 65 years or older fall one
or more times a year, with a significant proportion requiring
hospitalization. Proprioception is a critical factor in stable
standing (120, 230, 232), and both deteriorate with age (e.g.,
Refs. 50, 231, 232; see also Ref. 324). As people age they
become more frail. One expression of frailty is muscle weak-
ness from sarcopenia (for a review of sarcopenia, see Ref.
272). There is a correlation between muscle strength and pro-
prioceptive acuity in the stability of standing (50), with acuity
diminishing at high levels of muscle force. Thus an impor-
tant reason for the increase in falls in the elderly is a dete-
rioration in proprioception, and this has important clinical
and public health implications (see also sect. VI).
B. Sarcopenia
The term sarcopenia (337) refers to the loss of muscle mass
associated with ageing, but is usually meant to include the
age-related loss of muscle strength or performance. The
decline is greater in the lower limbs. Sarcopenia is associ-
ated with a loss of motor units (38, 389). There may be a
greater loss of units in lower limb muscles given that there is
evidence that distal limb muscles show a bigger decrease in
units than proximal muscles (132). The loss of motor units
decreases muscle cross-sectional area. Loss of contractile
tissue is accompanied by infiltration by fat and connective
tissue. These changes reduce muscle power that can be re-
duced by as much as 60% in elderly subjects (379).
Throughout life, skeletal muscle undergoes a continuous
cycle of denervation and reinnervation, but in old age it
seems the process of reinnervation cannot keep pace with
that of denervation, leading to a loss of motor units. This
appears to be related to a decline of neurotropic factors
concerned with motoneuron survival (174). In addition,
there is evidence of apoptosis of skeletal myocytes (94). The
evidence suggests that regular, intense physical activity can
keep age-related loss of muscle mass at bay by leading to
fiber hypertrophy, but it is unable to reverse the loss in fiber
numbers as a result of neuropathic processes (272).
C. Muscle Spindles and Aging
As muscle spindles play a key role in kinesthesia, does sar-
copenia affect muscle spindle responses? First, the changes
with age in muscle architecture are likely to impact on re-
sponses of spindles. Changes in fibers lengths, pennation
angles, and tendon compliance (273) are all likely to alter
spindle responses to passive and active lengthening. Changes
in tendon properties are likely to alter tendon organ re-
sponses and therefore change the relation between muscle
tension and tendon organ discharge. The losses of muscle
fibers and motoneurons that are a part of sarcopenia could
include intrafusal fibers and fusimotor neurons. In aged
human peripheral nerve (tibial), there is a significant fall in
numbers of nerve fibers. This includes a bigger proportion
of large-diameter fibers (375). Some of these could be sen-
sory fibers. Therefore, the possibility exists that aged muscle
contains denervated spindles. There is some evidence on
this point.
Spindles dissected from aged muscles have fewer intrafusal
fibers (376). Other structural abnormalities were consistent
with the signs of denervation atrophy. In addition, spindles
in aged rats may contain intrafusal fibers without any sen-
1682 Physiol Rev VOL 92 OCTOBER 2012
sory innervation altogether (91). Kim et al. (215) showed
that in aged rats the number of the sensory endings in each
spindle was about normal, but most primary endings had
lost their typical annulospiral configuration. They were ta-
pered and irregular in shape. Comparable abnormalities
could not be detected for secondary endings. This leads to
the question of the responsiveness of spindles in aged mus-
Responses have been recorded from spindles in muscles of
rats of different ages (215, 261). As expected, in muscles of
young animals, the conduction velocity of presumed pri-
mary spindle endings was significantly higher, and they had
larger diameter axons than the presumed secondary end-
ings. In contrast, in aged animals, the conduction velocity
and axon diameter spectra had shifted in the direction of
slower speeds and smaller axons. There was no longer a
clear difference between primary and secondary endings.
Dynamic responses of spindle primary endings in aged an-
imals were much lower than in young animals, including
responses to the depolarizing drug succinylcholine. The re-
sponses of aged primary endings resembled those of second-
ary endings, probably as a result of loss of their annu-
lospiral terminal structure.
These observations on muscle spindles have potentially im-
portant implications for proprioception. Given that pri-
mary endings of spindles contribute both to the sense of
limb position and the sense of limb movement (see sect. II),
the observations of Kim et al. (215) raise the possibility that
elderly human subjects may have a reduced sense of move-
ment as a result of a degraded dynamic sensitivity of spindle
primary endings. There is some psychophysical evidence on
this point. Human subjects were asked to discriminate be-
tween joint angles during movement of the ankle joint over
the range, 15–30°/s. Normal elderly subjects produced
larger errors than young adults. However, unexpectedly,
vibration of ankle dorsiflexors led to larger errors in the
elderly. It was concluded that there was no change with age
in the responses of spindles to vibration (391). However, in
a more recent report, dynamic position sense declined with
age if ankle rotations were extended beyond 30°/s to 90°/s,
to cover the full range of walking speeds over which sub-
jects are likely to fall (235). Another, more recent, observa-
tion of the effects of calf and neck muscle vibration sug-
gested that neck vibration produced a greater postural dis-
turbance in the elderly (284). Here, however, there are
potentially confounding influences from vibration affecting
the vestibular system. To conclude, while evidence has been
provided that muscle spindle structure and function change
with age, supporting observations of functional outcomes
remain limited.
D. Falls in the Elderly
Proprioception plays an important role in stable standing
(121, 122). This was shown by measuring body sway in
subjects whose bodies had been splinted, and they balanced
an equivalent weight with their feet. Under these condi-
tions, in normal subjects proprioception in leg muscles pro-
vided sufficient information to allow for a stable upright
stance. Vision and the effect of the segmented body were
important for maximal stability, while sensory input from
the feet had a smaller effect (36, 197, 210). It was concluded
that during simple standing vestibular input did not con-
tribute to body sway.
In studies of sensorimotor functions and balance under nor-
mal standing conditions in groups of subjects of different
ages, all of the sensory functions measured, particularly
lower limb sensations, showed age-associated differences in
their influence on standing stability, measured as body sway
(230, 231; see also Ref. 411). When conditions were made
more challenging by asking subjects to stand on foam, vi-
sion, strength of the quadriceps muscle group, and reaction
times to a visual stimulus played significant roles. In addi-
tion, standing on foam with the eyes shut revealed a vestib-
ular component. Up to the age of 65 years vision played a
major role in stable standing, but as visual acuity declined,
it became less important. In a cohort of elderly women, the
incidence of multiple falls was associated with poor posi-
tion sense at the knee, reduced cutaneous vibration sensi-
tivity, impaired lower limb strength, slower reaction times,
and low-contrast visual acuity (232; see also Ref. 236).
It is a common wisdom that exercise improves postural
stability (e.g., Ref. 292), perhaps as a result of the increases
in muscle strength. As we grow older and come more and
more under the influence of sarcopenia, it raises the ques-
tions, What is the relationship, if any, between muscle
strength and proprioception and can their decline with age
be offset by exercise? The relationship between muscle
strength and standing stability is not straightforward (50).
