Tristan D. M. Roberts (1996) Understanding Balance:the Mechanics of Posture and Locomotion. Psycoloquy: 7(02) Posture Locomotion (1)

Volume: 7 (next, prev) Issue: 02 (next, prev) Article: 1 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 7(02): Understanding Balance:the Mechanics of Posture and Locomotion

[Chapman & Hall 1995, xi+348 pp. ISBN 04126016051565934164]
Precis of Roberts on Posture-Locomotion

Tristan D. M. Roberts
(Formerly Reader in Physiology
University of Glasgow, Scotland)
11 Menzies Drive, Fintry
Glasgow G63 OYG


Balancing the body's loosely-jointed assemblage of heavy segments on one another during standing and locomotion calls for separate and continuous regulation of each of a very large number of independent motor units. Success depends on triggering appropriate anticipatory pre-emptive actions by gestalt recognition of specific developing trends in proprioceptive, and other, neural signals. This book discusses what forces are actually required and the nature of the signals available for guidance, concluding with a realistic scheme of operation for the intricately interconnected neurons in the brain, with implications for explaining the basis of gestalt recognition in other types of sensory perception also.


Balance, connectionism, gestalt, learning, locomotion, motor control, proprioception, recognition.
1. Balance is maintained by developing, against the available supports, forces suitably adjusted in timing, magnitude and direction. The oversimplified schoolroom physics of rigid bodies is not appropriate to body segments made up of soft tissue as well as bone, and the application of Newton's Laws has to be re-examined, particularly in relation to the distinction between stress and gravity and to D'Alembert's procedure for analysing problems in dynamics by the methods of statics.

2. Stress forces are the bulk manifestations of the intermolecular forces responsible for the cohesion and relative rigidity of solid objects. The convention of representing the forces in a structure in terms of an equivalent triangulated lattice demands that forces be regarded as acting at a point and that the struts and ties forming the lattice be linked by frictionless pin joints. These assumptions are unrealistic in that the relevant stresses are distributed over areas of significant magnitude. The locus of the resultant of the stresses at the area of contact with the supports can be determined with a force-plate of suitable design but analysis of the way forces are transmitted from one part of the body to another involves many uncertainties. Nevertheless, a number of interesting mechanisms can be identified.

3. A fully triangulated lattice structure is not deformable without changing the lengths of some of its members. The body uses muscles crossing a single joint in three-dimensional structures to convert tensions into thrusts, and to transmit forces from one part of the body to another. Muscles spanning two joints form four-bar linkages of two kinds. An open four-bar linkage is readily deformable in shear, but is useful in the body to transmit torque along the length of a jointed limb. Crossed four-bar linkages can act like a hinge, as in the case of the cruciate ligaments in the knee-joint, and in the support of the head on the neck. Muscles spanning more than two joints may form triangles with complex struts which are themselves made up of triangulated structures, as in the long muscles of the neck and back. Alternatively, they may be used to fold up the whole of a limb, as in the furling phase of a step, where the foot has to be lifted clear of obstacles at the start of the swing phase.

4. The limbs are required not only to resist longitudinal compression, as in propping, but also to transmit torque, in paddle action, to shift the effective points of application of forces against the trunk. The muscles that span two joints in open linkages are well placed for paddle action but do not contribute to propping except in conjunction with other muscles that span only a single joint. The geometry of the rolling joint at the knee, together with the spring-like action of the reflexly active single-joint extensor, results in action as a constant-force strut, developing the same overall thrust at any length in a specific range, a feature which has clear advantages in control system economy when supporting the trunk while it is being moved over the supports. The low friction at the synovial joints is accounted for by the quite complicated physical chemistry of articular cartilage. The forces which support the body are all derived from muscle activity. The necessary tensions are provided by the cooperation of a very large number of individually controlled "motor units", but the relationship between control signal and force generation is far from simple.

5. Regularly arranged filaments of different chemical composition within the muscle cells are caused to interact through a cascade of electrical and chemical changes. The energy for the performance of mechanical work is derived, ultimately, from the oxidation of foodstuffs. At one stage the breakdown of ATP provides the energy to prime a molecular spring-and-catch mechanism that stores elastic strain energy in the detached side-branches of the myosin filaments. When calcium ions are liberated by the incoming electrical signal, these side-branches become attached to the adjacent actin filaments and the catch is released. Tension then develops immediately in the resulting cross-bridges. This tension is transmitted, longitudinally, from sarcomere to sarcomere, and eventually to the attached tendon. When a muscle is made active, it will do one or more of three things: 1) it will shorten if it can; 2) it will develop tension against a resistance; 3) it will show an increased resistance to extension.

