Controlling stability of a complex movement system.

Controlling Stability of a Complex Movement System

Characteristics of Complex

Movement Systems

If there is one task shared among all sensorimotor systems, it is to transform sensory information about the environment into motor impulses that initiate appropriate responses. When the sytem requires only a single sensory stimulus, has a limited number of synaptic junctions, and relies on a limited musculoskeletal arrangement to produce its motor behavior (eg, the monosynaptic stretch reflex pathway), assumptions about direct relations between input parameters and response characteristics can be tested. Increasing complexity in the system, either through multiple sensory triggers or multiple means of goal attainment, exponentially increases the difficulty in identifying the mechanisms underlying control of the musculoskeletal system.

Identification of Multiple Muscle

Movement Systems

A musculoskeletal system that has the potential to move in multiple directions and that has more muscles surrounding its joints than are necessary for producing the functional range of movements, is considered an "overcomplete" movement system. Although these systems are often described in terms of the agonist action or as a balance between agonist and antagonist forces, there are actually multiple combinations of the group of muscles acting on the joint that could produce the same directional force output. Examples of such systems include the shoulder, the elbow joint, and the head-neck motor system. One example, the head, has 23 different muscles that directly link the skull on either side of midline to the vertebral skeleton. [1] The multiple muscle attachments might not be so surprising if the head were involved in the fine motor control and variety of motions found in the hand and fingers. Motions of the head relative to the trunk, however, are primarily directed toward orienting and stabilizing the position of the eyes and head in space, [2,3] even during fine motor activities such as eating or scanning the environment. A recent fluoroscopic study of free head movements in several animals, including monkeys, cats, and rabbits, suggested that only two of the joints in the cervical column are actually used for lowering or raising the head. [4] A consistent occurrence of head flexion and extension at the atlantooccipital joint or the cervicothoracic junction further limits the effective degrees of freedom of motion, making the quantity of muscles available for controlling the head seem even more extraneous.

Having more muscles than are necessary to control the musculosketal system potentially provides multiple solutions to a single motor task. In other words, the same action of extending the head could be accomplished with a variety of different muscle patterns. "Equifinality," a single behavior resulting from multiple muscle combinations, indicates that a functional movement pattern can be under the control of different motor programs and possible utilize different control mechanisms, yet still attain the same final behavior. [5] This means that a multisegmental, multimuscle system, like the head and neck, can potentially switch its control operations between proprioceptive reflexes, vestibulocollic reflexes, or mechanical resonant properties and still achieve the appropriate coordinated response. Thus, an overcomplete, complex movement system could rely on different sensory signals or different arrangements of coordinated muscle activation patterns to produce a single functional behavior. In this article, I will concentrate on the identification and analysis of two complex movement systems--the head and neck system and the whole body system--in order to examine muscle activation patterns and kinematic parameters during functional stabilizing actions. Models that have been proposed to explain how complex movement systems operate will also be discussed.

Stabilizing the Head-Neck

Motor System

Reflex and Voluntary Neck

Muscle Activation Patterns

To examine whether parties of muscle activation are dependent on the movement behavior (defined as the direction and speed of head movemement) or the functional definition of the task (defined as reflex versus voluntary actions), the same head movement can be compared across two tasks. Head movements can be controlled by a consistent, compensatory response, the vestibulocollic reflex, that acts at short latencies (eg, approxmately 50 milliseconds) to maintain stability of that segment in space. The vestibulocollic reflex is elicited by labyrinthine inputs signaling a change in the position of the head. If the head's position in space is unexpectedly changed, contralateral neck muscles are activated to stabilize or return the head to its original position. When neck muscle electromyographic (EMG) responses during the vestibulocollic reflex were compared with those of the voluntarily activated muscle producing exactly the same motions, head movements generated in a particular direction by the voluntary motor system used different muscle patterns that when the same head movements were generated by the reflex. [6,7]

