motor control theory

397

SCHOLARLY PAPER

Motor Control Theories - Insights for therapists

Patricia Bate

Key Words Motor control theory, rehabilitation, movement re-training, brain injury.

Summary Several approaches to treatment are described for the rehabilitation of movement following brain injury. It is likely that therapists’ choices of treatment approach and methods of application are influenced by their assumptions about how movements are produced. The present paper offers therapists an opportunity to clarify and update their understanding of recent developments in the scientific field of motor control theory. Some hierarchal, neural network and dynamic models of movement production are described, and some assumptions about movement production and learning commonly applied in rehabilitation following brain injury are placed within this theoretical context.

For a hypothetical example of direct application by a therapist of theory about movement production, consider the fact that the scapula rotates upwards during gleno-humeral flexion. Knowledge of this fact may lead a therapist to look for th i s movement component in a patient’s performance, and/or to focus on specific training of scapular movement while practising reaching. On a more abstract level, consider the work of a therapist who assumes that movements emerge from interactions of the mechanical characteristics of a patient with the environment. Such a therapist may use anti-gravity positions and common objects more often in rehabilitation than a therapist who believes that movements are controlled by motor programmes stored in the cerebral cortex. The first therapist may request a seated patient to reach for a cup; the second may ask a patient to lie on his or her side and perform shoulder flexion movements.

Introduction It is likely that therapists are influenced in their work by their assumptions about how movements are produced (Horak, 1991; Shumway-Cook and Woollacott, 1995). This paper provides therapists with an opportunity to clarify and update these assumptions and to increase their understanding of recent developments in the scientific field of theories of movement production. Some current models of movement production are described, and common assumptions about it are placed within a theoretical context. Implications of these assumptions for rehabilitation of movement following brain injury are discussed.

The material under discussion is complex. Key points are listed as notes after each major section and illustrative examples and exploratory sections are given in boxes. The first examples comprise illustrations of ways in which assumptions about motor production may influence the practice of rehabilitation.

The scientific field describing movement production is known as ‘motor control theory’. I t is different from the field known as ‘motor learning’ because it is centred on the concept of movement production, rather than that of acquiring movement skills. However, motor control theories often support predictions about how skills are acquired, and motor skill learning theories often incorpo- rate assumptions about how movements are controlled.

Within the field of motor control, models of movement production are developed from theory and then tested and modified. Some of these models are based primarily o n information flow in computers, some are based on theory of physiology and anatomy, some on theory of human behaviour and some on physics. Because of their diverse origins, motor control theories may appear very different from each other.

Classification of Motor Control Theories Motor control theories may be classified in various ways, the allocation of each theory t o a group being dependent on the classification criteria. The structure of the present classification reflects sets of common assumptions, allows assumptions central to the field of clinical neurology to be high- lighted, and permits research-based discussion

Physiotherapy, August 1997, vol 83, no 8

398

of some more complex and recent material. Alter- native classifications are presented by Reed (19821, Horak (19911, Abernethy and Swallow (1992) and Shumway-Cook and Woollacott (1995).

For the purposes of this paper motor control theories were initially categorised into one of two broad groups. Theories in the information processing category explain motor control in terms of information flow in the nervous system of the organism. Sensory information flows in from outside, t o be used and stored, and movement commands flow outward; there is also extensive information flow between these two systems. These theories are based on anatomy and physiology and information flow in computers, and the behaviour of animals and humans in laboratory experiments. In contrast, the action category includes theories which are less reliant on storage of information and central representation of movements, and place greater emphasis on the role of the environment in movement generation. Some of these theories are based on the assumption that movements are an emergent property of the dynamics of the physical system comprising the organism and its environment.

The present is an interesting time to examine motor control theories, because a shift of paradigm may be occurring within the motor control field (Abernethy and Swallow, 1992). The information processing models of motor control have dominated the field for the last three decades, but some interpretations of the experimental evidence supporting these models are now being questioned. Support for models of motor control based on physical principles is growing rapidly (Gentner, 1987; Heuer, 1988).

Information Processing Models Models of motor control falling within the information processing category may be further classified as hierarchal or neural network models (Horak, 1991).

Hierarchal Models One of the assumptions of models categorised as ‘hierarchal’ is that the organisation of the nervous system is ‘top down’ (Horak, 1991). Arguing that the parts of the brain of more recent phylogenetic development are more important, supporters of this view consider that adaptive control of complex movements is performed by the cortex, and the more automatic control of simpler functions occurs at lower levels of the nervous system. This point of view has developed from decades of debate over whether the peripheral or more central parts of the nervous system control movement (Reed, 1982). Clinical implications of this view include the concept of release of reflexes due

to disruption of higher level control, and the idea that movements controlled by lower levels of the nervous system occur in stereotyped patterns (Horak, 1991). These models fail to explain the experimental finding that some complex functions such as the gait cycle are driven from a spinal level.