6. The tension generated within a single sarcomere depends on the degree of overlap between actin and myosin filaments, since overlap is essential for the formation of cross-bridges. Furthermore, the structures lying mechanically in series with the tension-generating sites are all extensible to varying degrees. Some tendons are more extensible than others, depending on the precise mix of the constituent proteins, collagen and elastin. Collagen fibres are relatively inextensible, accounting for the useful properties of leather straps. Elastin fibres, on the other hand, have a specialised molecular architecture that confers properties resembling those of lightly vulcanised rubber. A tendon or ligament containing a large proportion of elastin fibres can thus act as a store of elastic strain energy, a property that is important in the bouncing phases of locomotion.

7. Changing the length of the tension-generating mechanism involves the release and reformation of the cross-bridges between the filaments. In consequence, less force is available during active shortening. Forced extension at moderate speeds is resisted, while more rapid forced extension results in slippage between the filaments. The calcium ions that catalyse the formation of cross-bridges are rapidly reabsorbed after each activity cycle, provided that the oxygen supply is maintained. The resulting cycle of mechanical activity is thus rather brief, and sustained tensions call for repetitive activations. There are disadvantages in increasing the frequency of repetitive activation beyond a certain point, and adjustments in the tension transmitted to bone are usually achieved by regulating the number of motor units that are active at any one time.

8. In pulling upon the bones, the muscles are called upon to perform tasks of different kinds: 1) setting the angle of an unloaded joint to a new position; 2) holding the prevailing joint-angle in the face of a varying load; 3) performing work by driving to a new position against a load; 4) absorbing work by yielding while continuing to exert tension. Such tasks arise continually at each and every one of the very many joints in the body. Coordination is achieved by a hierarchy of mechanisms, some reflex, others depending on learned voluntary components. Details are given of the reflexes of standing, of the reflexes of balance (including the interactions between neck reflexes and labyrinth reflexes), of the reaction to tilting and to overbalancing, and of the combinations of these that produce the various modes of locomotion.

9. The question of timing is important in all these reactions. It is a matter of threshold. Conditions describable as "below threshold" are readily distinguishable from those "above threshold", but it is not possible to be specific about the precise moment at which the threshold is crossed. The problem is particularly acute in many ongoing situations where both the input information streams and the output command streams are subject to continual updating.

10. Comparable occurrences of "achievement events" in the course of "ongoing processes" underly the phenomena of "gestalt recognition". This term applies where an input pattern is recognised and classified, as evidenced by the resulting behaviour, on the basis of roughly cotemporal presentation of components of sensory input of specific kinds. In the recognition of a gestalt, the full set of input components that might be supposed to constitute the "adequate stimulus situation" do not all have to be present at the same time. Some items from the set may be included or others omitted without affecting the resultant classification or the response. Since motor acts can be initiated without conscious perception, it is reasonable to speak of the triggering by the input stream in terms of "gestalt recognition", rather than of "gestalt perception".

11. When a person, standing on one leg with the other leg tethered, is overbalanced by an external force, a hop is executed. The crucial triggering signal might arise from several sources: the skin of the foot (detecting movements of the centre of pressure); the ankle joint (detecting tilt); the hip joint (involved in certain relevant swaying movements); and the accelerations (horizontal, vertical, and angular) of the head. Each type of signal is clearly relevant, but each can be eliminated in turn by appropriate experimental design without failure of the hop to be developed. It may be concluded that the trigger consists of the detection, not necessarily consciously, of the gestalt "that overbalancing is imminent".

12. Hops of two kinds are distinguishable. If the hop is made by the subject in response to a command, the hop is invariably preceded by a dip in the force record. Such a hop is conveniently labelled a "voluntary hop". Subjects who were at all apprehensive about the experimental situation invariably produced such hops also when they were overbalanced artificially. In contrast, experienced subjects who had got over their initial trepidation often produced hops with no preliminary dip in the force trace. Such hops may be labelled "reflex", on the grounds that they appear to be "automatic" and differ from hops produced on command.

