Professional Documents
Culture Documents
Feldman Levin PDF
Feldman Levin PDF
Feldman Levin PDF
CITATIONS
READS
34
124
1 author:
Anatol Feldman
Universit de Montral
169 PUBLICATIONS 6,780
CITATIONS
SEE PROFILE
Abstract This chapter is a brief account of fundamentals of the equilibriumpoint hypothesis or more adequately called the threshold control theory (TCT).
It also compares the TCT with other approaches to motor control. The basic
notions of the TCT are reviewed with a major focus on solutions to the
problems of multi-muscle and multi-degrees of freedom redundancy. The
TCT incorporates cognitive aspects by explaining how neurons recognize that
internal (neural) and external (environmental) events match each other. These
aspects as well as how motor learning occurs are subjects of further development of the TCT hypothesis.
Introduction
We begin with an explanation of two fundamentals of the equilibrium point
(EP) hypothesis the EP concept and the notion that changes in threshold
positions of the body or its segments underlie EP shifts and motor actions.
From the biophysical viewpoint, motor actions result from shifts in the
equilibrium state that the organism and the environment tend to achieve in
the process of the mutual interaction. Such shifts can be elicited voluntarily, by
the organism, or involuntary, following changes in environmental forces. It is
essential to emphasize that the equilibrium state is conditioned by both the
organism and the environment and therefore the organism can only influence
but not entirely predetermine it: the environment is an equal player in achieving
this state.
The notion of EP was introduced to characterize the equilibrium state in
terms of output, mechanical variables such as the equilibrium positions of body
A.G. Feldman (*)
Department of Physiology, University of Montreal, School of Physical
and Occupational Therapy, McGill University; Center for Interdisciplinary
Research in Rehabilitation, Montreal Rehabilitation Institute and Jewish
Rehabilitation Hospital, Canada
e-mail: feldman@med.umontreal.ca
699
700
701
only has survived numerous attempts of rejection during last 43 years but also
has been continuously advanced by solving several essential problems in motor
control (see below) suggests that its explanatory and predictive power is far
from having been exhausted.
In many aspects, the TCT challenges traditional views on how motor actions
are controlled and produced, which explains why the theory, in essence, simple
is often poorly understood and misinterpreted (Feldman and Latash 2005). We
will list five aspects of the TCT that conflict with traditional approaches to
motor control.
First, the TCT implies that neural control levels are not involved in the
specification of movement trajectories, muscle activation, forces and torques,
nor do they specify the EP the values of these variables emerge following the
central shifts in the threshold positions and interactions within the organism
and between the organism and the environment. Thus, the TCT rejects the idea
motivated by mechanics and robotics, that the nervous system is directly
involved in programming of EMG activity and mechanical variables describing
the motor output.
Second, the TCT implies that by specifying threshold positions of body
segments the neural control levels only narrow the set of possible EPs and the
repertoire of possible motor actions but do not specify per se a unique EP or
motor action. For example, in response to the same shifts in the threshold value
of a joint angle, the motor output may be different, depending on external
conditions: if there are no obstacles, the joint will move to another position or,
otherwise, generate an isometric torque. In other words, neural control levels
only diminish the redundancy in the possible EPs and motor actions, while a
unique sequence of EPs or a unique motor action emerge in the process of the
interaction of different neuromuscular components of the organism between
themselves and with the environment. This notion conflicts with the traditional
idea that solutions to redundancy problems in motor control rely on internal
computational procedures that compares different motor actions, selects and
specifies one of them based on some optimality criterion.
Third, in the cases when external forces acting on body segments are zero, the
threshold position (R) may coincide with the actual position (Q) of these
segments but generally these positions are different. This difference implies an
important notion. Each actual position, Q, can be considered as a point in the
space of all physically possible positions (postures) of the body. In other words,
body postures can be considered as coordinates in the space of all-possible body
configurations. By specifying an R posture, the nervous system defines the
origin point in this space (frame of reference). Because of the threshold nature
of the R posture, other body postures becomes non-equivalent: anatomically
and neurophysiologically, some muscles (usually called agonists) at a given
posture, Q, appear as being stretched above the thresholds of activation and
resist the deviation from the threshold, R posture whereas other muscles
(antagonists) remain silent. Physically, this imply that muscles react depending
on some measure (metrics) characterizing the closeness of a given posture to
702
703
reference for action and perception are united is reminiscent of the suggestion
that action and perception are related (Gibson 1966).
704
transform the computed torques into individual muscle forces, transform the
muscle forces into individual forces of motor units, transform forces into EMG
signals, transform EMG signals into postsynaptic potentials of hundreds of
individual motoneurons, decompose these potentials into millions of individual
synaptic potentials descending to motoneurons, and so on for interneurons
terminating on motoneurons. In other words, the system would need to solve an
exponentially increasing number of redundancy problems arising at each level
of these inverse transformations. Because of the strong non-linearity of neural
elements (electrical thresholds), some of the inverse transformations appear
impossible (Ostry and Feldman 2003). Thus, the programmed specification of
torques and EMG patterns with or without internal models appears physiologically unrealistic.
One major implication of the TCT is that muscle forces are emergent,
non-programmed variables (see above). The hypothesis acknowledges that,
physiologically, control processes can be expressed in terms of changes in the
membrane potentials of motoneurons and that in the presence of positional
feedback arising from muscle receptors, control processes result in shifts of the
threshold muscle length (l) at which muscles begin to be activated (Feldman
1986). Shifts in the threshold length are ambiguously related to the EMG
signals: the muscle is activated or not depending on the difference (and the
rate of its change) between the actual and the threshold length. The existence of
muscle activation thresholds and the possibility of their central regulation result
from a specific form of integration of afferent and central influences on motoneurons (see below). In other words, the TCT implies that nervous system
controls motor actions by changing neuro-mechanical parameters to which
neither mechanical variables, nor motor commands (EMG signals or their
efference copies) belong.