The hierarchal models emphasise control of movement by knowledge accumiilated through experience (Marteniuk et al , 1988) and by inten- tion (Proteau et a l , 1987) and they assume internal structure changes with learning. These assumptions appear to be held by most therapists, who expect that patients will remember some aspects of a movement practised the previous session.

The Generalised Motor Programme The concept of the motor programme was developed from studies of performances of simple motor tasks, to explain how high levels of the nervous system can store representations of a large number of movements. A generalised motor programme (GMP) is a rule for producing a proto- type movement; it can be viewed as a set of commands that is capable of carrying out movements without the use of feedback. The rule is general; values of various parameters have to be specified for a specific movement to be produced (Summers, 1981).

For example: think of playing tennis. The sight of the other player lining up a shot which will bring the ball to your forehand side may activate a GMP for forehand strokes; but the precise force and timing values may have to be specified further if your racquet is to contact the ball.

Evidence for the Use of GMPs 1. One of the strongest pieces of evidence for the existence of GMPs is the motor equivalence phenomenon. It is argued that for this phenomenon (the production of a functional movement such as writing, by different body parts) to exist, there must be a common set of instructions for which the body part to be used is an additional specification.

An apparent application of the notion of the GMP is seen in therapists’ assumptions that rehearsal of a skill using one body part may lead subsequently to more co-ordinated performance of the same skill by another part. While it is not known if this effect occurs in brain injured people, therapists sometimes ask for a required movement to be first performed by the opposite limb.

I


399

2. It has also been proposed that identification of aspects of movement which do not change with changing task conditions would support the existence of abstract motor programmes coded in such terms (Schmidt, 1988; Magill, 1989). This would be potentially useful in rehabilitation because knowledge of the ‘language’ in which movement is controlled may lead to more efficient design of motor retraining programmes. However, extensive experimentation has yielded conflicting findings. It has been proposed that some findings of invariances in temporal parameters of movement across different conditions (eg Shapiro et al, 1981) are effects of experimental design rather than real phenomena (Burgess-Limerick et al , 1992).

Producing Novel Movements Using the GMP One of the main advances made by the concept of a GMP was in its capacity to explain production of novel variations of movements (Summers, 1981). One of the experimental paradigms on which this model is based was pointing to a small target. Using this task as an example, such a novel variation could be produced by setting the target a different distance from the subject. Thus, if you have a GMP for the task of touching a call- button, then you can touch such a button 35 cm in front of you even if none of the call buttons you ever previously pressed were exactly that distance away. This type of novelty, the scaling of a pre- existing movement pattern, is the only type of new movement which can be explained by the concept of a GMP.

Without this assumption a therapist might consider it necessary to guide a brain injured patient through practice of every variation of every task he or she may ever need to perform.

However, the GMP concept cannot explain the development of new movement forms (that is acquisition of new GMPs).

Consider the first time you did a handstand. Did standing on your hands require a new GMP? If it did, then your performance could not be easily explained by this model.

[There is another possibility: a GMP for that task could have existed in your central nervous system, geneti- cally coded, never to be activated until that day. Perhaps you had to practise to find the appropriate values for the timing and force parameters of this pre-existing, never-used GMP.]

This limitation of hierarchal control theory offers insight in relation to the management of brain injured patients who are unable to perform a particular task, or who perform it very poorly. In considering the many factors contributing to

the poor performance therapists should include the possibility that the patient may no longer have access to a GMP for the movement. Unfor- tunately hierarchal models offer no prediction of how development of another GMP should be attempted.

Much of the empirical literature describes the improvements in outcomes of motor tasks which occur with practice as continuous functions (Newell, 1991), supporting the predictions of hierarchal models of motor control. However, learning in brain damaged subjects has been shown to comprise irregular increments in task outcome rather than smooth progressions (for example: Neilson and McCaughey, 1982; Bate and Matyas, 1992). This difference may be due to the fact that while the tasks practised by the normal subjects were already well learned (Newell et al, 1989), the task performances of brain injured subjects may have been at an earlier stage of learning. Alternatively, the irregular progress of brain damaged subjects may indicate that their learning in rehabilitation does not occur in the manner predicted by hierarchal models of motor control.