Patterns of neck muscle activation by the vestibulocollic reflex in three alert cats were compared with patterns used when the same animals voluntarily made the same head movements. [6,7] To elicit the vestibulocollic reflex, the animals were rotated in the dark with their head fixed to the rotating device so that there was no visual or neck proprioceptive feedback. The animals were trained to make voluntary movements by tracking a water spout that moved in the same 24 planes of motion that the whole body was rotated in for the reflex task. Because muscle activity during the vestibulocollic reflex serves to compensate for or oppose head movement, whereas activity during voluntary motion assists the movement of the head, maximum activation of the intramuscular EMG responses for matched head movements should be equal, but occur in opposite directions of head movement. For example, if the head and body were unexpectedly titled right, the vestibulocollic reflex in the left side neck muscles would bring the head toward an upright position. Voluntary motion of the head to the right, however, occurs through activation of the right side neck muscles. Although such equivalence was observed in the biventer cervicis muscle (primarily a head extensor muscle), other muscles did not behave in this fashion (Tab. 1). One explanation for these differences was the increased complexity of sensory inputs during voluntary movements. Unlike the vestibulocollic reflex, which was elicited by semicircle canal inputs, the voluntary responses could be organized by retinal, somatosensory, vestibular, and descending inputs.

Plots of the muscle activation patterns over a series of head movements in a single plane illustrated the muscle's response pattern. Response patterns for each muscle during voluntary head movements were consistent for individual cats over several months of testing, but the response patterns differed much more from animal to animal than did the reflex muscle patterns. For example, one cat used the splenius capitis muscle (SPL) for lateral flexion and the occipitoscapularis muscle for lateral rotation during voluntary tracking. In another cat, the preferred directions of these two muscles were reversed. This reversal was never observed in the reflex-activated EMG responses. These observations demonstrate the concept of equifinality. At the same speed of motion, an equivalent direction of head movement could be produced by the same animal with different patterns of muscle activation. The emergent pattern of muscle activation appeared to be more dependent on the available sensory inputs or the requirements of the task (action versus compensation) than on the mechanical advantage of an individual muscle. [8] Results also indicate that the voluntary motor patterns in this task were not composed of, nor supported by, the reflex-activated responses, but were distinct, learned responses to the relevant sensory inputs.

Muscle Activation Patterns

During Isometric Stabilization

These findings suggest that movement in a complex system is not a result of balancing the forces of an agonist and an antagonist pulling in a single plane of motion. The motor control picture emerging from this analysis coincides with current theories of muscle-activation patterns. In both the cat neck [8] and the human elbow joint, [9] EMG acivation of a muscle has been found to change in relation to the direction of motion or bending about the joint. The activation patterns of muscles around the joint will overlap, and synergistic relationships will depend on the desire motion, the muscle's preferred orientations, and the geometry of the joint.

Patterns of muscle activation during isometric head stabilization in humans have also been examined by Keshner et al. [10] The EMG activity of four right-sided neck muscles--the semispinalis capitis (SEMI), the SPL, the trapezius (TRAP), and the sternocleidomastoid (SCM)--was recorded with surface electrodes. Surface electrode placements were later verified with bipolar intramuscular eletrode recordings. Subjectx (N=15) were seated and received visual feedback of head positionat all times. They were instructed to counteract a moderate horizontal force applied to the head at 22-degree intervals around a 360-degree circumference via a weight-and-pulley system attached to a specially adapted helmet. At 0 degrees, the force was applied in a direction requiring neck extension to stabilize the head. A force at 90 degrees required rightward lateral flexion of the head,and a force at -90 degrees required left lateral flexion. Placing the pulley directly behind the subject resulted in pure forward flexion of the head (180[degrees]). Lateral rotation against resistance was achieved by placing a weight and pulley on either side of the helmet and having the subject resist the rotational torques. Average amplitudes of EMG responses for each muscle were calculated for each direction of applied force.