There are attempts to justify the internal model approach by claiming that it
offers explanations of predictive and anticipatory powers of the brain, most
often manifested during motor learning. While considering these aspects, it is
worth mentioning that Michael Turvey (personal communication) attracted
our attention to the existence of a modern classification of predictive systems
(Dubois 2001). Strong predictive systems are those in which predictive properties are inherent in the systems natural dynamics and thus do not rely on
internal models. Weak predictive systems are based on internal models of
themselves. Typically, proponents of internal models underestimate the possibilities of natural dynamic systems and consider observations of predictive and
anticipatory motor behavior as unequivocal evidence of such models (for
review see Ostry and Feldman 2003). They thus disregard the option that
biological systems may manifest strong predictive and anticipatory properties
in the absence of any internal models. In contrast, this option is offered by the
EP hypothesis, as illustrated in several studies from Ostrys and Gribbles
groups (for references see Ostry and Feldman 2003) as well as by empirical
studies of adaptation to changing external forces (Weeks et al. 1996; Foisy and
Feldman 2006). The claim that predictive, forward models are necessary to
705
overcome the destabilizing effects of feedback delays has also been rejected by a
recent demonstration that threshold control manifests predictive properties
allowing the system to achieve stability despite substantial feedback delays
(Pilon and Feldman 2006). Pilon et al. (2007) have also demonstrated that the
experimental observations of an anticipatory increase in the grip force during
arm motion and correlation between the grip and load forces can also be
produced in the absence of any internal models.
Thus, the TCT does not conflict with dynamic systems theory, but rather, is a
part of it. On the other hand, the TCT rejects, as physiologically unrealistic, the
notion that motor control is reduced to computation and programming of
motor commands and mechanical variables characterizing the motor output.
It also suggests that natural dynamic systems, to which biological organisms
belong, may manifest predictive and anticipatory properties in the absence of
any computational inverse or forward internal models of the organism interacting with the environment.
Past
Original Findings in Humans
The TCT resulted from an experimental comparison of involuntary arm movements elicited by unloading with voluntary arm movements produced intentionally by subjects.
Figure 1 shows the manipulandum that was used in experiments in which
changes in forearm position, torque, and EMG activity of biceps (BB) and
triceps brachii (TB) muscles resulting from unloading of the forearm were
recorded (Asatryan and Feldman 1965). Loads (weights) were hung on ropes
that were attached via pulleys to the manipulandum by small electromagnetic
locks. Subjects initially established a specified position while counteracting a
certain load torque. The combination of the current torque and position was
visible as a point on a large display of an oscilloscope and the subject was
required to bring it to the target point (an initial EP) on the same display. Once
the EP was established, the subject was required to close the eyes and an
electrical pulse was delivered to one of the two electromagnetic locks. The
load thus was suddenly decreased, resulting in motion of the forearm to another
combination of the static torque and position (a new EP). Then the initial load
was restored, the subject established the same initial EP, and the trial was
repeated with the same or a different randomly chosen final load. It was thus
possible to record a set of EPs resulting from unloading from the same initial
EP. The tonic EMG activity of pre-loaded muscles was not the same for
different EPs it monotonically decreased with the decreasing amount of the
residual load (Figs. 2 and 3). Together with the initial EP, the the final EPs
comprised a monotonic but non-linear torque-angle curve the displacement of
706
a
b
Fig. 1 Manipulandum that was used in unloading experiments that revealed that intentional
movements of the arm involve resetting of the threshold elbow angle. Horizontal forces
(weights) were applied to a lever (a) of the manipulandum via ropes, pulleys and small
electromagnetic locks (b). The elbow angle was measured with a potentiometer (c) and the
elbow torque was measured with a ring strain-gauge (d)
a
BB
TB
Angle
Torque
1s
d
40 0
10 Nm
Fig. 2 Unloading of pre-loaded elbow extensors. EMG activity of extensors (triceps brachi,
TB) and flexors (biceps brachii, BB) are shown together with the elbow angle and torque. In
repeated trials, subjects specified the same joint angle while compensating the same net load
torque. The amount of unloading in different trials (af) was randomly selected by choosing
different initial combinations of weights hung on the electromagnetic locks (see Fig. 1) and by
unlocking one or two of them. Note that the tonic EMG activity of TB progressively decreases
with the decreasing amount of the residual load. Similar unloading experiments were repeated
after subjects made intentional transition to different (up to seven) initial combinations of
elbow position and torque
707
c
BB
TB
Torque
5 Nm 500
Angle
f
Torque (Nm)
6
4
2
0
60
1s
80 100 1200
Elbow angle
the arm increased with the increasing amount of unloading (Fig. 3 h). The
subjects were instructed not to intervene voluntarily, i.e., behave in a natural
way without trying to intentionally modify the effects of unloading. The behavior of the arm was similar to that of a mass-spring system in response to
changes in load and, therefore, this arm behavior was called spring-like.