Schema Theory Schema theory (Schmidt, 1975), while developed primarily to describe motor learning, has many features characteristic of hierarchal theories of motor control. The model states that a GMP initi- ates a movement, and that feedback from the movement is compared with a generalised sensory representation to allow corrections and learning.

Learning with Schema Theory Unlike earlier models, schema theory predicts that variability in task conditions will enhance learning. That is, the GMP will become more accurate if variations on the movement are performed, rather as the predictions of a mathematical relationship become more reliable if more points are added to the data set. This prediction has been supported in a study of stroke patients learning a pursuit tracking task (Bate et al, 1992).

This model also predicts that we do not learn from passive movements. Assuming this to be true, therapists are challenged to create situations in which action is demanded from a paralysed limb.

Finally, schema theory predicts that learning requires active processing of feedback; in information theory terms this means it requires attention. Consonant with this model of learning, movement rehabilitation places heavy emphasis on patients’ attention to feedback (Carr et al , 1987; Bobath, 1990).

Physiotherapy, August 1997, vol83, no 8

400

Notes 1. The assumption that the cortex controls complex movements and lower centres control more automatic, simpler movements is probably wrong.

2. Hierarchal control theories assume that movements are controlled by knowledge accumulated through experience, and that internal structure changes with learning; these assumptions appear consistent with the practice of rehabilitation.

3. The concept of a GMP appears to explain posi- tive transfer of a skill to performance by an untrained part of the body.

4. The prediction of hierarchal motor control theories that improvement is a continuous progression mays not reflect the pattern of changes occurring during rehabilitation following brain injury.

5. Schema theory predicts that practice of variations of a movement will improve performance.

6. Schema theory predicts that we do not learn from passive movements; therapists create situations affording active movements.

7. Schema theory predicts that learning requires attention to feedback; this prediction is consistent with practice in rehabilitation.

Neural Network Models In these models of motor control, simple processing units consisting of neurones and their extensive systems of interconnecting dendrites are linked t o form nets. These networks take on activation levels and send excitatory and inhibitory signals. Communication within and between the nets is far more extensive than is assumed in hierarchal models, and at all levels and in all directions within the central nervous system, rather than predominantly in inward or outward directions as in the hierarchal models. Goal directed actions are believed to emerge from transmissions between overlapping neural networks which serve particular functions.

The possibility that movement is supported by neural networks is suggested by the findings of recent investigations of the speech system (Abbs et al, 19841, the eye-head-hand system (Carnahan and Marteniuk, 1991) and the prehension system (Cole and Abbs, 1986). These studies have shown that when one component of the system is unex- pectedly perturbed, interpretable speech or accurate touch is still achieved because of very rapid modifications of the movements of many components. Such control of a multi-movement system requires extensive ongoing communication by feedback and feedforward processes. It is argued that because this type of extensive communication is characteristic of neural nets, but does not feature in hierarchal models, it is more likely that complex functional movements

are controlled by neural nets (Abbs et al , 1978). These models assume that the parameter encoded by the controlling system is achievement of the functional goal rather than control of movement components. A therapist working under this assumption ensures that movement retraining is conducted in the context of achievement of function.

The complex fast communications possible within a neural network would allow computationally demanding methods of motor control such as those that require extensive calculations during movement performance. There is strong support for the operation of this type of control in human movement, and several such models offer intuitive appeal when considered from a rehabilitation frame of reference, because they support some assumptions inherent in rehabilitation practice.

Hogan (1984) has presented a mathematical model describing the organisation of voluntary movements so that the smoothest trajectory possible is achieved: that is, the rate of change of acceleration is minimised. The model has been shown accurately to predict position and velocity profiles of multi-joint movements (Flash and Hogan, 19821, and thus has some demonstrated validity as a control mechanism for human movement. The key assumption of the model, that movement is controlled to be as smooth as possible, is consonant with the high value often placed on smoothness of movement in rehabilitation following brain injury.

Three other models requiring the extensive computation capacity existing in a neural network reflect characteristics of movements often encouraged in rehabilitation. Investigating the movements of subjects instructed to move comfortably, Cruse and colleagues (1993) derived a mathematical model in which the characteristics of comfortable movements could be reproduced using a minimum cost principle requiring that joint movement be equally distrib- uted among the available joints. This model thus supports the assumptions that a ‘comfortable’ movement is one of minimal energy requirement, and that all available joints are involved when a ‘comfortable’ movement is produced.

Also along the lines of minimum effort, Hirayama and colleagues (1993) successfully tested a model in which the initial feedforward command was based on a minimum torque criterion, and Uno and colleagues (1989) developed a model of motor control based on the principle of minimum muscle tension change. These models support the assumptions that movements are generated which use the minimum force or muscle tension necessary to achieve the movement goal. These


401

assumptions also appear to be reflected in practice of rehabilitation with brain injured patients.