Muscle weakness, measured as ankle dorsiflexor strength,
by itself, does not reduce standing stability. However, there
was reduced proprioception in the weaker subjects, as mea-
sured by the amount of body sway, when they were asked to
stand with their eyes shut. This was the case despite there
being no differences in sensory functions between the weak
and strong groups, including a measure of knee position
sense, although the position sense data need to be con-
firmed. In a group of older subjects (70 yr), this relation-
ship between muscle strength and proprioception was not
apparent. In these subjects visual acuity and knee joint po-
sition sense had deteriorated significantly, and it was con-
cluded that older subjects relied more on the cues provided
by a range of sensory inputs and not just vision and propri-
oception to achieve stable standing. Nevertheless, in sub-
jects of all ages, weaker leg muscles mean poorer proprio-
ceptive control in the absence of vision. Why that might be
so remains unclear, but it implies an important link between
the contractile and sensory functions of muscles, a link
which is likely to apply to all skeletal muscles. Perhaps, the
1683Physiol Rev VOL 92 OCTOBER 2012
need to drive the motor system harder, because of weak-
ness, compromises the signaling capacity of the propriocep-
tive system (see FIGURE 16).
To conclude, it is clear that proprioceptive inputs are an
important determinant of stable standing. The evidence
suggests that there is not only a loss of muscle strength with
age, but this is accompanied by a deterioration of the sig-
naling capacity by muscle receptors. That, in turn, implies a
proprioception-related increase in the incidence of falls.
The specific relationship between muscle strength and pro-
prioception should be explored further as it may provide a
basis for the claim that exercise improves standing stability.
Interestingly, to achieve a reduction in the incidence of falls,
it is not sufficient to improve muscle strength alone as ex-
ercises are required which actually challenge standing sta-
bility (357).
A. Clinical Tests of Proprioception
Tests of proprioception are one important component of
the standard neurological examination. Proprioception
in the lower limbs is measured using movement detection
thresholds at the big toe. Position sense in the upper limbs
is determined by the patient’s ability to maintain their arm
in an extended position without undue positional drift. Fi-
nally, there is the well-known Romberg’s sign, maintenance
of an upright posture with the eyes closed.
The Romberg sign is probably the most commonly used
diagnostic test for proprioceptive abnormalities. The test is
attributed to M. H. Romberg, a German neurologist prac-
ticing in the mid-19th century (334). He presented the first
detailed account of tabes dorsalis, a demyelinating syphi-
litic condition. Romberg described the loss of balance dem-
onstrated by the erect patient when he closes his eyes while
standing with his feet together. This behavior is to be dis-
tinguished from the condition in which the patient already
sways with his eyes open, a symptom typical of cerebellar
For the lower limb, a decision about whether there is any
proprioceptive deficit is based on a comparison of the move-
ment detection thresholds of the big toes. Unfortunately,
detection thresholds at the big toe, when expressed in terms
of proportional changes in fascicle length for the muscles
acting at the big toe, are much poorer than for the ankle,
knee, or hip joints (317). The reason for this is the anatomy
of the foot and ankle, leading to poor mechanical coupling
between the toe and the muscles that operate it (318). It is
important to keep this in mind when comparing the pa-
tient’s proprioceptive performance at different joints.
Another important consideration when making measure-
ments of position and movement sense is to take into ac-
count any thixotropic effects that may be present in the
muscle or its spindles (see sect. IIB3). Particularly when
making measurements of position sense in the relaxed limb,
before the measurement is made, it is advisable to ask the
subject to contract their muscles isometrically at the test
B. Parkinson’s Disease
One common neurological condition in which there have
been claims of disturbed proprioception is Parkinson’s dis-
ease. Such a view is supported by the casual observation of
a Parkinsonian patient while they are eating. Provided the
person does not have too pronounced a tremor, these pa-
tients are able to eat liquid foods with a spoon, but they
have difficulty in locating their mouth unless they concen-
trate on visually tracking the arm movement. Such behavior
suggests not just disturbed proprioception but implies a
derangement of spatial maps of the body.
The experimental evidence for a disturbance of propriocep-
tion in Parkinson’s disease has not yet revealed a clear pic-
ture. It has been reported that illusions of movement and
displacement of a limb evoked by muscle vibration are nor-
mal in Parkinson’s disease (e.g., Refs. 265, 335). It implies
that muscle receptor responses to the vibration are normal.
That view has been backed up by microneurographic re-
cordings of muscle spindle responses in patients with Par-
kinson’s disease (48, 237). It leads to the conclusion that if
proprioception is disturbed in Parkinson’s disease, it is not
due to malfunction of the peripheral sensory apparatus but
results from problems with central processing of proprio-
ceptive information.
In a forced-choice comparison of elbow angles by patients
with Parkinson’s disease they achieved fewer correct
matches than normal subjects (428). In another test, pa-
tients’ wrist extension movements were monitored during
vibration of the flexor tendon. Vibration led to a smaller
undershoot of the target than it did in healthy control sub-
jects (326). It was concluded that in Parkinsonian patients
there was an abnormality in the higher level integration of
proprioceptive information. One possible source for such
an abnormality was in the comparison between afferent
feedback and the corollary discharge of the motor com-
mand (264). It has been suggested that there is a saturation
of central sensory processing in Parkinson’s disease (326), a
view that is supported by animal observations. An animal
model of Parkinson’s disease can be produced by injection
of methyl phenyl tetrahydropyridine (MPTP). In monkeys
with symptoms of Parkinson’s disease produced by MPTP,
pallidal neurons had increased sensitivity and reduced se-
lectivity in their responses to peripheral sensory inputs
1684 Physiol Rev VOL 92 OCTOBER 2012
Motor cortex Sensor y cortex
of afferent
Nucleus Z
Clarke’s column
Spindle afferent
Anterior cerebellum
Forward model
sensory feedback
FIGURE 16. Possible mechanism (shown in diagrammatic form) for how the brain distinguishes between
impulses coming from muscle spindles that are generated by muscle stretch (exafference) or by fusimotor
activity (reafference). When the muscle is stretched, spindle impulses travel to sensory areas of the cerebral
cortex via Clarke’s column, the dorsal spinocerebellar tract (DSCT), Nucleus Z, and the thalamus (shown in
red). Collaterals of DSCT cells project to the anterior cerebellum. When a motor command is generated, it
leads to coactivation of skeletomotor and fusimotor neurons (shown in blue). A copy of the motor command is
sent to the anterior cerebellum where a comparison takes place between the expected spindle response based
on that command and the actual signal provided by the DSCT collaterals. The outcome of the match is used to
inhibit reafferent activity, preventing it from reaching the cerebral cortex. Sites of inhibition could be at Nucleus
Z, the thalamus, or the cortex itself. For further details, see text.