As mentioned above, the purpose of these experiments was to try to understand what variable(s) was changed by the nervous system when an intentional
movement was made to another position, rather than describe how the arm
reacted to unloading. With this purpose, after recording one torque-angle
curve, the subject was asked to establish a new initial EP, thus producing a
voluntary action. Then, the unloading trials were repeated with the new initial
EP, yielding a similar torque-angle curve. We thus recorded a family of torqueangle curves starting from different initial EPs (Fig. 4, filled dots) in each of 7
subjects. This family was complemented by an additional torque-angle characteristic recorded during slow passive rotation of the manipulandum when the
subject was instructed to fully relax elbow muscles, using EMG feedback
(passive-torque-angle characteristic; Fig. 4, dashed curve).
The spring-like responses of the arm to unloading suggested that there was
some parameter that subjects were reluctant to change (kept invariant) when
unloading was done but they were forced to change this parameter when
instructed to intentionally move the arm to another initial EP and thus establish
another torque-angle relationship. The suggestion of the existence of a parameter specifically related to intentional actions followed from the mono-parametric
708
Torque (Nm)
60
0
100
160
Elbow angle
709
710
Torque (kg)
0.6
0.5
500
320
380
320
150
0.4
0.3
20
S2
0.2
0.1
100
g
S1
500
580
90
400
110 130
Elbow angle ()
b
(10; 10)
50
(5; 15)
180
(15; 5)
440
(5; 15)
(10; 10)
120
(5; 15)
(10; 10)
140
(15; 5)
c
140
100
440
240
50
7.5
130
0 90
Nm
d
1s
Fig. 5 Invariant properties of the torque-angle curves obtained by the unloading method. a:
Arm unloading was produced from the same initial torque-angle combination (initial EP,
upper dot). Different load characteristics (dashed lines S1 and S2) were used to produce final
torque-angle combinations (final EPs, the point of intersection of lines S1 and S2 with the solid
curve). Despite the essential change in the load characteristics, the different final EPs comprise
a single torque-angle characteristic (solid curve). bd: The same final EP can be obtained
despite the temporal variations of the pattern of unloading. The numbers in brackets in each
panel in b and c show the weights (in N) removed in each of two sequential unloadings; the
bottom number shows the interval (in ms) between the two unloadings. In d, full unloadings
were produced by removing the two weights simultaneously. eg: the unloading effect can be
reversed (e, f) or not (g) depending on whether or not the loading after unloading was
produced in a smooth or an abrupt way, respectively
changing the activity of g-motoneurons that influence muscle spindles and, via
afferent feedback, -motoneurons (Fig. 6).
Matthews also found that the threshold could be changed by continuous
(tonic) stimulation of cutaneous nerves (skin reflexes) or afferents responsible
for the crossed extensor reflex. These findings showed that the stretch reflex
could not be considered as producing stereotyped responses to perturbations
but represented a well-organized, parametrically tunable and thus flexible
structure. At the time when the experiments were done, the dominating view
was that EMG patterns were basically generated independently of reflexes and
that the latter only modulated the centrally specified EMG output (the servoassistant hypothesis). This may explain why Matthews was unable to suggest
that threshold control was a fundamental characteristic of any central influences on -motoneurons, not only mediated by g-motoneurons but also those
targeting -motoneurons pre- or post-synaptically or via spinal interneurons.
711
1400
a
b
1200
1000
800
c
600
400
200
0
0
6
8
Extension (mm)
10
12
14
a
1 kg
1 sec
b
0.5 mV
Fig. 6 Experiments by Methews (1959) showing that g-motoneurons influence the threshold
of the stretch reflex. In the upper panel, an increase in the reflex threshold of the soleusgastrocnemius muscles in the decerebrated cat is illustrated by shifts of the tension-extension
curve from initial position a to the right (be) when the g-fibers in the muscle nerve are
progressively blocked by an anesthetic. In the lower panel, the same effect is illustrated by
showing that a greater lengthening of the muscles is required to evoke EMG activity in the
muscles when g-fibers are blocked (lower traces, compare with upper traces obtained without
anesthesia)
Feldman and Orlovsky (1972) not only reproduced Matthews results but
also showed that many descending systems (including the vestibulo-, reticuloand cortico- and rubro-spinal systems) have the capacity to modify muscle
activation thresholds (Fig. 7; see also Nichols and Steeves 1986; Capaday
1995). The experiments demonstrating threshold position control in humans
and decerebrated cats provided a solid empirical foundation for the EP
hypothesis.