Transformation of Information The neural network models require that information is transmitted between neural nets which serve particular functions. It is proposed that information is encoded in different frames of reference within each net; and that the information must be transformed so that the meaning is retained, for each transmission between nets.

Consider a linguistic analogy: it is thought that communication within each subsystem of the neural network is in a different language. In order for subsystems to talk to each other, the information must be translated. Alternatively, consider the transformations which may be required in processing information for a person reaching to an object. Knowledge of the position of the object relative to the gaze, the orientation of the eyes relative to the head, the head to the t runk and the arm to the trunk are necessary to determine the position of the hand relative to the object. Each of these pieces of information may be stored as co-ordinates in a different sensory co-ordinate subsystem, t h u s four transformations would be required in processing this perceptual information. Further, because of the nature of our receptors, perceptual information is probably coded in kinematic terms as spatio-temporal patterns. However, many motor control models suggest that movements are controlled dynamically; that is, they propose that the language of motor control includes force in addition to space and time information. If this is so, production of the movement of reaching for an object would involve additional transformations. A s well as the transformations between extrinsic and intrinsic kinematic co-ordinates described in the preceding paragraph, transformations between kinematics and dynamics and between dynamics and muscle activation would be necessary (Soechting, 1989).

The concept of transformation of information between frames of reference or co-ordinate systems is interesting when considered in relation to rehabilitation for brain injured patients. It offers the possibility that movement re-training is effective by increasing the accuracy of these transformations. Perhaps when movements are a little awkward or halting or innaccurate this is due to imprecision in transformation of information between subsystems. Rehabilitation may offer opportunity for re-calibration or fine-tuning between subsystems. Bingham (1988) has proposed a definition of learning which is congruent with this possibility. He suggested that learning is the selection or discovery of constraints that minimise the computations necessary t o transform information between frames of reference.

Also interesting in terms of rehabilitation is the concept of coding information within each subsystem. It may be that task practice increases the accuracy with which points are located within a co-ordinate system. Investigating possible co-ordinate systems for a target-directed arm movement, Soechting (1982) demonstrated that the arm movements were performed within a frame of reference centred at the shoulder, relative to absolute vertical and horizontal planes. This finding offers clues for potentially useful imagery in re-training. If the co-ordinate systems of particular motor functions are known, it may be effective to use these systems in movement instructions.

Notes 8. The multi-movement model of motor control assumes that the goal encoded by the controlling system is achievement of the function rather than control of contributing movement components.

9. Neural network models assume that motor control is computationally demanding.

10. Some neural network models assume movements are controlled in terms of smoothness of trajectory, or minimisation of force, energy require- ments or muscle tension, or in terms of distribution of movement through all available joints.

11. Rehabilitation may be effective through increasing the accuracy with which information is encoded within subsystems of the neural net, or the accuracy of transformations of information between subsystems.

Action Theories Action theories (Reed, 1982) apply a more macro- scopic lens to the topic of movement production than the information processing approach. They also accent different aspects of movement production, and as yet have not clearly addressed the concept of learning. Functional actions are viewed broadly as emerging from the relationship between the organism and the environment. Rather than assuming two sequential processes within the organism, one in which movement results from activation of a pre-existing motor programme and another in which movement is modified upon receipt of feedback of perceived changes, as in hierarchal models, action theories identify the mutualism of the relationship between the organism and the environment as the source of movement. The emphasis upon the integrative nature of the movement production process extends to the assumption that in this process perception and movement are so entwined that it is not valid to consider either in isolation. Instead, they are considered to operate together


402

in an arena commonly known as the ‘perceptuo- motor workspace (perceptuo-motor field)’ (Newell et al, 1989).

Field-oriented Theory of Perception An understanding of the integral nature of the relationship between perception and movement emphasised in action theories can be gained by interpreting actions in terms of Gibson’s (1979) field-oriented theory of perception.

This theory emphasises the complex role of the optic flow field in movement production. The optic flow field is the pattern of stimuli on the retina produced by the interaction between the organism and the environment. This optic flow field is controlled by movement, and it also provides information for both perceiving and action.

Imagine a horse running fast on an open plain. As it runs, the long grass in front of it appears to the horse to move closer, and then to flow past on each side. The horse sees the grass stems on either side moving backwards very fast, and the mountains moving more slowly. The horse swerves in the direction of a tree, and the tree appears to it to move in the opposite direction, towards the centre of the field of view, and to move closer. In this example the optic flow field has provided information about speed and direction of running. Also, change in the speed and direction of running caused concomitant change in the position and movement of the stimuli impinging on the retina. That is, movement changed the optic flow field, and the optic flow field provided information controlling both perception and action.