1685Physiol Rev VOL 92 OCTOBER 2012
A feature of Parkinson’s disease is bradykinesia. Patients
with bilateral asymmetry in expression of the disease over-
estimate the trajectory of the more affected limb in a match-
ing task using slow, active movements of the arms (264).
One interpretation of this result is a lower-than-normal
central modulation of the afferent input, perhaps the result
of the previously postulated saturation mechanism.
C. Schizophrenia
As described in section IVC, we know that in a subject with
a paralyzed, anesthetized arm attempts to move the hand at
the wrist leads to illusory sensations of movement (148,
398). This observation shows unambiguously that central
motor mechanisms are, by themselves, able to generate kin-
esthetic sensations in a normal subject. They are therefore
available to contribute to proprioceptive sensations during
load bearing as well as being potentially capable of gener-
ating phenomena such as the phantom limb. Subjects over-
estimate externally generated forces by 20% when they gen-
erate them themselves (353, 402) (FIGURE 12). Similarly,
attenuation of the muscle sense has been reported during
active movements (77). Such suppression mechanisms may
allow us to distinguish between self-generated and exter-
nally generated actions; that is, during active movements,
the efference copy is used not only for predicting the reaf-
ference (FIGURE 11), but it gives rise to a sense of body
“ownership”. Such a sense appears to be reduced in schizo-
In Parkinson’s disease the patient fails to transform a willed
intention into action, leading to symptoms of bradykinesia,
and this is alleviated by giving dopamine. Schizophrenics
who complain that they are under alien control appear to
have a reduced awareness of self-initiated motor acts that
they have just carried out (127, 129). The failure to form a
conscious percept for an intended action leads to the nega-
tive symptoms of schizophrenia (e.g., Ref. 128). Thus a
breakdown of the awareness of generation of willed acts
can have profound consequences.
D. Clinical Symptoms of Fatigue
The subject of fatigue is extensive and complex because it
includes both central and peripheral components (for re-
view, see Ref. 137). Fatigue is relevant to a review of pro-
prioception because sensations of fatigue are typically ac-
companied by increases in the sense of effort. Throughout
this review the term effort has been used to imply sensations
generated by motor commands, in isolation, or in combi-
nation with inputs from the periphery. An abnormal sense
of effort therefore has a place in clinical considerations of
disturbed proprioception. The fatigue felt as a result of
mental effort, mental calculations, etc. is not included here.
Many pathological conditions are accompanied by abnor-
mal perceptions of effort and fatigue and are therefore rel-
evant to a discussion of proprioception. These include cor-
ticospinal lesions, spinal cord injury, and multiple sclerosis
(for a review, see Ref. 137). The influence of joint pathol-
ogy, particularly when associated with arthritis, can reduce
voluntary strength and proprioception (388), perhaps as a
result of the action of the nociceptive input generated by
joint movement (179). Twitch interpolation, when used
with patients with chronic fatigue syndrome, fibromyalgia,
and heightened effort syndromes has led to rather variable
results (e.g., Refs. 212, 228, 259). So, the clinical picture for
patients exhibiting symptoms of fatigue syndromes, with or
without other accompanying pathologies, remains unclear,
at least with respect to the role played by proprioception.
A. To Summarize and Look to the Future
It is interesting to compare the material included in the
present account with that covered in a similar review 35
years ago by McCloskey (244). That account not only pro-
vided an up-to-date picture of the state of knowledge at the
time, but it included a comprehensive review of the histor-
ical background. While in the present review some histori-
cal material has been included, it is not as detailed or wide-
ranging. Much of the background provided by McCloskey
is relevant to the present review, and the reader is encour-
aged to consult it.
In looking back to the state of knowledge in 1978, it is
astonishing to see the increase in breadth of coverage that
has been necessary since then to include the many recent
developments. An obvious example of an area that did not
exist 35 years ago and where rapid progress is currently
being made, is in the application of neuroimaging tech-
niques and their application in different experimental con-
texts (e.g., Refs. 83, 177, 270, 289). As imaging techniques
are further refined, and new methods are introduced, such
as diffusion tensor imaging (202), more progress will be
made in our understanding of the brain areas concerned
with the processing of proprioceptive information and how
proprioceptive inputs are combined with information com-
ing from the other senses during normal movements.
Another important development is that derived from obser-
vations on motor cortical neurons (e.g., Refs. 151, 152). It
has brought about the realization for the kinesthetic senses
that the brain is not concerned with information about
muscle length changes provided by individual afferents, but
with the population of muscle afferent signals arising in
groups of muscles. Furthermore, studies on second-order
neurons point to the regulated property as being limb move-
ment end-point (28) not the disposition of limb segments,
one relative to another. Here it is interesting to reflect on the
fact that there is already significant processing of the infor-
mation at the first relay in the central projection pathway of
1686 Physiol Rev VOL 92 OCTOBER 2012
proprioceptive afferents. The population approach has
been successfully applied to disturb limb muscles with vi-
bration to generate perceptions of complex movements
At the time that McCloskey wrote his review, the important
concepts of efference copy, reafference, and exafference
were already known. However, they had not yet been fully
integrated into the thinking about proprioceptive mecha-
nisms. The idea that afferent information generated as a
result of motor commands is processed differently from
activity generated by external stimuli has only recently be-
gun to be exploited more directly (86, 87). Models have also
been generated to account for the events underlying the
generation of predicted information in association with mo-
tor commands for movements, the forward models, and
how these might interact with activity arising in the periph-
ery (19). Such models may have to operate differently de-
pending on whether the regulated property is posture or
movement (65). This is another area where it is likely that
further important steps forward will be taken in the future.
How is the outcome of a comparison processed and what
are the sensory consequences?
One area of proprioception where entirely new proposals
have been put forward concerns the senses of effort, force,
and heaviness (233). Here the ideas of von Holst and Mit-
telstaedt (393) have been directly applied to the problem.
The currently prevailing view is that the motor command is
directly involved in the generation of sensations of effort,
force, and heaviness. Luu et al. (233) have taken a broader
view and proposed that the reafferent signal generated by
the motor command plays a dominant role in generating the
sensation. The peripheral input is proposed to include sig-
nals from muscle spindles that have traditionally been
thought of as length and movement sensors, not force sen-
sors. If future experiments support such a view, it will re-
quire a rethink of views about the role of muscle spindles
and the fusimotor system. Most current theories of the role
of the fusimotor system have focused on the reflex action of
muscle spindles and how this can be supported by fusimotor
activity. The prospect of fusimotor-mediated spindle dis-
charges generating proprioceptive sensations represents a
significant departure from current views of the role of spin-
dles in motor control.