13. It was observed that the threshold angle of tilt at which the voluntary hops were initiated was less than the threshold tilt for the reflex hops. The voluntary hops can therefore be regarded as examples of a new class of action that may be termed "anticipatory pre-emptive actions". This expression indicates that the actions are initiated by a recognition of a developing trend such that a reflex response is about to be initiated. The actions are "pre-emptive" because they have the effect that the conditions are not allowed to develop to the point of threshold for the reflex response itself.

14. Anticipatory pre-emptive actions have the status of habits. They are learned behaviours that are so well rehearsed that they can be invoked without the subject being aware of the occurrence of the triggering gestalt.

15. It turns out that there are a great many instances in ordinary living where anticipatory pre-emptive actions intervene to provide very rapid smooth corrections to cope with the varying environment. They have the advantage of avoiding the delays and hunting oscillations associated with conventional servomechanisms.

16. The anatomy of the semi-circular canals in the inner ear indicates that they function as detectors of angular acceleration. On the other hand, clinical experience suggests a function as detectors of angular velocity, that is, of the speed at which the head is being turned. This confusion of opinion is resolved by computer simulation which shows that the effect of the damping inherent in the structure of the canals is to convert the response to the short-lasting angular accelerations of normal head movements into a signal whose time-course closely resembles that of the changes in angular velocity. The simulation also explains the various bizarre sensations of movement experienced in special conditions such as those of the Barany chair and those in aircraft performing banked turns, where Coriolis effects are dangerously significant.

17. The otolith organs, on the other hand, are shown to act as differential-density accelerometers, responding to linear stress-gradients rather than to gravitational forces as has hitherto been generally believed. The responses to tilting are, however, somewhat complex, different sensory units showing different patterns of signal. The labyrinth signals, taken together, produce reflex effects on the weight-bearing functions of the limbs, with different response patterns according as the stimuli are transient or sustained. The nature of the responses is predominantly stabilising, in conditions that may be classified as "normal".

18. While a tilt of the head would produce, by labyrinth reflexes alone, an extension of the limbs on the downhill side, to stabilise the head, a corresponding movement of the head on the neck would produce, by neck reflexes, changes in the limbs in such directions as to tend to straighten the neck. Thus, if the trunk remains stationary, a tilt of the head should produce opposing effects on the limbs from the labyrinth and from the neck respectively. In practice, the conflict is resolved centrally, and it is shown, by the use of a model, that the effect of the combination of the two sets of reflexes is to stabilise the trunk, although the relevant directional sensor lies in the skull, not in the trunk.

19. On a moveable platform, the effect of the stabilising reflexes is that the trunk tends to remain in the same place when the platform is moved about within a certain range, even though the feet are necessarily carried along with the platform. Such a system can only work for small displacements. More extensive movements of the platform involve a risk of overbalancing. This is corrected for by a rearrangement of the attitude of the body, with stepping movements of the feet if necessary.

20. The problems of maintaining balance when the supports are moving, as in a moving vehicle, make it clear that, in spite of what is generally believed, the direction of the gravitational vertical is irrelevant. The reactions of the body are, in contrast, organised around a direction referred to as the "behavioural vertical", which may be thought of as the "best direction in which to push against the supports to avoid falling over". This direction is very much dependent on circumstances, and changes when the platform undergoes linear acceleration. The choice of what is currently to count as the direction of the behavioural vertical is a matter of gestalt recognition, as the trigger for initiating "anticipatory pre-emptive actions" when there is imminent risk of overbalancing. Locomotion is achieved by deliberately resetting the behavioural vertical to initiate forced toppling, which then calls in a succession of rescue reactions which carry the body forward along the chosen path.

21. A number of separate types of reaction can be distinguished as contributing to the rearrangement of the attitude of the body to take account of the conditions. In general, these appear to be ways of adjusting the sensitivity of the stretch reflex. In this reflex, when a load is applied to a muscle, so as to tend to stretch the detectors lying within it, lengthening of the muscle is resisted. This is achieved by the recruitment into activity of an increased number of motor units as well as by a small increase in the frequency at which the motor units are repetitively activated. The stretch reflex is to be distinguished from the better-known jerk reflex in that the stimulus is different, the neural mechanism is different, and the response is different. Jerk reflexes depend on synchronised volleys of impulses, both on the sensory side and on the motor side, the synchrony being essential for the unusual synaptic effects on which the reflex response depends. In the stretch reflexes, and indeed in most of the activities of the nervous system, synchronised volleys of impulses are very seldom encountered except in special experimental conditions such as those involving electrical stimulations.