Consider that you are landing an aeroplane in a computer simulation. You hold a joystick, and are watching a moving graphical display of the view from the pilot’s seat. This includes ground features such as bushes and grass around the landing strip, the edges of the landing strip, and lines painted across the strip every 20 metres. You have no other information. You watch the relative positions and changes in position of these visual stimuli and use this information to control the height, pitch, yaw and roll of the plane through the joystick. That is, the optic flow field changed the movement of your arm, and hence the aeroplane. In this example also, while movement changed the optic flow field, the optic flow field provided information controlling both perception and action.

Following publication of Gibson’s theory, it was proposed that the organising role of optic flow in movement production may be extended to other modalities (Newell et al, 1989). That is, in the language of action models, not only the visuo- motor field, but more broadly, the perceptuo- motor field, may organise movement.

Action theories, in which movements are gener-

ated in real time from interaction between the organism and the environment, predict that novel movements would arise if the environment, the organism, or their interaction was altered. Thus, some action theories predict that people could effectively handle a joystick if their environment suddenly included an unconscious pilot, even if they had not done so before.

The environment is given more importance in this model than in traditional forms of rehabilitation. The implication for rehabilitation following brain injury is that therapists could change the environment in such a way that the interaction of the patient with the environment generates the required movement. In addition to changing the patients’ cognitive environment by giving specific instructions, or altering their biomechanical environment by applying a mod- ality such as ultrasound or stretching, their external environment could be modified so that it facilitates better performance.

Action theories such as the field-oriented theory of perception (Gibson, 1979), which emphasise the role of perception in motor control, do not clearly address the issue of learning. It is interesting to speculate what effects movement training may have in terms of this model. Presumably finer discrimination of sensory stimuli, and finer tuning of motor mechanisms, could lead to generation of more finely co-ordinated movements. However, the way in which such increases in capacity of the organism would occur is not specified in many action models.

Dynamic Emphasis in Action Theories Emphasising the importance of interaction between the organism and the environment in movement production Kugler et al (1980) eluci- dated a constraint perspective of motor control which was based on the work of Feldman (1986), Polit and Bizzi (1978) and Bernstein (1967). This model explains the use of constraints in movement production in terms of physics and biology; we obey the laws of mechanics, acting like springs and pendulums and more complex systems in interacting with the environment. In contrast to neural network theories, which support a large amount of information processing occurring in parallel systems, one of the effects of control based on dynamic principles is minimisation of the computational load on the nervous system.

For each task some of the dynamic properties of the human action system are assembled, temporarily, into a ‘low-dimensional deterministic machine that is used to achieve the goals’ of the task (Bingham, 1988, page 240). The term ‘low- dimensional’ here refers to the small number of variables which would have to be controlled by computational processes.


403

Picture yourself returning from the supermarket with a bag of groceries. You put them on a low bench in the kitchen. Among assorted items is a large plastic container of juice with a handle formed from the body of the container. You reach out with your right hand and hook two fingers around the handle of the juice container as you turn toward the refrigerator. You let the container drop to the full length of your arm, it swings out a little away from your body dragging your arm th rough a curvilinear path to the refrigerator. Toward the end of its arc the container gains just sufficient height to graze the edge of the lowest shelf. You add a little shove and it slides to rest. Notice the effortlessness of the way in which the juice was transported. What sort of mechanical system (ie temporary low-dimensional deterministic machine) did your arm become? What principles of physics were exploited? Perhaps the muscles of the arm were constrained to act like a pendulum, the falling mass of the bottle providing momentum for the swing u p to the refrigerator shelf at the far end of its arc.

A Dynamic Perspective of Learning While action models make few specific predictions about learning, Bernstein (19671, the father of dynamic theory, stated that learning is developing groupings of muscles set to have the particular properties which will enable them t o interact mechanically with the forces of the environment in such a way that the task is performed mechanically except for minor tuning. It appears that rehabilitation could provide extensive opport- unities to experience movement in various environments, while also experiencing a demand for reduction of the energy used in moving. These circumstances may lead to development of the properties referred to by Bernstein (1967).