The emergence of proposals for a reafferent contribution to
force and effort sensations comes at a time when it has
become possible to study sensations arising entirely cen-
trally (148, 398). These findings are supported by observa-
tions made on subjects with a large-fiber sensory neuropa-
thy (233). So, presumably, we routinely have available the
capacity to access sensations generated by signals arising
from motor commands, but in everyday situations it re-
mains uncertain what is the precise mix of signals contrib-
uting to the senses of position and movement during load
bearing, the senses of force, effort, and heaviness. It is in-
teresting that chronically deafferented subjects do not ex-
perience phantom limb phenomena and are able to imagine
movements visually, without feeling them. Any central
sense of motor command appears to fade without regular
sensory recalibration.
Another important concept that has evolved rapidly is the
idea of a body image and a body schema, central maps to
which incoming proprioceptive information is referred for
determination of starting points and endpoints during the
generation of movements. The body schema is a labile map,
constantly updated by the immediately previous afferent
input (e.g., Refs. 108, 147, 198), yet remarkably stable, as
shown by persistent phantoms after limb amputation. The
other is a cognitive representation based on stored knowl-
edge and used for perceptual judgements. Recent observa-
tions suggest an additional map incorporating the body’s
metric properties (229). Interestingly the map is distorted
and crudely resembles the maps drawn years earlier by Pen-
field and colleagues using electrical stimulation. Yet sub-
jects consciously perceive only the undistorted map. These
findings emphasize that there must be multiple representa-
tions of body form, some accessible to consciousness, some
unconscious. It will be interesting in the future to explore
further the different body forms and their distortions pro-
duced experimentally. The phantom limb phenomenon em-
phasizes the potency of central processes in maintaining
representations of body form. A question for the future
concerns the integration of the different body parts into a
perception of the body as a whole.
Other recently expanding areas include the study of propri-
oception after fatigue from exercise and the realization that
some of the clumsiness in movements experienced follow-
ing intense exercise may be proprioceptor-mediated. In our
ageing society, an important topic where there is a growing
need for further progress is in documenting the changes in
proprioception with age. The evidence points, at least in
part, to declining proprioception as responsible for the in-
creased incidence in falls in the elderly, a major public
health issue. Finally, the knowledge obtained from experi-
ments on proprioception has provided new insights into
pathological aspects and allowed more objective assess-
ments in the clinic. There are likely to be further important
advances in the clinical arena.
B. Some Speculations
As proprioceptive sensations, the senses of effort, force, and
heaviness are distinct in that they are always associated with
sensory input reafferent in origin, being generated exclu-
sively by motor commands. The same is not true of the
senses of limb position and movement. Here, as Sherrington
noted (355), kinesthetic sensations can arise in the passive
limb, in the absence of any motor commands, that is, the
1687Physiol Rev VOL 92 OCTOBER 2012
sensory information can be exclusively exafferent in origin.
Yet we can also accurately perceive the position and move-
ment of our limbs when these are accompanied by volun-
tary muscle contractions, for example, moving our arm
when it is supporting a load. So position and movement
signals can provide accurate kinesthetic information, re-
gardless of whether they are of exafferent or reafferent or-
igin. It raises the question, How does the brain extract the
relevant kinesthetic information from a reafferent signal?
Differences in the central processing of self-generated and
externally generated afferent activity have been discussed
previously in relation to the proprioceptive system (242),
vestibular system (87), the tactile system (25), and the elec-
trosensory system in fish (262, 263).
Recordings in alert, behaving monkeys have shown that
vestibular afferents respond equally well to passive and self-
generated head movements. However, the vestibular nuclei
included cells, the vestibular-only cells that were less sensi-
tive to active, compared with passive movements. It was
proposed by Cullen et al. (87) that during active head move-
ments an efference copy is processed by an internal model
which computes the sensory consequences of the motor
command. Neck proprioceptive inputs are compared with
this estimate of reafference in a matching center, probably
located in the cerebellum. If there is a match, a cancellation
signal is sent to the vestibular-only cells, thereby diminish-
ing their activation by reafferent inputs.
How this works for the proprioceptive system remains un-
certain, but there are a few clues. It was observed by Mc-
Closkey (243) and subsequently confirmed (13) that the
illusion of limb movement and displacement evoked by
muscle vibration is at its most vivid in a passive or slightly
contracted muscle. If the illusion is measured during graded
voluntary contractions, it becomes progressively weaker
until a point is reached, at 25% MVC, where it disappears
altogether (13). This result is unlikely to be due simply to
the changing mechanical conditions in the muscle from the
extrafusal contraction (37), since the vibration was applied
to the muscle belly, not the tendon. It suggests that if the
muscle contraction is sufficiently strong, the exafferent ac-
tivity, represented by the vibration response of spindles, is
no longer able to access consciousness.
Combining the vibration observations with the proposals of
Cullen et al. (87) and incorporating them into the known
central projection pathways for muscle afferents allows
speculation about a possible mechanism for distinguishing
between exafferent and reafferent muscle inputs during
limb movements (FIGURE 16). Animal studies indicate that
spindle afferents from the hindlimb of the cat have their first
synapse in the lumbar spinal cord, in Clarke’s column. The
second-order neurons comprise the DSCT which projects to
the anterior cerebellum with collaterals terminating in Nu-
cleus Z in the brain stem. Nucleus Z cells project to the
thalamus and cerebral cortex (249). The forelimb projec-
tion is not quite the same but conforms to a similar pattern
(see Ref. 248). While this level of detail is not available for
human subjects, the general pattern is likely to be similar.
FIGURE 16 shows these arrangements and some hypothe-
sized signal processing. If a muscle is stretched passively,
spindle exafferent signals are generated and project to the
cortex to generate a sensation of movement corresponding
to elongation of the muscle. If, on the other hand, spindle
activity is generated as a result of muscle contraction, in-
cluding, in part, through the fusimotor system, this repre-
sents a reafferent signal. A copy of the motor command, the
efference copy, is sent to the anterior cerebellum, a region
thought to be concerned with differentiation between
movements, depending on their sensory consequences (25).
Here the efference copy accesses an internal model (e.g.,
Ref. 424), which has been generated, based on past experi-
ence, to calculate the anticipated afferent signal for the
given motor command. The anticipated signal is compared
with the afferent signal coming from the muscle and pro-
jecting there via the DSCT. Depending on the precision of
the match, an inhibitory signal is generated which sup-
presses the reafferent component of the signal, allowing
passage to the cortex of only the exafferent component. The
inhibitory interaction could occur anywhere along the path-
way (FIGURE 16). There is evidence that for suppression of
cutaneous reafferent signals spinal presynaptic inhibitory
mechanisms are available (e.g., Ref. 349). Similar inhibitory
interactions, perhaps acting at brain stem, thalamus, or cor-
tical levels, would enable suppression of signals generated
by self-generated activity while leaving unaffected activity
generated by external sources like, for example, by muscle
length changes. Such a mechanism would resolve the prob-
lem of how the brain deals with signals from the same
receptor, but generated in two very different ways, only one
of which has access to consciousness.