22. Modifications of stretch reflex activity are responsible for such reactions as the supporting reaction that converts a limb into a relatively rigid pillar when the foot comes to the ground, and for the various sway reactions that deal with lateral perturbations. They are also involved in the withdrawal reflexes and the changes that occur in other limbs to provide the necessary redistribution of the contributions to weight-bearing. Certain types of perturbation lead to stepping, to reposition the feet. A succession of such reactions produces locomotion. This can be initiated by a deliberate adjustment of the direction of the behavioural vertical, since this necessarily invokes reactions that would produce overbalancing if not compensated for by the rescue reactions of stepping.

23. The pattern of limb movements in locomotion can be altered to achieve different speeds over the ground. A number of different gaits can be distinguished. No less than sixty gait variants have been identified in the horse on the basis of the relative timing of the placings and liftings of the feet. In addition to the choice of gait pattern, it is necessary for the aiming of each foot to be adjusted to ensure that it lands on a potentially safe support rather than on an obstacle or other hazard.

24. Special types of locomotor patterns are required in certain conditions, like those involving the wearing of specialised footwear such as skates or skis, and in the riding of a bicycle. Clearly, such new locomotor patterns must be learned, rather than innate. In considering the learning process, a distinction should be made between "conditioning" and "true learning". In classical conditioning, an alteration is induced in the gestalt forming the trigger for a specific reflex response. The response itself is not changed. In true learning, on the other hand, it is the motor performance that is altered, and this may take a quite new form after the relevance of certain features of the environmental situation has been recognised by the subject. The process of learning is reinforced when the resulting new behaviour triggers some form of reward-recognition process.

25. Traditional accounts suggest that, in the performance of its various coordinating functions, the brain receives information from the sense organs to indicate what is going on in more peripheral parts of the body. A detailed study of the neural signals generated by proprioceptive sensors of various kinds, including the muscle-spindles, semi-circular canals, and otolith organs, reveals that this notion is unrealistic. Many factors contribute to influencing the pattern of the impulse streams in the afferent nerves. For each mechanoreceptor, deformation, its rate-of-change, and the recent history of changes, all make their contributions to the firing-frequency of impulses in the sensory axon. The situation is even further complicated by the presence of an efferent innervation where outgoing messages from the nervous system can have profound effects on the sensitivity of the receptor mechanisms themselves. While it is possible, in a few cases, to model and simulate the transduction process, it is very hard to conceive any mechanism by which the primary "physicist's variables", such as relative displacement, force, and the rates-of-change of these, could ever be recovered from the time-course of the impulse stream.

26. Whatever the neural coding in the sensory axons, even more serious complications are encountered when the messages enter the central nervous system. The incoming axons branch extensively, and each branch makes multiple functional connections (synapses) with a number of dendrites in a complex feltwork of neuropile made up of the interlacing elaborate dendritic trees of large numbers of neurons, including interneurons of several different types as well as motoneurons. At each junctional site, the synaptic effects of the incoming nerve impulses depend on chemical and pharmacological influences on the permeability of the subsynaptic membrane of the postsynaptic cell. The changes in permeability produce electrical changes which, in turn, influence other parts of the cell. Thus each central nervous neuron, of whatever type, effectively acts as a kind of sense organ, responding to the many influences of events occurring all over its surface, including the cell body as well as the whole of its dendritic expansions, and initiating new impulses in its own axon at a repetition frequency which, again, is dependent not only on local changes, but also on their rates-of-change and on the history of recent changes.

27. The axons of the interneurons pass to other regions of the brain, both nearby and far away, branching there and each influencing a number of other interneurons. Synaptic effects are produced by the liberation of chemical transmitters, of which several kinds have been identified. Some of these have only local effects, in the close vicinity of the active nerve terminal. Others, distinguished as "neuromodulators", have more widespread and longer lasting influences, some of which involve changes in the microanatomy of the synaptic junctions themselves. In this way the pattern of connectivity within the brain may become modified as a result of a history of particular patterns of afferent input. This process would form a basis for learning, and it may be argued that some form of learning is required for the nervous system to be able to interpret the incoming impulse streams.