In Figure 1, the mean percentage of maximum EMG output for each muscle in each direction of head orientation is presented for the group as a whole in the study for Keshner et al. [10] Eight directions in the frontal plane (including flexion, extension, and right and left lateral flexion) and the two directions of lateral rotation are presented. The circumference of the circle represents maximum output of the muscle, and the gradually increasing and decreasing density of the shaded area indicates that each muscle had a restricted range of excitation, with its maximum output appearing consistently in a well-defined (or preferred) direction. Minimal or absent responses of each muscle in directions opposite to that for maximum excitation (see depiction of head orientations in Fig. 1) indicate that the muscles were reciprocally activated during this task. For all of the muscles tested, activation in lateral rotation far exceeded that for other directions. Head movements do not normally work against a force in this direction of motion, because gravity is not influential during lateral rotation in the upright position. Thus, the greater output of each muscle might be indicative of a response to the unusual (or nonfunctional) demands of this task. Yet, even in this plane of motion, the muscles demonstrated directional preferences.

A s shown in Figure 1, the SEMI functioned primarily in extension with some lateral rotation (0[degrees]-45[degrees]), and the SCM functioned in flexion with lateral rotation (90[degrees]-180[degrees]). [10] The SPL, However, gave an unexpected result. This muscle has always been described as producing a lateral rotation and extension action; yet, half of the subjects presented a strong rightward roll response with flexion (135[degrees]), rather than extension (45[degrees]). This finding was confirmed with intramuscular EMG recordings. Thus, each subject used a consistent pattern in opposing applied forces, but different subjects used the SPL in a different fashion. The TRAP presented low levels of activation and a large area of variation in all of the tested directions. Because the TRAP surface and intramuscular EMG responses were greatest when the subject was asked to perform isolated movements of the shoulder joint, the authors concluded that superior fibers of the TRAP participated in scapular depression in order to stabilize the scapula during head movements.

In three subjects, tjhe robustness of the relationship between direction and activation level was tested by gradually increasing the amount of force applied to the head. [10] A linear relationship between increased force and EMG response was observed in the preferred directions of activation for each muscle. In the other directions, however, each subject produced either a nonlinear increase or decrease in EMG activation. The main findings from this study are that all muscles are preferentially activated in defined directions of joint motion, yet these preferred patterns of activation can differ between subjects. The central program organizing EMG responses to the directional parameters of the task then is either preempted or combined with other programs when the force parameters of the task are altered. This finding would suggest that the central motor program depends both on previous experience with a task on the individual biomechanical constraints of each subject's head-neck motor system to organize and plan the muscle activation patterns that impose stability.

Kinematics of the Head-Neck

Motor System

Mechanisms controlling the response organization appear to be switched or gated by changes in the movement parameters or task constrainst. One method of analysis that has been used to assess the relative influence of the different control mechanisms on head movement is measurement in the frequency domain rather than in the time domain. Usually, time of stimulus onset is controlled, and response latencies are measured from this point. With this method, there is no specific onset and offset of the stimulus. Instead, frequency of a continuous sinusoidal stimulus is the controlled variable, and the relationship of head amplitude to perturbing-stimulus amplitude is calculated at specified frequencies (gain). The phase, or the delay between the stimulus direction and the head movement direction is also calculated at each frequency. If the amplitudes and phase relationships between the head and the stimulus are invariant across frequencies, then the motor output is linear. Models of the behavior of a cat's neck, when all central influences have been removed and only passive mechanics control position, predict that if head stability were controlled by inertial and viscoelastic mechanisms alone, the gani (relation of input to output) should rise because of mechanical resonance as frequency increases. This rise in gain should produce greater amplitudes and velocities of head movement than for trunk movements. [3] The phase should drop from 0 degrees to an 180-degree lag, indicating that the head and trunk are moving independently of each other. If neutral elements were responsible for head stability, then gain and phase would stay relatively constant (at 1[degree] and -180[degrees], respectively). The model and experimental data on which it is based are discussed in detail by Goldberg and Peterson. [3]