One final point about the proposed arrangement is that it
must be able to accommodate the ability to perceive sensa-
tions of motor commands in the absence of peripheral feed-
back, as a result of an acute experimental deafferentation or
in chronically deafferented subjects. Therefore, an addi-
tional direct projection from motor cortex to sensory areas
has been postulated.
We are grateful to our many collaborators and colleagues
but particularly to Drs. Trevor Allen and Richard Fitzpat-
rick for comments on a draft manuscript. We thank Dr.
Sabine Giesbrecht and Andrea Riley for assistance in prep-
aration of the manuscript.
Address for reprint requests and other correspondence: U.
Proske, Dept. of Physiology, Monash University, Victoria
3800, Australia (e-mail: or S.
1688 Physiol Rev VOL 92 OCTOBER 2012
Gandevia, Neuroscience Research Australia, Randwick,
NSW 2031, Australia (e-mail:
U. Proske and S. Gandevia have each received research sup-
port over many years from the National Health and Medi-
cal Research Council of Australia.
No conflicts of interest, financial or otherwise, are declared
by the authors.
1. Adrian ED, Umrath K. The impulse discharge from the pacinian corpuscle. J Physiol 68:
139 –154, 1929.
2. Aimonetti JM, Hospod V, Roll JP, Ribot-Ciscar E. Cutaneous afferents provide a
neuronal population vector that encodes the orientation of human ankle movements.
J Physiol 580: 649 – 658, 2007.
3. al-Falahe NA, Nagaoka M, Vallbo AB. Response profiles of human muscle afferents
during active finger movements. Brain 113: 325–346, 1990.
4. Albert F, Bergenheim M, Ribot-Ciscar E, Roll JP. The Ia afferent feedback of a given
movement evokes the illusion of the same movement when returned to the subject
via muscle tendon vibration. Exp Brain Res 172: 163–174, 2006.
5. Allen TJ, Ansems GE, Proske U. Effects of muscle conditioning on position sense at the
human forearm during loading or fatigue of elbow flexors and the role of the sense of
effort. J Physiol 580: 423– 434, 2007.
6. Allen TJ, Ansems GE, Proske U. Evidence from proprioception of fusimotor coacti-
vation during voluntary contractions in humans. Exp Physiol 93: 391–398, 2008.
7. Allen TJ, Leung M, Proske U. The effect of fatigue from exercise on human limb
position sense. J Physiol 588: 1369 –1377, 2010.
8. Allen TJ, Proske U. Effect of muscle fatigue on the sense of limb position and move-
ment. Exp Brain Res 170: 30 –38, 2006.
9. Anema HA, van Zandvoort MJ, de Haan EH, Kappelle LJ, de Kort PL, Jansen BP,
Dijkerman HC. A double dissociation between somatosensory processing for per-
ception and action. Neuropsychologia 47: 1615–1620, 2009.
10. Angelaki DE, Cullen KE. Vestibular system: the many facets of a multimodal sense.
Annu Rev Neurosci 31: 125–150, 2008.
11. Aniss AM, Gandevia SC, Burke D. Reflex changes in muscle spindle discharge during a
voluntary contraction. J Neurophysiol 59: 908 –921, 1988.
12. Aniss AM, Gandevia SC, Milne RJ. Changes in perceived heaviness and motor com-
mands produced by cutaneous reflexes in man. J Physiol 397: 113–126, 1988.
13. Ansems GE, Allen TJ, Proske U. Position sense at the human forearm in the
horizontal plane during loading and vibration of elbow muscles. J Physiol 576:
445– 455, 2006.
14. Armel KC, Ramachandran VS. Projecting sensations to external objects: evidence
from skin conductance response. Proceedings 270: 1499 –1506, 2003.
15. Aschkenasy MT, Rothenhaus TC. Trauma and falls in the elderly. Emerg Med Clin
North Am 24: 413– 432, 2006.
16. Banks RW, Hulliger M, Saed HH, Stacey MJ. A comparative analysis of the encapsu-
lated end-organs of mammalian skeletal muscles and of their sensory nerve endings. J
Anat 214: 859 – 887, 2009.
17. Bastian H. The “muscular sense”; its nature and localization. Brain 10: 1–36, 1888.
18. Bastian HC. The Brain as an Organ of the Mind. New York: Appleton, 1880, p. 18.
19. Bays PM, Wolpert DM. Computational principles of sensorimotor control that mini-
mize uncertainty and variability. J Physiol 578: 387–396, 2007.
20. Bell C. On the nervous circle which connects the voluntary muscles with the brain..
Philos Trans R Soc 116: 163–173, 1826.
21. Bergenheim M, Ribot-Ciscar E, Roll JP. Proprioceptive population coding of two-
dimensional limb movements in humans. I. Muscle spindle feedback during spatially
oriented movements. Exp Brain Res 134: 301–310, 2000.
22. Berlucchi G, Aglioti SM. The body in the brain revisited. Exp Brain Res 200: 25–35,
23. Blakemore SJ, Frith CD, Wolpert DM. Spatio-temporal prediction modulates the
perception of self-produced stimuli. J Cogn Neurosci 11: 551–559, 1999.
24. Blakemore SJ, Wolpert D, Frith C. Why can’t you tickle yourself? Neuroreport 11:
R11–16, 2000.
25. Blakemore SJ, Wolpert DM, Frith CD. Central cancellation of self-produced tickle
sensation. Nat Neurosci 1: 635– 640, 1998.
26. Blanke O, Arzy S. The out-of-body experience: disturbed self-processing at the tem-
poro-parietal junction. Neuroscientist 11: 16 –24, 2005.
27. Bosco G, Poppele RE. Proprioception from a spinocerebellar perspective. Physiol Rev
81: 539 –568, 2001.
28. Bosco G, Poppele RE. Reference frames for spinal proprioception: kinematics based
or kinetics based? J Neurophysiol 83: 2946 –2955, 2000.
29. Bosco G, Poppele RE. Representation of multiple kinematic parameters of the cat
hindlimb in spinocerebellar activity. J Neurophysiol 78: 1421–1432, 1997.
30. Bosco G, Poppele RE, Eian J. Reference frames for spinal proprioception: limb end-
point based or joint-level based? J Neurophysiol 83: 2931–2945, 2000.
31. Botvinick M, Cohen J. Rubber hands “feel” touch that eyes see. Nature 391: 756, 1998.
32. Boyd I. The structure and innervations of the nuclear bag muscle fibre system and the
nuclear chain muscle fibre system in mammalian muscle spindles. Philos Trans R Soc
Lond B Biol Sci 245: 81–136, 1962.
33. Boyd IA, Roberts TD. Proprioceptive discharges from stretch-receptors in the knee-
joint of the cat. J Physiol 122: 38 –58, 1953.
34. Bravata DM, Smith-Spangler C, Sundaram V, Gienger AL, Lin N, Lewis R, Stave CD,
Olkin I, Sirard JR. Using pedometers to increase physical activity and improve health:
a systematic review. JAMA 298: 2296 –2304, 2007.