28. The establishment of a system of interpretation to suit a particular type of environment may be assisted by inherited patterns in the anatomical layout of the central connections, helpful changes accumulating through evolution. When the environmental circumstances are unusual, sensory misunderstandings may occur. Some of these are seen in vehicles moving in a curved path, with more marked effects being produced at higher speeds, as in modern fighter aircraft, where certain consequences can be disastrous if not anticipated.

29. There is a good deal of recent information about the microanatomy of central connections, particularly in the visual and oculomotor systems. (Recall, in this context, that the detection of image streaming in the peripheral retina provides important clues about movements of the head in space.) Some of the anatomical information comes from the effects of localised lesions occurring in disease or injury. It has emerged, for example, that some of the reactions to tilting, normally generated from the labyrinth, must be mediated through the basal ganglia, since they are absent in certain cases of basal ganglion disease. Lesions in specific parts of the basal ganglia are found to affect the process of deciding on and manipulating the behavioural vertical. Some patients are unable to overbalance deliberately, as is required for the initiation of locomotion. Others behave as though overbalancing when they are not, being forced by the consequent rescue reactions into headlong rushes of festination.

30. Some patients, crippled for many years by basal ganglion disease, have shown dramatic temporary relief of symptoms after administration of L-Dopa. This substance has also been shown to release, in certain reduced preparations, motor rhythms to some extent resembling those of normal locomotion. It is this observation that has prompted speculation about the possibility that normal locomotion may be driven by some sort of central rhythm generator. This notion is made less plausible by the wide variety of gaits that have been identified and by the need for the aiming of each foot-placement to be separately adjusted.

31. In designing the motor commands needed to produce appropriate aiming of a limb, the nervous system must be able to recognise the relevant features of the environment that distinguish potentially safe supports from obstacles and other hazards. Indeed, recognition processes operating on the sensory information are involved in triggering even the simplest of reflex responses.

32. Until recently it has been difficult to imagine what sort of detailed neural activity could account for the performance of acts of recognition. Developments in computing with parallel distributed processors have now opened up new possibilities. Certain artificial networks have been shown to be capable of surprisingly complex tasks, such as assigning a sex to the subject of a portrait photograph presented to an array of photodetectors, or the control of a chemical plant to take account of fluctuations in the quality of its raw materials.

33. One example of a learning machine is Robinson's simulation of gaze stabilisation in the oculomotor system. A set of angular-velocity sensors corresponding to detectors in the three planes of the semicircular canals is linked to an output system controlling eye rotations about three axes, not oriented in the same way as the planes of the canals. The system could be trained to perform pursuit movements as well as saccades. Another simulator converts the angular-velocity information from the canals into position information for the eye-movement actuators. These devices were provided with a set of "correct" model responses to aid the convergence of the training algorithm.

34. Simulated learning can also be achieved without any inbuilt "teacher" that knows what response is to be regarded as correct. What is required here is some sort of reward function to simulate the expermenter's reinforcement in operant conditioning. Edelman's large computer model for simulating visual processes employs weight modifications based on the observed neural synaptic changes that are dependent on frequency of path use, together with an inbuilt "reward function".

35. The model receives input from a colour camera. Random values are initially assigned to the synaptic efficiencies of the various connections and these weights are automatically modified during the course of exposure to different inputs. This model spontaneously develops a mosaic of unit groupings resembling the columnar organisation found in the primary visual cortex, together with a variety of complex-feature extractors corresponding to those found in other areas. The system successfully learns to discriminate a moving pattern from its background, using various cues to establish coherence between the parts of the pattern, and performs many of the other functions of the mammalian visual system, including the binding of colour cues with motion cues in the discrimination of objects. The model can be provided with an output corresponding to the control of eye movement to bring an object of interest to the centre of attention, and a "saliency system" with diffuse projection that serves as the reward function needed for operant conditioning. Such a system can exhibit learned discriminatory behaviour.