Guitton et al [11] hypothesized that at high frequencies (eg, 2-4 Hz), inertial forces increased and long-latency voluntary mechanisms presented too much of a phase lag to be effective stabilizers. Thus, biomechanical and reflex mechanisms would be expected to predominate at higher frequencies of rotation. This hypothesis was tested by rotating seated, healthy subjects (N=4) in the horizontal plane with a sum-of-sines stimulus at frequencies ranging from 0.18 to 4.12 Hz. [6] That is, five discrete sinusoidal stimuli were introduced randomly into a single period of chair rotation. As a result, the subject could not predict which frequency of rotation would be encountered at any moment within a 20-second stimulus train. The subjects were rotated about the vertical axis at a constant velocity of 80[degrees]/s. Head velocity, chair velocity, and two surface EMG recordings (SCM and SPL) were collected. Neck velocity was derived from the difference between the head and chair velocities. Velocities of the trunk were found to be equivalent to those of the chair, thus permitting the assumptions that the trunk was supported by the chair and that the subject need only actively stabilize the head and neck.

Each subject underwent rotations in three conditions of varying sensory inputs: (1) Voluntary stabilization--as the body is rotated, the subject actively stabilizes the head by matching a light worn on the head to a stationary spot of light projected on the wall. This condition supplies vestibular, visual, proprioceptive, and voluntary signals. (2) Imaginary stabilization--the subject attempts to stabilize the head in the dark, imagining the stationary spot of light on the wall. This condition eliminates visual signals. (3) Mental arithmetic--the subject is distracted by performing mental arithmetic during body rotations in the dark, thus making the vestibular and proprioceptive inputs most influential. [6]

As shown in Figure 2, the subjects demonstrated similar results in each of the three experimental conditions. [6] In voluntary and imaginary stabilization, when the subjects were actively attempting to stabilize their heads, response gains and phases were kept relatively constant and close to the values indicating good head stability at stimulus frequencies up to 1 Hz. At stimulus frequencies above 3 Hz, there was a rise in gain and a drop in phase as a result of mechanical resonant activity dominating the response. Between 1 and 3 Hz, there was a gain plateau that correlates with an increase in EMG activity. With mental arithmetic, the head was poorly stabilized during low frequencies of rotation in which the voluntary mechanisms operated most effectively, but the same gain plateau appeared in this condition as in the voluntary stabilization conditions. The flattening of the gain and its correlation with muscle EMG activity is suggestive of a period of head stability, possibly attributable to reflex control at frequencies between 1 and 3 Hz.

Therefore, voluntary, reflex, and mechanical mechanisms all contribute to final head position, but each mechanism appears to be dominant at a different frequency range. This finding does not mean that we can train patients to move within a frequency range over which their central nervous system (CNS) is capable of exerting some control. Functional movements, when combined with the nonlinear characteristics of the musculoskeletal system, have been found to simultaneously produce greater than one frequently of response at the head. [12,13] Three-dimensional power spectrum characteristics of the multisensory head-neck system have been measured during functional activities, such as walking in place, [14] and during normal locomotion on a walkway. [15] Frequencies of horizontal (0.6-1.0 Hz) and vertical (0.6-4.2 Hz) head movements when walking in place fell below those values that could be voluntarily developed by actively shaking the head (1.4-5.4 Hz) in the horizontal and vertical planes. During normal locomotion, as the subjects increased their speed of locomotion from slow to fast, frequencies of the head in the signal plane (flexion-extension) fell only within a range of 1.0 to 2.5 Hz. Horizontal plane rotations (lateral rotation) ranged from 0.6 to 1.0 Hz, and frontal plane rotations ranged from 1.5 to 2 Hz for the head. Frontal plane rotations (lateral flexion) had little power; however, they often presented flat power curves at all speeds of locomotion.