35. Brockett C, Warren N, Gregory JE, Morgan DL, Proske U. A comparison of the
effects of concentric versus eccentric exercise on force and position sense at the
human elbow joint. Brain Res 771: 251–258, 1997.
36. Brocklehurst JC, Robertson D, James-Groom P. Clinical correlates of sway in old
age–sensory modalities. Age Ageing 11: 1–10, 1982.
37. Brown MC, Enberg I, Matthews PB. The relative sensitivity to vibration of muscle
receptors of the cat. J Physiol 192: 773– 800, 1967.
38. Brown WF. A method for estimating the number of motor units in thenar muscles and
the changes in motor unit count with ageing. J Neurol Neurosurg Psychiatry 35: 845–
852, 1972.
39. Browne K, Lee J, Ring PA. The sensation of passive movement at the metatarso-
phalangeal joint of the great toe in man. J Physiol 126: 448 – 458, 1954.
40. Bryson W. A Short History of Everything. London: Black Swan, 2004.
41. Buonomano DV, Merzenich MM. Cortical plasticity: from synapses to maps. Annu Rev
Neurosci 21: 149 –186, 1998.
42. Burgess PR, Clark FJ. Characteristics of knee joint receptors in the cat. J Physiol 203:
317–335, 1969.
43. Burgess PR, Wei JY, Clark FJ, Simon J. Signaling of kinesthetic information by periph-
eral sensory receptors. Annu Rev Neurosci 5: 171–187, 1982.
44. Burke D, Aniss AM, Gandevia SC. In-parallel and in-series behavior of human muscle
spindle endings. J Neurophysiol 58: 417– 426, 1987.
1689Physiol Rev VOL 92 OCTOBER 2012
45. Burke D, Gandevia SC, Macefield G. Responses to passive movement of receptors in
joint, skin and muscle of the human hand. J Physiol 402: 347–361, 1988.
46. Burke D, Hagbarth KE, Lofstedt L, Wallin BG. The responses of human muscle spindle
endings to vibration of non-contracting muscles. J Physiol 261: 673– 693, 1976.
47. Burke D, Hagbarth KE, Skuse NF. Voluntary activation of spindle endings in human
muscles temporarily paralysed by nerve pressure. J Physiol 287: 329 –336, 1979.
48. Burke D, Hagbarth KE, Wallin BG. Reflex mechanisms in Parkinsonian rigidity. Scand
J Rehabil Med 9: 15–23, 1977.
49. Bushnell MC, Smith DV (Editors). The Senses: A Comprehensive Reference. San Diego,
CA: Academic, 2008.
50. Butler AA, Lord SR, Rogers MW, Fitzpatrick RC. Muscle weakness impairs the pro-
prioceptive control of human standing. Brain Res 1242: 244 –251, 2008.
51. Cafarelli E, Bigland-Ritchie B. Sensation of static force in muscles of different length.
Exp Neurol 65: 511–525, 1979.
52. Cafarelli E, Kostka CE. Effect of vibration on static force sensation in man. Exp Neurol
74: 331–340, 1981.
53. Cafarelli E, Layton-Wood J. Effect of vibration on force sensation in fatigued muscle.
Med Sci Sports Exercise 18: 516 –521, 1986.
54. Calford MB, Tweedale R. C-fibres provide a source of masking inhibition to primary
somatosensory cortex. Proc R Soc Lond B Biol Sci 243: 269 –275, 1991.
55. Calford MB, Tweedale R. Immediate and chronic changes in responses of somatosen-
sory cortex in adult flying-fox after digit amputation. Nature 332: 446 – 448, 1988.
56. Calford MB, Tweedale R. Immediate expansion of receptive fields of neurons in area
3b of macaque monkeys after digit denervation. Somatosens Mot Res 8: 249 –260,
57. Capaday C, Cooke JD. The effects of muscle vibration on the attainment of intended
final position during voluntary human arm movements. Exp Brain Res 42: 228 –230,
58. Cardinali L, Frassinetti F, Brozzoli C, Urquizar C, Roy AC, Farne A. Tool-use induces
morphological updating of the body schema. Curr Biol 19: R478 – 479, 2009.
59. Carpenter JE, Blasier RB, Pellizzon GG. The effects of muscle fatigue on shoulder joint
position sense. Am J Sports Med 26: 262–265, 1998.
60. Carruthers G. Reply to Tsakiris and Fotopoulou, “Is my body the sum of on-line and
off-line body representations?” Conscious Cogn 17: 1321–1323, 2008.
61. Carruthers G. Types of body representation and the sense of embodiment. Conscious
Cogn 17: 1302–1316, 2008.
62. Carson RG, Riek S, Shahbazpour N. Central and peripheral mediation of human force
sensation following eccentric or concentric contractions. J Physiol 539: 913–925,
63. Case LK, Wilson RC, Ramachandran VS. Diminished size-weight illusion in anorexia
nervosa: evidence for visuo-proprioceptive integration deficit. Exp Brain Res 217:
79 – 87, 2012.
64. Chambers MR, Andres KH, von Duering M, Iggo A. The structure and function of the
slowly adapting type II mechanoreceptor in hairy skin. Q J Exp Physiol Cogn Med Sci 57:
417– 445, 1972.
65. Chew JZ, Gandevia SC, Fitzpatrick RC. Postural control at the human wrist. J Physiol
586: 1265–1275, 2008.
66. Christakos CN, Windhorst U. Spindle gain increase during muscle unit fatigue. Brain
Res 365: 388 –392, 1986.
67. Ciancia F, Maitte M, Coquery JM. Reduction during movement of the evoked poten-
tials recorded along the extralemniscal pathways of the cat. Electroencephalogr Clin
Neurophysiol 48: 197–202, 1980.
68. Clark FJ, Burgess RC, Chapin JW, Lipscomb WT. Role of intramuscular receptors in
the awareness of limb position. J Neurophysiol 54: 1529 –1540, 1985.
69. Clark FJ, Grigg P, Chapin JW. The contribution of articular receptors to propriocep-
tion with the fingers in humans. J Neurophysiol 61: 186 –193, 1989.
70. Clark FJ, Horch KW, Bach SM, Larson GF. Contributions of cutaneous and joint
receptors to static knee-position sense in man. J Neurophysiol 42: 877– 888, 1979.
71. Cole J. Large-fiber sensor neuropathy. In: Encyclopedia of Neuroscience, edited by
Binder MD, Hirokawa N, Windhorst U. Berlin: Spinger, 2009, p. 2102–2107.
72. Cole J. The phenomenology of agency and intention in the face of paralysis and
insentience. Phenom Cogn Sci 6: 309 –325, 2007.
73. Cole J. Pride and a Daily Marathon. Boston: MIT Press, 1995.
74. Cole J, Gallagher S, McNeill D. Gesture following deafferentation: a phenomenolog-
ically informed experimental study. Phenom Cogn Sci 1: 49 – 67, 2002.