712
DN
b
1.2
Force (kg)
4.5
3
2
0
12.5
2.5
15
15 + .3
12
10
15 + 2
7
0
10 12.5 + 2
0 2
0
0
20
15
15
d
20 mm
Length (mm)
0
5
DN DN+TA
Length
Force
EMG
1 kg
1s
Fig. 7 The vestibulo- and cortico-spinal pathways as well as heteronymous reflexes may
influence the threshold position at which muscles are activated (Feldman and Orlovsky
1972). a: Continuous (tonic) stimulation (30/s) of the vestibular Deiters nuclei (DN) decreases
the activation threshold length of the plantaris muscle as shown by shifts of the tensionextension curve to the left (numbers near the curves show the strength, in Volts), of bipolar
stimulation of the DN; the trace without a number is the tension-extension curve for the
inactive muscle. b: Tonic stimulation of the pyramidal tract decreases the effect of stimulation
of the DN (curve DNPYR) by increasing the activation threshold (shifting the curve to the
right). c: Stretching the antagonist muscle (the tibialis anterior, TA) by hanging a 3-kg load to
its tendon detached from the bone when the DN is stimulated results in an increase in the
activation threshold of the plantaris muscle (curves DNTA). d: Electromyographic demonstration of a decrease in the activation threshold of the plantaris elicited by tonic stimulation
of the DN (upper curves: no DN stimulation; lower curves: with DN stimulation)
713
b
MN
Afferent
input
a
Output
Membrane potential
V+
Central
Input
Vi
xi
xi
c
V+
Vi
Fig. 8 Schematic diagrams that explain the physiological origin of the control of threshold
position resulting from changes in the membrane potential of motoneourons. a: The threshold position at which a neuron is recruited becomes controllable only if central control
inputs are combined with position-dependent afferent feedback at the level of the neuronal
membrane. The existence of electrical threshold at which the neuron is recruited is an
additional condition of the existence of threshold control in a spatial domain. b: In the
absence of a control input, stretching the muscle results in an increase in the membrane
potential and eventually recruitment of motoneurons at a certain, threshold muscle length,
l. Control inputs to the motoneurons (vertical arrows) can increase the membrane potential, resulting in a decrease in the threshold muscle length l. c: Threshold muscle length may
also be changed by inpiuts from the brain stem by influencing the electrical threshold (V) of
motoneurons
influence on the motoneuron is the shift in the threshold muscle length. Note
that threshold muscle length can be regulated centrally, even if the electrical threshold of the motoneuron remains constant. It has recently been
shown that electrical threshold of motoneurons can also be modified by
some inputs from the brain stem (Fedirchuk and Dai (2004) which might
be an additional source of shifting in the threshold length, l (Fig. 8c).
Note that the electrical control signals to motoneurons could not be
decoded into the spatial dimensional threshold in the absence of proprioceptive feedback. Therefore, threshold position control does not exist in
proprioceptively deafferented subjects. As a result, deafferentation results
in substantial motor deficits (see e.g. Levin et al. 1995).
Control levels must be able to elicit activation or, conversely, relaxation of
the muscle at any length within the biomechanical range [x, x]. To meet these
requirements, the threshold must be able to be regulated in a range [l, l] that
exceeds the biomechanical range. In some subjects with hemiparesis and cerebral palsy, the range of threshold regulation is reduced, resulting in weakness,
spasticity, and deficits in inter-joint coordination (Levin et al. 2000; Jobin and
Levin 2000).
714
Fig. 9 Basic rules describing
threshold control, EMG and
force regulation for a single
muscle
Figure 9 lists basic rules underlying the threshold position control, EMG
activity and muscle force regulation for a single muscle. In particular, physiological data indicate that the threshold length is comprised of several additive
components with only one component controlled centrally (Matthews 1959;
Feldman and Orlovsky 1972; Feldman 2007). To reflect these findings, we use
the symbol l* for the composite (net) threshold whereas symbol l is reserved
for its central component:
l l v "t
(1)
(2)
715
716
changes in the muscle length signaled by the central and afferent components of
position sense cancel each other so that the joint angle is perceived as unchanged.
Typical kinesthetic illusions are explained by influences of tendon vibration
on the afferent component of position sense whereas inverted kinesthetic illusions are explained by the influences of vibration on the central component
(Feldman and Latash 1982).
One can assume that -motoneurons and g-motoneurons are facilitated in
parallel and this facilitation sets a threshold length (l) for each muscle or a
threshold joint angle (R) for all muscles of the joint (see below). In this case,
the same threshold angle provides a referent position for motor actions and for
interpretation of proprioceptive signals in position sense, thus giving an illustration of the notion that action and perception are provided in united spatial frames
of reference (Fig. 10). Note that the threshold position, R, should be understood
as the origin of the united frame of reference for motor action and position sense,
rather than as a copy of motor commands (efference copy) considered in
traditional approaches to action and perception. Also note that there is a spatial
equivalent of proprioceptive signals at the level of perception these signals can
be graded in proportion to the deviation of the joint from the R position, an
aspect of position sense that is essential for the explanation of how some anticipatory actions are produced.
Position sense
Joint torque
Load
central
component
R
afferent
component
Fig. 10 United frames of reference for motor actions and position sense exemplified for a
single joint. By setting a certain threshold joint angle (R), the system constraints the set of
possible equilibrium points the points of a static torque-angle characteristic. A specific
equilibrium point from this redundant set (e.g., point a) emerges following the interaction of
muscle with the load. In order to identify and thus adequate perceive the actual joint angle
(e.g., for point a), the system needs to take into account the threshold, R position (central
component of position sense) and the deviation from it signaled by proprioceptive signals
recalibrated in terms of position (afferent component of position sense). Motor actions are
produced by shifting the R position thus simultaneously resetting the central component of
position sense relative to which proprioceptive signals are interpreted. Sense of effort can be
generated based on a recalibration of the same afferent signals (including those delivered by
afferents form muscle spindles and tendon organs) in terms of force or torques
717
Joint torque
arm torque
Fig. 11 Preventing changes in the arm position when the load (e.g. a book held on the palm of
the hand) is removed by the subject himself. Subjects can specify an R position so that,
deviated from the R position by gravity, the arm arrives at an actual position (Q). When
somebody places a book on his palm, the arm is deviated to position Q0 . The change in the arm
position elicited by the load is stored in motor memory. While removing the book with the
other arm, the subject can produced a similar but opposite changes in the R position (arrow)
so that the arm remains at approximately the same position (Q0 ) as before the unloading
Anticipatory Actions
In restaurants, we often see waiters lifting a heavy beer stein from the tray while
steadily holding the latter. Indeed, they anticipate a change in the load and do
something to prevent motion of the arm when the load is lifted. Similarly, if a heavy
book held on the palm of the hand is suddenly removed by the subject himself, the
arm motion is usually prevented, in contrast to the case when the book is suddenly
removed by another person. To qualitatively explain this behavior, we need to take
into account that subjects initially specify an R position and the arm arrives at an
equilibrium position Q. This position changes to Q0 when the book is added. The
changes in the proprioceptive signals expressed in terms of changes in the position
resulting from adding the book can be remembered. While lifting the book, the
subject can simultaneously change the R position by the distance defined by these
remembered afferent signals. In other words, by influencing the R, the subject may
compensate the loss of the afferent contribution into position sense elicited by
unloading and thus keep the arm at approximately the same position (Fig. 11).