36. It is claimed that such models are also capable of processes comparable to concept-formation and to many other stages in what has, hitherto, been regarded as the domain of mental events, including the detection of illusory boundaries in certain optical illusions. It seems reasonable to suppose that feature extractors, similar to those found in the visual system but operating on signals from proprioceptors, discriminate overall limb positions, together with the magnitudes of the forces exerted against the supports. In addition, we may postulate higher order systems that detect the "imminence of overbalancing", which may be regarded as a concept that includes both a direction and a rate-of-change. Other feature extractors record what patterns of motor activity produce limb thrusts of specific direction and force. A memory system stores the effectiveness of each thrust pattern in reducing the urgency of the imminence of overbalancing. Operant conditioning then builds up, from this type of experience, a repertoire of thrust patterns appropriate to each of a variety of conditions of imminence of overbalancing. What is now required is a mechanism operating, like a feature extractor in reverse, to formulate patterns of activation of motor units in order to achieve a specific direction and force of thrust. Mechanisms of this sort must be available for any form of voluntary movement to be possible.

37. It may be noted that the direction of the behavioural vertical itself is not something that can be directly sensed. It has the status of a concept of the limiting condition in which the imminence of overbalancing is reduced to zero. It is built up from experience of the effectiveness of various thrusts that have been exerted in the past. Its direction, in relation to the trunk, is continually changing, for reasons that are explained. For stability, we aim to reduce the risk of overbalancing, while to initiate locomotion, we may develop a thrust that deliberately leads to overbalancing, so that a horizontal acceleration in a desired direction can be obtained from the interaction with gravity.

38. From what we now know, from computer simulations, of how interconnected groups of neurons can be expected to behave, we appear to be approaching a position in which we can account for the way the central nervous system organizes the strategic choice of sequences of offset thrusts by which we maintain our balance and perform acts of locomotion.


Boyd, I.A. (1980) The isolated mammalian muscle spindle. Trends. Neurosci., 3: 258-65.

Boyd, I.A. and Roberts, T.D.M. (1953) Proprioceptive discharges from stretch-receptors in the knee-joint of the cat. J. Physiol., 122: 38-58.

Lindsay, K.W., Roberts, T.D.M., and Rosenberg, J.R. (1976) Asymmetric tonic labyrinth reflexes and their interaction with neck reflexes in the decerebrate cat. J. Physiol., 261: 585-601.

Lowenstein, O. and Roberts, T.D.M. (1949) The equilibrium function of the otolith organs of the Thornback Ray (Raja clavata). J. Physiol., 110: 392-415.

Purdon Martin, J. (1967) The Basal Ganglia and Posture. London: Pitman.

Roberts, T.D.M. (1952) Reflex interaction of synergic extensor muscles of the cat hind limb. J. Physiol., 117: 5P.

Roberts, T.D.M. (1963) Rhythmic excitation of a stretch reflex, revealing (a) hysteresis and (b) a difference between the responses to pulling and to stretching. Quart. J. Exp. Physiol., 48: 328-345.

Roberts, T.D.M. (1968) Labyrinthine control of the postural muscles. In: Third Symposium on the Role of the Vestibular Organs in Space Exploration, pp 149-168. Washington: NASA SP-152.

Roberts, T.D.M. (1971) Standing with a bent knee. Nature, 230: 499-501.

Roberts, T.D.M. (1973) Reflex balance. Nature, 244: 156-158.

Roberts, T.D.M. (1975) The behavioural vertical. In: Mechanisms of Spatial Perception and Orientation as Related to Gravity, pp 192-198. Ed. Schone, H. Fortsch. Zool., 23.

Roberts, T.D.M. (1976) The role of vestibular and neck receptors in locomoton. In: Neural Control of Locomotion, pp 539-560. Eds. Herman, R.M. et al. New York: Plenum Press.

Roberts, T.D.M. (1977) The relationship of gravity to vestibuar studies. In: Life Sciences Research in Space, pp 139-145. ESA SP-130.

Roberts, T.D.M. (1978) Neurophysiology of Postural Mechanisms. 2nd. Ed. London: Butterworth.

Roberts, T.D.M. and Stenhouse, G. (1976) The nature of postural sway. Agressologie 17A: 11-14.

Roberts, T.D.M. and Stenhouse, G. (1979) Reactions to overbalancing. Prog. Brain. Res. 50: 397-404

Roberts, T.D.M. (1982) Problems in the understandng of balance and locomotion in man. In: Proprioception, Posture and Emotion, pp 142-152. Ed. Garlick, D.

Roberts, T.D.M. (1992) Equestrian Technique. London: J.A. Allen.

Volume: 7 (next, prev) Issue: 02 (next, prev) Article: 1 (next prev first) Alternate versions: ASCII Summary