75. Cole J, Paillard J. Living without touch and peripheral information about body position
and movement: studies with deafferented participants In: The Body and the Self, edited
by Bermudez JL, Marcel A, Eilan N. Cambridge, MA: MIT Press, 1995, p. 24549 –266.
76. Cole JD, Sedgwick EM. The perceptions of force and of movement in a man without
large myelinated sensory afferents below the neck. J Physiol 449: 503–515, 1992.
77. Collins DF, Cameron T, Gillard DM, Prochazka A. Muscular sense is attenuated when
humans move. J Physiol 508: 635– 643, 1998.
78. Collins DF, Prochazka A. Movement illusions evoked by ensemble cutaneous input
from the dorsum of the human hand. J Physiol 496: 857– 871, 1996.
79. Collins DF, Refshauge KM, Gandevia SC. Sensory integration in the perception of
movements at the human metacarpophalangeal joint. J Physiol 529: 505–515, 2000.
80. Collins DF, Refshauge KM, Todd G, Gandevia SC. Cutaneous receptors contribute to
kinesthesia at the index finger, elbow, and knee. J Neurophysiol 94: 1699 –1706, 2005.
81. Cooke JD, Brown S, Forget R, Lamarre Y. Initial agonist burst duration changes with
movement amplitude in a deafferented patient. Exp Brain Res 60: 184 –187, 1985.
82. Cordo P, Gurfinkel VS, Bevan L, Kerr GK. Proprioceptive consequences of tendon
vibration during movement. J Neurophysiol 74: 1675–1688, 1995.
83. Corradi-Dell’Acqua C, Tomasino B, Fink GR. What is the position of an arm relative to
the body? Neural correlates of body schema and body structural description. J Neu-
rosci 29: 4162– 4171, 2009.
84. Crago PE, Houk JC, Rymer WZ. Sampling of total muscle force by tendon organs. J
Neurophysiol 47: 1069 –1083, 1982.
85. Craske B. Perception of impossible limb positions induced by tendon vibration. Sci-
ence 196: 71–73, 1977.
86. Cullen KE. Sensory signals during active versus passive movement. Curr Opin Neurobiol
14: 698 –706, 2004.
87. Cullen KE, Brooks JX, Jamali M, Carriot J, Massot C. Internal models of self-motion:
computations that suppress vestibular reafference in early vestibular processing. Exp
Brain Res 210: 377–388, 2011.
88. Day BL, Fitzpatrick RC. The vestibular system. Curr Biol 15: R583–586, 2005.
89. De Vignemont F. Body schema and body image–pros and cons. Neuropsychologia 48:
669 – 680, 2010.
90. Decety J, Lindgren M. Sensation of effort and duration of mentally executed actions.
Scand J Psychol 32: 97–104, 1991.
91. Desaki J, Nishida N. A further observation of muscle spindles in the extensor digito-
rum longus muscle of the aged rat. J Electron Microsc 59: 79 – 86, 2010.
92. Dijkerman HC, de Haan EH. Somatosensory processes subserving perception and
action. Behav Brain Sci 30: 189 –201, 2007.
93. Dimitriou M, Edin BB. Human muscle spindles act as forward sensory models. Curr
Biol 20: 1763–1767, 2010.
94. Dirks AJ, Leeuwenburgh C. The role of apoptosis in age-related skeletal muscle
atrophy. Sports Med 35: 473– 483, 2005.
95. Donaldson IM. The functions of the proprioceptors of the eye muscles. Philos Trans R
Soc Lond B Biol Sci 355: 1685–1754, 2000.
96. Downing PE, Jiang Y, Shuman M, Kanwisher N. A cortical area selective for visual
processing of the human body. Science 293: 2470 –2473, 2001.
1690 Physiol Rev VOL 92 OCTOBER 2012
97. Dummer T, Picot-Annand A, Neal T, Moore C. Movement and the rubber hand
illusion. Perception 38: 271–280, 2009.
98. Edin B. Cutaneous afferents provide information about knee joint movements in
humans. J Physiol 531: 289 –297, 2001.
99. Edin BB. Finger joint movement sensitivity of non-cutaneous mechanoreceptor affer-
ents in the human radial nerve. Exp Brain Res 82: 417– 422, 1990.
100. Edin BB. Quantitative analyses of dynamic strain sensitivity in human skin mechano-
receptors. J Neurophysiol 92: 3233–3243, 2004.
101. Edin BB. Quantitative analysis of static strain sensitivity in human mechanoreceptors
from hairy skin. J Neurophysiol 67: 1105–1113, 1992.
102. Edin BB, Abbs JH. Finger movement responses of cutaneous mechanoreceptors in the
dorsal skin of the human hand. J Neurophysiol 65: 657– 670, 1991.
103. Edin BB, Johansson N. Skin strain patterns provide kinaesthetic information to the
human central nervous system. J Physiol 487: 243–251, 1995.
104. Edin BB, Vallbo AB. Stretch sensitization of human muscle spindles. J Physiol 400:
101–111, 1988.
105. Ehrsson HH. The experimental induction of out-of-body experiences. Science 317:
1048, 2007.
106. Ehrsson HH. How many arms make a pair? Perceptual illusion of having an additional
limb. Perception 38: 310 –312, 2009.
107. Ehrsson HH, Holmes NP, Passingham RE. Touching a rubber hand: feeling of body
ownership is associated with activity in multisensory brain areas. J Neurosci 25:
10564 –10573, 2005.
108. Ehrsson HH, Kito T, Sadato N, Passingham RE, Naito E. Neural substrate of body size:
illusory feeling of shrinking of the waist. PLoS Biol 3: e412, 2005.
109. Ehrsson HH, Spence C, Passingham RE. That’s my hand! Activity in premotor cortex
reflects feeling of ownership of a limb. Science 305: 875– 877, 2004.
110. Ehrsson HH, Wiech K, Weiskopf N, Dolan RJ, Passingham RE. Threatening a rubber
hand that you feel is yours elicits a cortical anxiety response. Proc Natl Acad Sci USA
104: 9828 –9833, 2007.
111. Eklund G. Position sense and state of contraction: the effects of vibration. J Neurol
Neurosurg Psychiatry 35: 606 – 611, 1972.
112. Ellaway PH, Prochazka A, Chan M, Gauthier MJ. The sense of movement elicited by
transcranial magnetic stimulation in humans is due to sensory feedback. J Physiol 556:
651– 660, 2004.
113. Fallon JB, Macefield VG. Vibration sensitivity of human muscle spindles and Golgi
tendon organs. Muscle Nerve 36: 21–29, 2007.
114. Felician O, Anton JL, Nazarian B, Roth M, Roll JP, Romaiguere P. Where is your
shoulder? Neural correlates of localizing others’ body parts. Neuropsychologia 47:
1909 –1916, 2009.
115. Ferrell WR. The adequacy of stretch receptors in the cat knee joint for signaling joint
angle throughout a full range of movement. J Physiol 299: 85–99, 1980.