Equilibrium Point Hypothesis This can be considered t o be one of the more simple dynamic models, although there are several current versions of this model and they appear to be growing in complexity. Simply represented, it is assumed that the muscles moving a body segment can be represented by springs whose resting length can be set at some value. Motor programmes dictate relative lengths and stiffnesses of muscles, the values of which define positions of the end of the limb in space (Polit and Bizzi, 1978; Feldman, 1986; Hogan et aZ, 1987). I t is possible t o imagine ratios of muscle stiffness set up in such a way that the arm will swing like a simple pendulum, coming to rest at the set equilibrium point.

Examination of the model predicts a mechanism by which movement rehabilitation may be

effective; that is, by re-setting the resting 1engtWtension ratios of muscle. This could lead to changes in resting posture, sensations of greater ease of movement, and availability of a broader range of movement combinations.

Control of Complex Movement Newel1 et aZ(1989) have pointed out that co-ordination of complex movements has not been extensively studied. Most motor control models explain performance and learning of simple, bidirectional, well learned movements; for example, schema theory, and the equilibrium point hypothesis (Feldman, 1986).

Knowledge about the control of complex functional tasks such as locomotion or balance, which continue cyclically over time, may be particularly relevant in the rehabilitation of movement following brain injury, in which such functions are important goals. This control can be modelled as pattern formation in dynamical systems (Turvey, 1990).

Dynamic systems are systems in which behav- iours evolve in time. Using gait as the system to be illustrated: as the limbs start moving, kinematic and kinetic measures of gait can be seen to pass through some variations and then to settle into repeating patterns which can be described by cadence and speed. The mechanics of the organism and the environment, the goals and energy supply of the organism, and its perceptual feedforward and feedback subsystems, define these regions of stable equilibrium. As the system tends to return to regions of stable equilibrium, a regular walking pattern will emerge.

Another example of a region of stable equilibrium, or a constraint, which may operate in the perceptuo-motor workspace of a dynamical system, is provided in the Equilibrium Point Hypothesis. In this model of motor control, set ratios of stiffness around a joint represent a state to which the system will tend to return - ie t o which it will be attracted. This type of constraint is termed a ‘point attractor’. Another type of constraint is the stable limit cycle attractor. Oper- ation of this type of attractor underpins the easy way in which a seated person can rhythmically swing a leg from the knee. Metabolic energy is injected periodically into the system, potentially increasing the kinetic energy; the energy is dissi- pated as heat in such a way that the velocity and position of the limb cycle regularly in time.

Particular functions, such as gait, are identified with particular kinematic patterns. These are assembled by the generation of particular dynamical attractor states in the perceptuo-motor


404

work-space specific t o that function. The form of a movement pattern or co-ordinative mode is identified with the stable equilibrium regions in the perceptuo-motor workspace.

Dynamic systems may also be characterised by their capacity to change state. When values of a particular variable are changed and reach a crit- ical level, a new region of stable equilibrium is generated (Shumway-Cook and Woollacott, 1995). The gait system appears to show such behaviour; if the speed of walking increases, a t a certain speed the form of the movement pattern changes and it becomes running.

These models hold an important implication for rehabilitation in relation to the passive components of a patient's biomechanics. Normal lengths and stiffness ratios of soft tissue must be achieved before the movement may settle to a region of stable equilibrium close to normal.

Learning of Complex Movements A dynamic perspective of control can describe learning of both simple and complex movements. The learner can be seen as exploring the perceptuo-motor workspace for attractors, and mapping the annihilation and creation of these properties as they evolve as a function of motion through space and time (Newel1 et al, 1989).

This description of the learning process appears to match the structure for movement development offered in the form of creative dance and awareness through movement (Feldenkrais, 1980) lessons. In these situations constraints such as unusual skeletal configurations, unusual movement goals, and proximity t o the floor are introduced, and students are encouraged t o explore this workspace. The exploration is sometimes conducted in a rhythmic fashion.

Considering the dynamic perspective in relation t o the generation of novel movements, i t seems likely that the type of exploration described could lead to discovery of new movement solutions or forms. The model could thus be applicable in rehabilitation in any case where a patient is viewed as requiring new movement patterns.

The dynamic models do not predict continuous improvements in task performance. They predict that regular measures of performance conducted over a practice period will demonstrate irregular increments in skill as new movement forms are generated and explored. This model of learning may match the process of motor skill acquisition following brain damage better than the predictions of continuous improvement during learning made from hierarchal information processing models.

Notes 12. Action theories assume that the relationship between the organism and the environment is the origin of movement.

13. Gibson's (1 979) theory assumes that the optic flow field controls, and is controlled by, movement.

14. Dynamical models assume that the human action system is assembled temporarily into mechanical systems suitable for performing particular tasks by exploiting the laws of physics.

15. The equilibrium point hypothesis predicts that postures and movement styles are determined by IengthAension relationships in muscles.