Resonant responses of the head (ie, those resulting mechanically from the force of the stimulus on the head with no control exerted by neural pathways) appeared at two frequencies during attempts at vertical head stability. Viviani and Berthoz [12] found that when human subjects resisted a force applied in the pitch (flexion-extension) plane, resonant activity occurred at 3 and 6 Hz. Similar resonant frequencies of the head (3 and 7 Hz) appeared in the vertical plane when human subjects experienced whole-body instability in response to dorsiflexion rotations of a posture platform. [13] As a result of their data, Viviani and Berthoz [12] developed a model treating the head and neck as a system with two degrees of freedom. They suggested that the two resonant frequencies were due to two centers of rotation for head movement in the sagittal plane, each having distinct elastic and viscous properties. Although their data fit their model at low frequencies, at high frequencies the fit was poor. A poor fit at higher frequencies could be due to nonlinear interactions between the two degrees of freedom or to a model that was too simple to explain the behavior of a multisensory system. [16] Thus, it is probably the sum of the forces produced by all of the central mechanisms acting at their dominant frequencies that is used to overcome mechnics and inertia in order to achieve a final, stable position.

Whole-Body Stability

Muscle Activation Patterns

and Kinematics

Some studies [17,18] of postural patterns of muscle activity have concentrated on muscles in the lower limb, from which most of the descriptions about restabilizing actions have been drawn. In these studies, subjects stood on a platform that could be translated in an anterior and a posterior direction or rotated so that the ankles are either plantar flexed or dorsiflexed. The expected response to anterior motion of the platform was that the subject would sway backward (base of support moved in front of the center of mass) and that the ankle muscles on the anterior surface of the body (ie, tibialis anterior muscles [TA]) would be stretched. If the platform moved posteriorly, the subject would sway forward (base of support moved behind the center of mass), and the gastrocnemius and soleus (SOL) muscles would receive stretch inputs. Although the monosynaptic stretch reflex did not act functionally to replace the center of mass over the base of support, EMG analysis of the lower limb muscles revealed that, in order to bring the body back over the base of support, the muscles being stretched still tended to respond first, but at latencies longer than the stretch reflex. The restabilizing ankle muscle responses (at latencies of 90-120 milliseconds) were followed within 10 to 20 milliseconds by the responses of the muscles in the upper leg on the same side of the body (ie, SOLs followed by the hamstring muscles; TAs followed by the quadriceps femoris muscles). Thus, from these early studies, patterns of muscle activation apparently initiated by ankle proprioceptive inputs and arising from the most distal to the most proximal lower limb muscles were identified as ascending muscle synergies that were responsible for restabilization after platform movement.

Keshner et al [19] examined the EMG onsets of 11 leg, trunk, and neck muscles in standing human subjects (N=10) in response to support-surface anterior and posterior translations and plantar-flexion and dorsi-flexion rotations on a hydraulic platform. The objective of the study was to test the hypothesis that the responses radiating upward from distal leg muscles would represent part of a large ascending muscle activation synergy encompassing axial muscles along the entire length of the body. Timing of postural muscle responses within and between body segments was analyzed in order to determine whether the muscles maintained a consistent temporal relationship under translational and rotational platform movement paradigms. The results did not support a strict ascending pattern of activation (Fig. 3). In response to posterior platform translations, an ascending pattern of muscle responses along the extensor surface of the body was observed. In addition, responses elicited in the neck flexor (NK FLX) and abdominal (ABD) muscles occurred as early as those of the stretched SOLs, suggesting a simultaneous descending pattern of activation. The crossed ascending-descending pattern was not as clear for anterior platform displacements, in which early NK FLX responses were observed at the same time as neck extensor (NK EXT) and ankle flexor (TA) activation. Temporal differences between muscle activation patterns to platform perturbations in the forward or backward directions were also revealed (Fig. 3). Platform rotations caused fewer responses in the neck and upper trunk muscles than forward or backward platform translations, and all muscle responses occurred simultaneously, rather than sequentially, during plantar-flexion platform rotations.