Present
Different Forms of Threshold Position Control
The TCT underwent some recent developments by applying the scheme in Fig. 8
to different neurons. For example, suppose there are neurons that receive
afferent facilitation in proportion to the distance (aperture) between the
718
thumb and the index finger, for example, during the task of gripping an object
between them. Then, similarly to the scheme described for motoneurons, independent, central inputs can be interpreted as setting the threshold position
(threshold or referent aperture) for activation of the muscles involved in the
grip (Fig. 12a). Thereby, the grip force will depend on the distance between the
actual aperture (Qa) defined by the size of the object) and the threshold aperture
(Ra) specified by the control inputs. The feasibility of these notions has been
demonstrated by simulation of the elevation of the grip force during arm
motion (Pilon et al. 2007). Similarly, suppose there are neurons that receive
afferent facilitation in proportion to the elbow angle. Then independent, central
inputs to these neurons can be interpreted as setting the threshold elbow angle
for activation of all muscles spanning the joint. In other words, the notion of
threshold position implies the existence of the joint angle at which all muscles
from this group, regardless of their biomechanical functions, reach their activation thresholds (Fig. 12b), as experimentally was confirmed (Ostry and Feldman
2003; Foisy and Feldman 2006).
Membrane potential
Aperture neuron
Aperture (Qa)
b
Elbow neuron
Fig. 12 Different forms of threshold position control a: If some neurons receive afferent
signals that monotonically related to the distance (aperture) between the index finger and the
thumb then central inputs can control the referent (threshold) value (Ra) of the aperture at
which the neurons (aperture neurons) and appropriate hand muscles begin to be recruited.
b: Similarly, if neurons receive afferent inputs that monotonically depend, say, on the elbow
joint angle, then the effect of control inputs is measurable in terms of a change in the threshold
angle at which these neurons (elbow neurons) begin to be recruited. Note that the diagrams
in a, b (see also Fig. 8) not only implies the existence of different neurons that provide different
forms of threshold position control but also explain the important cognitive ability of neurons
to identify when the central and afferent events delivered to the inputs of these neurons (the
actual and threshold positions) match each other. Thereby the neurons act depending on the
degree of the discrepancy between these events
Level
719
These examples shows that the notion of threshold position with appropriate modifications can be referred to different neurons and different levels of
the neuromuscular system, as exemplified in Table 1. At the level of the arm,
threshold position can be defined for its several segments (threshold aperture,
threshold configuration of the hand) or for the whole arm (threshold configuration of the arm; Archambault et al. 2005; Foisy and Feldman 2006). Threshold
position can also be defined for all skeletal muscles of the body (threshold or
referent body configuration), the notion that underlies the explanation of how all
muscles of the body are controlled without redundancy problems (Lestienne et al.
2000; St-Onge and Feldman 2004).
The list of possible threshold positions can be continued by having in mind
that motor actions are often directed to objects in an external space (e.g., a tea
cup on the table). In these cases, the position of effectors (e.g. the hand) or the
whole body should be associated with the environment, rather than with body
parts. The concept of effector neurons that control levels may use to shift the
threshold localization of the effectors or the whole body in the environment is
defined at this level (Table 1).
720
In fact, redundancy problems exist even at the level of single joint control and
here too, these problems are solved automatically without any explicit efforts
of the nervous system. Specifically, at the level of a single joint, the control
process is practically finished with the specification of the threshold angle for
activation of agonist and antagonist muscles. Neither levels of activation, nor
muscle forces, torques, or even a final EP are pre-determined by these control
signals. The IC of the joint defined by these signals is a collection of possible EPs
without indication of how a single EP is selected from this redundant set (see
above). How is the redundancy problem actually solved? Once the thresholds
are specified, the control levels allow different muscles (including flexors and
extensors) to interact between themselves, at mechanical and neural levels, and
with external forces. This interaction itself reduces the redundant number of
EPs to a single EP (Fig. 13).
The notion that neural control levels per se do not solve redundancy problems is crucial to the understanding of how multiple muscles of the body are
controlled. Specifically, it has been suggested that muscles of the whole body are
controlled as a coherent unit by a global factor the difference between the
actual, emerging configuration of the body and its threshold (referent) configuration modulated by the nervous system (Table 1). This allows the nervous
system to guide multiple muscles without redundancy problem: each skeletal
muscle of the body is only activated if its current muscle length at the current
body configuration exceeds the threshold muscle length defined by the referent
body configuration. This global factor in combination with local anatomical,
biomechanical and reflex factors determines whether a muscle is recruited or
Load:
isometric
b
spring-like
Joint torque
c
zero
a
R
Fig. 13 A redundancy problem and its solution at the level of single-joint control. By shifting
the threshold joint angle, the nervous system only predetermines a specific torque-angle
characteristic that represents a set of possible equilibrium points (EPs). A specific EP (a, b,
or c), i.e. the final combination of the position or torque is established in the process of the
interaction of the joint segments with the external forces (loads). Thus, control levels of the
nervous system only diminish the set of possible motor actions whereas the redundancy
problem yielding a unique motor action from this set emerges following the interaction of
the neuromuscular components between themselves and with the environmental forces
721
not (see Levin and Dimov 1997 and Feldman et al. 2007 on how coactivation of
opposing muscle groups influences this process). Experimental confirmations
and modeling of several movements involving substantial number of muscles,
including human walking and sit-to-stand movements, based on these notions,
have been provided (Gunther and Ruder 2003; St-Onge and Feldman 2004;
Archambault et al. 2005; Pilon et al. 2007; Feldman et al. 2007).