16. Interpreting motor control in terms of dissipative dynamic systems, it is assumed that the perceptuo- motor system generates particular combinations and values of variables (eg particular constraints) that makes it tend to move in a particular way for a particular function. Learning is assumed to be exploration for optimal variables and settings, and improvements in task outcomes may not be continuous.

Conclusion The disparate and apparently conflicting nature of the various models of motor control is probably a function of their diverse sources of origin, of different levels of observation, and of great versa- tility in modes of action of the human organism. The range of motor control methods utilised in various situations may be very great; it is con- ceivable that one person could demonstrate characteristics of equilibrium point control to a physiologist measuring activity in the brachialis muscle in response to stretch, and also display the physics of a dissipative dynamic system to a physicist observing running at different speeds on a treadmill.

While a paradigm which incorporates all aspects of movement has not yet emerged in the field of motor control, the current models offer extensive challenge and scope to therapists. Because we are likely to apply our assumptions about motor control in our work, either explicitly or even without being aware that we are doing so, it is important that we remain informed of the sources and limitations of these assumptions, and thus of the situations in which they are likely t o be valid. We can also use the field of motor control as a source for new directions in rehabilitation of movement.

This review of the information processing and action models of motor control has identified aspects of the theories of movement production that appear to have potential to explain some of the practices in rehabilitation following brain injury. I t has also identified issues worthy of


405

further discussion by therapists. Some of the predictions of the theories reviewed are empiri- cally testable within the rehabilitation paradigm, offering opportunity for sound research within our field.

Author and Address for Correspondence Patricia Bate MAppSci BAppSc(Phty) MAPA is a senior lecturer in the School of Physiotherapy, Faculty of Health Sciences, La Trobe University, Plenty Road, Bundoora, Victoria, Australia.

This article was received on October 10, 1996, and accepted on December 2, 1996.

References Abbs, J H and Cole, K J (1978). 'Neural mechanisms of motor equivalence and goal achievement' in: Wise, S P (ed) Higher Brain Functions, Wiley, USA.

Abbs, J H, Gracco, V L and Cole, K J (1984). 'Control of multi- movement co-ordination: Sensorimotor mechanisms in speech motor programming', Journal of Motor Behavior, 16, 195-231.

Abernethy, B and Swallow, A (1992). 'The rise and fall of domi- nant paradigms in motor behaviour research' in: Summers, J J (ed) Approaches to the Study of Motor Control and Learning, Elsevier Science, Amsterdam, pages 343-384.

Bate, P J and Matyas, T M (1992). 'Negative transfer of training following brief practice of elbow tracking movements with electromyographic feedback from spastic antagonists', Archives of Physical Medicine and Rehabilitation, 73, 1050-58.

Bate, P J, Matyas, T and Rogers, D R (1992). 'Movement rehabilitation following brain damage: Transfer of training after consultant or variable practice,' presented at Allied Health Research Seminar, Austin Hospital, Melbourne.

Bernstein, N A (1967). The Co-ordination and Regulation of Movement', Pergamon Press, Sydney.

Bingham, G P (1988). 'Task specific devices and the perceptual bottleneck', Human Movement Science, 7,225-264.

Bobath, B (1 990). Adult Hemiplegia: Evaluation and treatment, Heinemann, London, 3rd edn.

Burgess-Limerick, R, Neal, R J and Abernethy, B (1992). 'Against relatiye timing invariance in movement kinematics', Quarterly Journal of Experimental Psychology, 44A, 4k, 705-722.

Carr, J H, Shepherd, R S, Gordon, J, Gentile, A M and Held, J M (1 987). Movement Science: Foundations for physical therapy in rehabilitation, Aspen, Maryland.

Carnahan, H and Marteniuk, R G (1991). 'The temporal organisation of hand, eye and head movements during reaching and pointing', Journal of Motor Behavior, 23, 2, 109-1 19.

Cole, K J and Abbs, J H (1986). 'Kinematic and electromyographic responses to perturbation of a rapid grasp', Journal of Neurophysiology, 57, 1498-1 51 0. Cruse, H, Bruwer, M and Dean, J (1993). 'Control of three- and four-joint arm movement: Strategies for a manipulator with redun- dant degrees of freedom', Journal of Motor B e / ~ v i o r ,

Feldman, A G (1986). 'Once more upon the equilibrium-point hypothesis (lambda model) for motor control', Journal of Motor Behavior, 18, 17-54.

Feldenkrais, M (1 980). Awareness Through Movement, Penguin, Middlesex.