116. Ferrell WR, Gandevia SC, McCloskey DI. The role of joint receptors in human kin-
aesthesia when intramuscular receptors cannot contribute. J Physiol 386: 63–71, 1987.
117. Ferrell WR, Smith A. The effect of digital nerve block on position sense at the proximal
interphalangeal joint of the human index finger. Brain Res 425: 369 –371, 1987.
118. Ferrell WR, Smith A. Position sense at the proximal interphalangeal joint of the human
index finger. J Physiol 399: 49 – 61, 1988.
119. Filion M, Tremblay L, Bedard PJ. Abnormal influences of passive limb movement on
the activity of globus pallidus neurons in parkinsonian monkeys. Brain Res 444: 165–
176, 1988.
120. Fitzpatrick R, Burke D, Gandevia SC. Task-dependent reflex responses and move-
ment illusions evoked by galvanic vestibular stimulation in standing humans. J Physiol
478: 363–372, 1994.
121. Fitzpatrick R, McCloskey DI. Proprioceptive, visual and vestibular thresholds for the
perception of sway during standing in humans. J Physiol 478: 173–186, 1994.
122. Fitzpatrick R, Rogers DK, McCloskey DI. Stable human standing with lower-limb
muscle afferents providing the only sensory input. J Physiol 480: 395– 403, 1994.
123. Flor H, Elbert T, Knecht S, Wienbruch C, Pantev C, Birbaumer N, Larbig W, Taub E.
Phantom-limb pain as a perceptual correlate of cortical reorganization following arm
amputation. Nature 375: 482– 484, 1995.
124. Fortier S, Basset FA, Billaut F, Behm D, Teasdale N. Which type of repetitive muscle
contractions induces a greater acute impairment of position sense? J Electromyogr
Kinesiol 20: 298 –304, 2010.
125. Fotopoulou A, Rudd A, Holmes P, Kopelman M. Self-observation reinstates motor
awareness in anosognosia for hemiplegia. Neuropsychologia 47: 1256 –1260, 2009.
126. Freund HJ. Somatosensory and motor disturbances in patients with parietal lobe
lesions. Adv Neurol 93: 179 –193, 2003.
127. Frith C. The self in action: lessons from delusions of control. Consciousness Cognition
14: 752–770, 2005.
128. Frith CD. The positive and negative symptoms of schizophrenia reflect impairments in
the perception and initiation of action. Psychol Med 17: 631– 648, 1987.
129. Frith CD, Blakemore S, Wolpert DM. Explaining the symptoms of schizophrenia:
abnormalities in the awareness of action. Brain Res 31: 357–363, 2000.
130. Frith CD, Blakemore SJ, Wolpert DM. Abnormalities in the awareness and control of
action. Philos Trans R Soc Lond B Biol Sci 355: 1771–1788, 2000.
131. Fuentes CT, Bastian AJ. Where is your arm? Variations in proprioception across space
and tasks. J Neurophysiol 103: 164 –171, 2010.
132. Galea V. Changes in motor unit estimates with aging. J Clin Neurophysiol 13: 253–260,
133. Gallagher S. How the Body Shapes the Mind. New York: Oxford Univ. Press, 2005.
134. Gallagher S, Cole J. Body image and body schema in a deafferented subject. J Mind
Behav 16: 369 –390, 1995.
135. Gandevia SC. Illusory movements produced by electrical stimulation of low-threshold
muscle afferents from the hand. Brain 108: 965–981, 1985.
136. Gandevia SC. Kinesthesia: roles for afferent signals and motor commands. In: Hand-
book of Physiology. Exercise: Regulation and Integration of Multiple Systems. Bethesda,
MD: Am. Physiol. Soc., 1996, sect. 12, p. 128 –172.
137. Gandevia SC. Spinal and supraspinal factors in human muscle fatigue. Physiol Rev 81:
1725–1789, 2001.
138. Gandevia SC, Hall LA, McCloskey DI, Potter EK. Proprioceptive sensation at the
terminal joint of the middle finger. J Physiol 335: 507–517, 1983.
139. Gandevia SC, Killian K, McKenzie DK, Crawford M, Allen GM, Gorman RB, Hales JP.
Respiratory sensations, cardiovascular control, kinaesthesia and transcranial stimula-
tion during paralysis in humans. J Physiol 470: 85–107, 1993.
140. Gandevia SC, Mahutte CK. Theoretical requirements for the interpretation of signals
of intramuscular tension. J Theor Biol 97: 141–153, 1982.
141. Gandevia SC, McCloskey DI. Changes in motor commands, as shown by changes in
perceived heaviness, during partial curarization and peripheral anaesthesia in man. J
Physiol 272: 673– 689, 1977.
142. Gandevia SC, McCloskey DI. Interpretation of perceived motor commands by refer-
ence to afferent signals. J Physiol 283: 493– 499, 1978.
143. Gandevia SC, McCloskey DI. Joint sense, muscle sense, and their combination as
position sense, measured at the distal interphalangeal joint of the middle finger. J
Physiol 260: 387– 407, 1976.
144. Gandevia SC, McCloskey DI. Sensations of heaviness. Brain 100: 345–354, 1977.
145. Gandevia SC, McCloskey DI, Potter EK. Alterations in perceived heaviness during
digital anaesthesia. J Physiol 306: 365–375, 1980.
146. Gandevia SC, Miller S, Aniss AM, Burke D. Reflex influences on muscle spindle activity
in relaxed human leg muscles. J Neurophysiol 56: 159 –170, 1986.
147. Gandevia SC, Phegan CM. Perceptual distortions of the human body image produced
by local anaesthesia, pain and cutaneous stimulation. J Physiol 514: 609 – 616, 1999.
1691Physiol Rev VOL 92 OCTOBER 2012
148. Gandevia SC, Smith JL, Crawford M, Proske U, Taylor JL. Motor commands contrib-
ute to human position sense. J Physiol 571: 703–710, 2006.
149. Gandevia SC, Wilson L, Cordo PJ, Burke D. Fusimotor reflexes in relaxed forearm
muscles produced by cutaneous afferents from the human hand. J Physiol 479: 499 –
508, 1994.
150. Gauthier I, Tarr MJ, Moylan J, Skudlarski P, Gore JC, Anderson AW. The fusiform “face
area” is part of a network that processes faces at the individual level. J Cogn Neurosci
12: 495–504, 2000.
151. Georgopoulos AP, Kalaska JF, Caminiti R, Massey JT. On the relations between the
direction of two-dimensional arm movements and cell discharge in primate motor
cortex. J Neurosci 2: 1527–1537, 1982.
152. Georgopoulos AP, Schwartz AB, Kettner RE. Neuronal population coding of move-
ment direction. Science 233: 1416 –1419, 1986.
153. Gilhodes JC, Roll JP, Tardy-Gervet MF. Perceptual and motor effects