Kinematics of the lower limb muscles have been examined during platform dorsiflexion rotations. [13] Activation in the SOLs and the TAs were found to occur concurrently and with significantly correlated EMG response areas at both medium (120 milliseconds) and long (approximately 75 milliseconds later) latencies, thereby confirming a hypothesis of functional co-activation of these muscles. At short latencies, there was no correlation between the two muscles, as would be expected of responses having different onset latencies (ie, 50 milliseconds for the SOLs and 80 milliseconds for the TAs) and different reflex origins (proprioceptive versus vestibulospinal).

Interrelationships between a biomechanical measurement (ankle torque) and muscle EMG activity were examined through a multivariate regression analysis. [13] Actual torque output at the ankle was divided by the weight of the subject to remove variability attributable to subject size. A partial regression analysis of either ankle muscle on resultant ankle torque produced poor correlations (Pearson r=.1-.6). When the coactivated action of both ankle muscles was taken into account, however, significant correlations (r=.8-.88) were found between the two muscles and ankle torque at short, medium, and long latencies with eyes both open and closed.

The motor control picture emerging from this analysis suggests that at short latencies, the SOL monosynaptic reflex had the greater effect, but that the SOLs must act in conjunction with the TAs to prevent activation of the SOL monosynaptic reflex from further destabilizing the body. At medium latencies, the TAs exerted the greater influence on ankle torque when the activity of the SOLs was considered. The action of the TAs was to increase forward sway torque about the ankle joint to counteract the rearward thrust from the platform. Longer-latency torque was best predicted by the SOLs when controlling for activity in the TAs. In this case, the more influential SOL action produced a negative sloping regression line, suggesting a braking torque action ont eh forward sway produced earlier at medium latencies. [13]

We can conclude from these results that postural response patterns are not fixed patterns of activation, but differ as a result of variations in vestibular and neck proprioceptive inputs, as well as in diffrent mechanical demands presented by the posture platform paradigms. Patterns of EMG activity at individual joints demonstrate yet againt that stability is not simply contraction of an agonist muscle working to produce a single joint force, or even the stiffening of the joint by maximal output of agonist and antagonist, but a net torque comprising the coordinated action of several muscles at each joiny. Remember that total body posture is a function of the position of all of the different joints, and restabilizing actions are not caused only by changes at the support surface, but by changes within the body segments as well.

Effects of Movement Disorders 

on Postural Stability

Clinical studies are a means of exploring where the controls lie for certain motor functions. Postural strategies during rotational perturbations at the ankle tend to be replicable within a population, and the temporal organization and magnitude of EMG responses, as well as resultant torques, have been shown to be significant indicators of vertibulospinal dysfunction. [15] The area under the EMG record of the ankle muscles (SOLs and TAs) and ankle torque recordings of patients with bilateral labyrinthine deficit [13] were found to be significantly diminished when compared with those of healthy subjects with both eyes open and closed. Short-latency TA responses, which appeared at about 80 milliseconds in healthy subjects, were essentially absent in the patient population, indicating a vestibulospinal origin to this response. Amplitude of EMG recordings correlated with extent of peripheral vestibular deficits, [20] suggesting that lower limb postural reflexes could be triggered by proprioceptive stretch reflexes, but that their amplitude modulation is under the control of vestibulospinal signals. Electromyographic activity in the neck muscles was not obviously altered in these patients, suggesting local control by segmental stretch reflexes.

In another study, [21] latencies and areas under ankle and neck muscle EMG responses and torque exerted on the support surface during platform dorsiflexion perturbations with eyes open and closed were compared across three populations: healthy young adults (20-40 years of age), healthy elderly adults (50-80 years of age), and parkinsonian patients (50-80 years of age). A stepwise discriminant analysis clearly distinguished between the three groups on the basis of the variables shown in Table 2. The elderly adults produced a three-peak EMG response in the ankle muscles that could be identified as the short-, medium-, and long-latency responses seen in the healthy young adults. Mean latencies of each response, however, were shown by Bonferroni t test to occur significantly later in the elderly adults. The areas under the curves at short and long latencies were also significantly greater in the elderly adults, as were EMG response areas between the short- and medium-latency responses (SLb) in the TAs and the SOLs (SLb in Tab. 2 equivalent to 80- to 120-millisecond latencies). Medium-latency torque, a variable found to be predictive of stability, [13] was diminished, despite the larger EMG response. Impaired balance in the elderly is apparently produced by delayed vestibulospinal and propriospinal reflex responses that are poorly compensated for by enhanced response magnitudes. Balance problems in age-matched parkinsonian patients is a function of both age-and disease-related variables.