Flash, T and Hogan, N (1982). 'Evidence for an optimisation strategy in arm trajectory formation', Society of Neuroscience Abstracts, 8, 282.

Gentner, D R (1987). 'Timing of skilled motor performance: Tests of the proportional duration hypothesis', Psychological Review,

Gibson, J J (1979). The Ecological Approach to Visual Percep- tion, Houghton Mifflin, Boston.

25,3, 131-1 39.

94,255-276.

Heuer, H (1988). 'Testing the invariance of relative timing: Comment on Gentner (1 987). Psychological Review, 95,

Hirayama, M, Jordan, M I and Kawato, M (1993). 'The cascade neural network model and a speed-accuracy trade-off of arm movement', Journal of Motor Behavior, 25, 3, 162-1 74.

Hogan, N (1984). 'An organising principle for a class of voluntary movements', Journal of Neuroscience, 4, 2745-54.

Hogan, N, Bizzi, E, Mussa-lvaldi, F and Flash, T (1987). 'Control- ling multijoint motor behaviour', Exercise and Sports Science Reviews, 15, 153-190.

Horak, F (1 991). 'Assumptions underlying motor control for neurologic rehabilitation', in: Contemporary Management of Motor Control Problems, Proceedings, 11 Step Conference, Foundation for Physical Therapy, New York.

Kugler, A N, Kelso, J A S and Turvey, M T (1980). 'On the concept of co-ordinative structures as dissipative structures. 1: Theoretical lines of convergence' in: Stelmach, G E and Requin, J (eds) Tutorials in Motor Behaviour, Elsevier Science, Amsterdam.

Magill, R A (1989). Motor Learning: Concepts and applications, W C Brown, Indianapolis, 3rd edn.

Marteniuk, R G, McKenzie, C L and Leavitt, J L (1988). 'Repre- sentational physical accounts of motor control and learning: Can they account for the data?' in: Colley, A M and Beech, J R (eds) Cognition and Skilled Behaviour, Elsevier, Amsterdam.

Neilsen, P D and McCaughey, P (1 982). 'Self regulation of spasm and spasticity in cerebral palsy', Journal of Neurology, Neuro- surgery and Psychiatry, 45, 320-330.

Newell, K M, Kughler, P N, Van Emmerik, R A E and McDonald, P V (1989). 'Search strategies and the acquisition of co-ordination', in: Wallace, S A (ed) Perspectives on the Co-ordination of Movement, Elsevier, Amsterdam.

Newell, K (1991). 'Motor skill acquisition', Annual Review of

Polit, A and Bizzi, E (1978). 'Processes controlling arm movements in monkeys', Science, 201, September 29, 1235-37.

Proteau, L, Marteniuk, R G, Gerouard, Y and Dugas, C (1987). 'On the type of information used to control aiming movements after moderate and extensive training', Canadian Journal of

Reed, E S (1982). 'An outline of a theory of action systems, Journal of Motor Behavior, 14, 2, 98-134.

Schmidt, R A (1975). 'A schema theory of discrete motor skill learning', Psychological Review, 82, 225-260.

Schmidt, R A (1 988). Motor Control and Learning: A behavioral emphasis, Human Kinetics, Illinois, 2nd edn.

Shapiro, D C, Zernicke, R F, Gregor, R J and Diestal, J D (1981). 'Evidence of generalised motor programmes using gait pattern analysis', Journal of Motor Behavior, 13, 33-47.

Shumway-Cook, A and Woollacott, M (1995). Motor Control: Theory and practical application, Williams and Wilkins, Baltimore.

Soechting, J F (1982). 'Does position sense at the elbow reflect a sense of joint angle or one of limb orientation?' Brain Research, 248, 393.

Soechting, J F (1 989). 'Elements of co-ordinated arm movements in three-dimensional space' in: Wallace S A, Perspectives on the Co-ordination of Movemenf, Elsevier, Amsterdam.

Summers, J J (1981). 'Motor programs' in: Holding, D (ed) Human Skills, John Wiley and Sons, New York. Summers, J J (1992). 'Movement behaviour: A field in crisis?' in: Summers, J J (ed) Approaches to the Study of Motor Control and Learning, Elsevier, Amsterdam. Turvey, M T (1 990). 'Co-ordination', American Psychologist,

Uno, Y, Suzuki, R and Kawato, M (1989)..'Minimum muscle tension change model which reproduces a human arm movement', Proceedings of the Fourth Symposium on Biological and Physical Engineering, Fukuoka, Japan, 299-302.

552-557.

Psychology, 42,213-237.

PSyCholOgy, 6, 181-199.

August, 938-953.


motor control theory

Documents