The notion that interactions within the neuromuscular system and between
this system and the environment play a decisive role in solving the redundancy
problem equally refers to the redundancy problem at the level of mechanical
degrees of freedom. To describe this solution, we need to characterize a general
principle guiding these interactions. Threshold control implies that given an
appropriate task-specific form of threshold position (R), the neuromuscular
elements are activated and tend to diminish the activity and internal interactions as well as the interaction between the system and the environment in order
to diminish the difference of the actual position of the body from the referent
position (R) to reach, if possible, an equilibrium state. In this state, the difference between Q and R values becomes minimal in the limits determined by
internal and external constraints. This process is a manifestation of the principle
of minimal action in the functioning of the neuromuscular system (cf. Gelfand
and Tsetlin 1971). It guides the motor action in the following way (Fig. 14).
Neural control levels select an (leading) form of threshold position control,
Rl, which is most appropriate for a given motor task. By changing this threshold, they change the activity (Al ) of appropriate neurons at this level. This
activity influences the form of threshold control, Rs , in neurons of subordinate
levels. The activity, As , of these neurons eventually influences the l thresholds
of motoneurons of muscles. A motor action emerges following the control
influences and the tendency to minimize the activity at other levels. The process
goes until external and internal constraints (e.g. intentional or mechanical
restrictions of the mobility of some degrees of freedom) limit the depth of the
minimization. In the absence of constraints, the system may reach an absolute
minimum at which the referent (threshold positions) are reached at all neuromuscular levels, as indicated by the bottom line in Fig. 14.
Figure 15 illustrates how, in qualitative terms, threshold control and the
principle of minimal action guide the motion of the hand to a cup. The movement is produced by shifting the referent coordinates (R) of the hand by
influencing neurons that receive afferent inputs related to the actual coordinates
(Q) of the hand in space (see the bottom section of Table 1). The movement will
proceed until the difference between these variables becomes minimal, which
occurs when neurons of subordinate levels, including motoneurons, also minimize their activity (Fig. 15). In each trial, there will be no uncertainty in
choosing one coordination pattern of a set of many other patterns each
time, the minimizing process will produce a unique coordination pattern. If
necessary, with additional corrective shifts in R, the target will be reached. The
coordination pattern leading to the same motor goal (motor equivalency) can,
indeed, naturally vary with task repetitions, history-dependent changes in the
722
Al = [Ql R*l]+
min
As = [Qs R*s]+
Am = [x *]+
Absolute minimum:
Al = As = Am = 0
Ql = Rl ; Qs = Rs ; x = for all muscles
Fig. 14 Principle of minimal action and its capacity to guide multiple muscles and degrees of
freedom without redundancy problems. Different forms of threshold position control are
tools that the nervous system can select by influencing respective neurons in order to reach the
motor goal. Like usual tools in everyday life, these forms do not represent the motor goals.
This strategy is reminiscent of using the steering wheel to direct the car motion: the focus is on
the direction of car motion, rather than on the means (turning the steering wheel) used to
accomplish this. Once the choice of the appropriate (leading) form (Rl) is made, the nervous
system influences respective neurons to guide the body or its segments to reach the goal. By
generating the activity (Al) depending on the difference between the actual and the referent
position at this level, these neurons influence the referent position (Rs) at subordinate neural
levels. The activity (As) neurons at these levels eventually influence the individual thresholds
of muscle and motoneurons. The functioning of the whole system is guided by the tendency to
reach activity minima at all levels, in the limits defined by internal and external constraints.
Because of the minimization principle, the system produces a unique action without concerns
with redundancy problems. Because of variation in the properties of neuromuscular elements
(e.g., due to fatigue, intentional or mechanical restrictions of the mobility of some degrees of
freedom), the emerging action can vary with movement repetition. With appropriate choice of
the leading form of threshold position control, the goal (e.g. reaching for a cup, see Fig. 15)
can be achieved comparatively easy either by a single shift in the Rl or after additional
corrections. The system may use several forms of threshold position control to reach several
motor goals (for example, to reach a cup on the table while simultaneously leaning the trunk
to better hear a person on the opposite side of the table)
723
R*s
Qs
Rl*
Ql
Fig. 15 Reaching for a cup guided by shifts in the referent hand position and by the principle
of minimal action. Some spinal and supraspinal neurons projecting to motoneurons mono- or
poly-synaptically may integrate somatosensory, visual and vestibular inputs and proprioceptive signals from muscle, joints and skin receptors to receive information about coordinates
(Q) of the hand in an external frame of reference (FR). Like for motoneurons, independent,
control influences on these neurons can be measured by the amount of shifts in the threshold
(referent) position (R) of the hand (hand neurons). Such neurons may or may not be
recruited depending on the difference between Q and R. In the task of reaching for a cup
control levels shift the R to move the hand in the desired direction. As long as the hand
approaches the cup, it is not essential whether or not the actual and referent trajectories of the
hand coincide. Reaching for a cap is thus controlled by shifting the origin R of the leading FR
in which the hand position (Q) is localized. According to the principle of minimal action (see
Fig. 14), the activity of hand neurons influences the referent body configuration set by changes
in the membrane potentials of subordinate neurons. The output of the latter, in turn,
influences the threshold lengths of individual muscles and eventually of motoneurons. The
activity of neural elements in each FR tends to minimize (min) the discrepancy between the
actual and referent coordinates, forcing the arm and other body segments to move until the
hand reaches a final position at which a minimum in the system, in the limits of intrinsic and
external constraints, is reached. Because of the weights of the body segments, the absolute
minimum (zero activity at all subordinated levels will not be reached: the body will continue to
move until at some equilibrium position, Q, the activity of appropriate muscles will produce
torques balancing the weight torques. The system may correct the Rl until the resulting hand
position becomes convenient for grasping the cup
these principles lack the essential element describing forms of threshold positions underlying motor control. Therefore, their applicability and explanatory
power are restricted, as would be the case if optimization of motor behavior
implied by EP control were considered without identifying control variables
threshold positions. These principles may appear to be reduced forms of the
more general principle of minimal action.
A similar relationship exists between the threshold control theory and the
synergy concept which was formulated without consideration of the relationship between neural control variables and emerging output variables (see chapters
by Kelso and Latash in this book). It remains to be seen whether or not the
synergy concept can be reformulated in the framework of the TCT.
724
Future
Both empirical and modeling studies have also demonstrated the usefulness of
the TCT in the explanation of learning, memory, kinesthesia, and neurological
movement disorders (Feldman et al. 2007) and no doubt these aspects of the
TCT are worthy of further development. Even these future advances will not
spell the end of the development of the TCT. Where can one go from here? Note
that the scheme in Fig. 8 is not just an explanation of the physiological origin of
threshold control. It implies that motoneurons and many other neurons are
cognitive devices that recognize when the values of respective physical variables
delivered to their inputs from sensory receptors match the centrally specified,
referent values of the same variables. Moreover, these devices signal to muscles
or other neurons the degree of the discrepancy between these variables and thus
tend to diminish the discrepancy, as the principle of minimal action suggests.
Threshold control thus provides a principal answer to the question of how the
nervous system can identify that some internal (referent) and external (physical)
events match each other: motoneurons and neurons are skillful in this function.
The development of these, cognitive aspects of the TCT might be beneficial for
the understanding of the functioning of different brain structures. Indeed, the
cognitive aspects can be developed within the TCT based on the notion that
frames of reference for action and perception are united (see Introduction).
The history of the EP hypothesis and its development might be a good
illustration of the notion that problems cannot be solved by thinking within
the framework in which the problems were created (Albert Einstein; a quote
from Pound 2004). In essence, the posture-movement problem and redundancy
problems in multi-muscle and multi-joint control appeared as such within the
mechanical framework. As soon as thinking went beyond the mechanical
framework by taking into account threshold position control and the principle
of minimal action, these problems are solved in a natural way. Thereby, it came
out that neural control levels are not pre-occupied in solving these problems:
these levels guide motor actions by specifying where, rather than which and
how, neuromuscular elements should work and rely on the capacity of these
elements and the environment to yield each time a unique action.
Recommended reading for students: Von Holst and Mittelstaedt (1950/1973);
Matthews (1959); Feldman and Orlovsky (1972); Nichols and Steeves (1986);
Feldman and Latash (1982); Hogan and Flash (1987); Latash (1993); Feldman
and Levin 1995; Won and Hogan N (1995); Gribble et al. 1998; Levin (2000);
Ostry and Feldman (2003); St-Onge and Feldman (2004); Feldman (2008); Foisy
and Feldman (2006); Feldman et al. 2007; Pilon et al. (2007)
References
Adamovich SV, Archambault P, Ghafouri M, Levin MF, Poizner H, Feldman AG (2001)
Hand trajectory invariance in reaching movements involving the trunk. Exp Brain Res
138:411423.
725
Archambault P.S., Mihaltchev P., Levin M.F. and Feldman A.G. (2005) Basic elements of
arm postural control analyzed by unloading. Exp. Brain Res. 164:22541.
Asatryan DG, Feldman AG (1965) Functional tuning of the nervous system with control of
movements or maintenance of a steady posture: I. Mechanographic analysis of the work of
the joint on execution of a postural tasks. Biophysics USSR 10: 925935.
Bernstein NA (1967) The coordination and regulation of movements. Pergamon Press, Oxford.
Capaday C (1995) The effects of baclofen on the stretch reflex parameters of the cat. Exp
Brain Res. 104: 287296.
Colby CL (1998) Action-oriented spatial reference frames in cortex. Neuron 20 :1524.
Dubois DM (2001) Computing Anticipatory Systems. AIP Conference Proceedings, Vol. 573.
2001, XI, p. 706.
Fedirchuk B, Dai Y (2004) Monoamines increase the excitability of spinal neurones in the
neonatal rat by hyperpolarizing the threshold for action potential production. J Physiol.
557:355561.
Feldman AG (1986) Once more on the equilibrium-point hypothesis (l model) for motor
control. J Mot Behav 18(1):1754.
Feldman AG (2008) Equilibrium point control (an essay). In: Encyclopedia of Neuroscience.
Field: Computational Motor Control (ed. Amir Karniel), in press.
Feldman AG, Goussev V, Sangole A, Levin MF (2007) Threshold position control and the
principle of minimal interaction in motor actions. Progr Brain Res 165:267281.
Feldman AG, Latash ML (1982) Afferent and efferent components of joint position sense;
interpretation of kinaesthetic illusions. Biol Cybern 42 : 205214.
Feldman, A.G., Latash (2005) Testing hypotheses and the advancement of science: Recent
attempts to falsify the equilibrium-point hypothesis. Exp. Brain Res. 161: 91103
Feldman AG, Levin MF (1995) The origin and use of positional frames of reference in motor
control. Target article. Beh Brain Sci 18:723806.
Feldman AG, Orlovsky GN (1972) The influence of different descending systems on the tonic
stretch reflex in the cat. Exp Neurol 37:48194.
Ferris T (1977) The whole shebang. A state-of-the-universe(s) report. Simon & Schuster Inc, p. 13.
Foisy M, Feldman AG. (2006) Threshold control of arm posture and movement adaptation to
load. Exp Brain Res. 175(4):72644.
Gelfand, I. M., & Tsetlin, M. L. (1971) On mathematical modeling of mechanisms of central
nervous system. In: Models of the structural-functional organization of certain biological
systems, ed. I.M. Gelfand, V.S. Gurfinkel, S.V. Fomin & M.L. Tsetlin. MIT Press.
Gibson JJ (1966) The senses considered as perceptual systems. George Allen & Unwin Ltd,
Ruskin House, London.
Gribble PL, Ostry DJ, Sanguineti V, Laboissie`re R (1998) Are complex control signals
required for human arm movement? J Neurophysiol 79:14091424.
Gunther M, Ruder H (2003) Synthesis of two-dimensional human walking: a test of the
lambda-model. Biol Cybern 89:89106.
Henneman E and Mendell LM (1981) Functional organization of motoneuron pool and its
inputs. In |Handbook of Physiology, sec 1, vol ii. The nervous system: Motor Control, Part 1,
VB Brooks (ed.) Bethesda, Am Physiol Society, pp. 423507.
Hogan N, Flash T (1987) Moving gracefully: quantitative theories of motor coordination
Trends Neurosci 10:170174.
Hulliger M, Nordh E, Vallbo AB (1982) The absence of position response in spindle afferent
units from human finger muscles during accurate position holding. J Physiol. 322:
167179.
Jobin A, Levin MF (2000) Regulation of stretch reflex threshold in elbow flexors in children
with cerebral palsy: a new measure of spasticity. Dev Med Child Neurol 42 : 531540.
Lashley KS (1951) The problem of serial order in behaviour. In: Jefress La, ed. Cerebral
mechanisms in behaviour. Wiley, New York.
Latash ML (1993) Control of human movement. Human Kinetics Publishers, Champaign.
726
Levin MF (2000) Sensorimotor deficits in patients with central nervous system lesions:
Explanations based on the l model of motor control. Hum Mov Sci 19:107137.
Levin MF, Dimov M (1997) Spatial zones for muscle coactivation and the control of postural
stability. Brain Res 757:4359.
Levin MF, Selles RW, Verheul MHG, Meijer OG (2000). Deficits in the coordination of
agonist and antagonist muscles in stroke patients: Implications for normal motor control.
Brain Res. 853:352369.
Lestienne FG, Thullier F, Archambault P, Levin MF, Feldman AG (2000) Multi-muscle
control of head movements in monkeys: the referent configuration hypothesis. Neurosci
Lett 283: 6568.
Lippold OCJ (1952) The relation between the integrated action potentials in human muscles
and its isometric tension. J. Physiol. (L) 117: 492499.
Matthews PBC (1959) The dependence of tension upon extension in the stretch reflex of the
soleus muscle of the decerebrated cat. J Physiol (London) 147:5246.
Nichols TR, Steeves JD (1986) Resetting of resultant stiffness in ankle flexor and extensor
muscles in the decerebrated cat. Exp Brain Res 62: 401410.
Ostry DJ, Feldman AG (2003) A critical evaluation of the force control hypothesis in motor
control. Exp Brain Res 153:275288.
Pilon JF, Feldman AG. (2006) Threshold control of motor actions prevents destabilizing
effects of proprioceptive delays. Exp Brain Res, 174: 229239.
Pilon J-F, De Serres SJ, Feldman AG (2007) Threshold position control of arm movement
with anticipatory increase in grip force. Exp Brain Res. 181: 4967.
St-Onge N, Feldman A.G. (2004). Referent configuration of the body: A global factor in the
control of multiple skeletal muscles. Exp Brain Res 155:291300.
Vallbo AB (1974) Human muscle spindle discharge during isometric voluntary contractions.
Amplitude relations between spindle frequency and torque. Acta Physiol Scand 90 :319336.
Von Holst E, Mittelstaedt H (1950/1973) Daz reafferezprincip. Wechselwirkungen zwischen
Zentralnerven-system und Peripherie, Naturwiss. 37:467476, 1950. The reafference principle.
In: The behavioral physiology of animals and man. The collected papers of Erich von Holst.
Martin R (translator) University of Miami Press, Coral Gables, Florida, 1 p. 139173.
Weeks D.L., Aubert M.P., Feldman A.G. and Levin M.F. (1996) One-trial adaptation of
movement to changes in load. J. Neurophysiol 75:6074.
Won J, Hogan N (1995) Stability properties of human reaching movements. Exp Brain Res
107:125136.