Examination of similar parameters in other populations should help to identify mechanisms that underlie diminished stability and has produced thef ollowing results. Patients with peripheral vestibular deficits tend to restabilize with greater motion at the ankle than at the hip and with the neck so stiff that little free head movement occurs. [22] Medium-latency responses at the ankle were absent in patients with spinal lesions. [23] A test of sway-stabilizing responses in patients with atrophy of the anterior lobe of the cerebellum revealed that response latencies were within normal limits following dorsiflexion rotations on a platform, but htey lacked balance between opposing muscle forces. [24]

Overall, results of these clinical studies suggest that the medium-latency response to postural instability is a spinal response modulated by supraspinal structures. These findings raise two questions that are pertinent to further study and analysis of postural stability. First, is a consistent spatial pattern in the EMG response an indication of a central program that does not compensate for temporal delays in sensory processing? Second, are the kinematics of the movement modified by changes in the patient's ability to process information as well as by the availability of sensory inputs into the central processing system?

Modeling a Complex

Movement System

Models can help us predict or hypothesize the relations between a signal and the mechanisms controlling complex movements. When we test a model, we begin to derive the range of functional variability for each behavioral task. Several types of modeling are currently being used in neuroscience. A popular model is the lumped-parameter model, which reduces the number of variables involved in the task by determining those that do not appear to be critical for successful performance. This type of model imlies many a priori assumptions about movement performance and can produce distorted results. For example, Bizzi et al [25] attempted to simplify the head-neck motor system by modeling the openloop mechanics of the monkey's head as a lumped second-order mechanical system. Components of this model included inertial characteristics and both passive (values not dependent on the level of neural input) and active (valves dependent on level of neural input) properties of viscosity and elasticity. Inertia was estimated from the dynamics of a severed, perfused monkey's head, and passive elasticity and viscosity determined by application of steps of horizontal torque to the head of anesthetized, rhizontomized, and vestibulotomized animals in order to reduce spontaneous and reflex inputs to the neck muscles. Active properties were measured by application of the same step torque perturbations to the awake animals.

Monkey head mechanics were found to have a natural frequency of about 2 Hz in the horizontal plane. [25] Through modeling the head's response to the increased muscle stiffness that would be expected in the presence of neural inputs, the authors concluded that the neck muscle stretch reflexes (including the cervicocollic reflex) were responsible for less than 10% to 30% of compensatory torques to an unexpected disturbance of the head. Mechanical properties of the neck muscles were considered to be responsible for the greater portion of force compensation. Limitations of the lumped-parameter model are illustrated by the fact that muscle nonlinearities found under normal conditions were ignored in this model, and the inability of the monkeys to fully compensate for unexpected loads was never explained by the model. Frequency characteristics of the torque disturbance were not considered and might explain the predominance of mechanical properties for head stabilization in this study, because studies of cats [3] and humans [6,11] have revealed that each of the mechanisms involved in head stabilization, including reflex, voluntary, and mechanical mechanisms, emerge at well-defined frequencies of head rotation.

More complex models of the head-neck motor system have been developed. One example, the sixth-order homeomorphic model [26] includes elements corresponding to the anatomical, physiological, biomechanical, and neural elements of the systems. First, all of the model components must be characterized (eg, muscles, ligaments, tendons), and then their interactions are studies through indenfication of a well-defined set of variables (eg, position, force, velocity). The components are combined into a set of equations stating that the rate of change of the variables depends on the current state of the system and on the external forces and neural inpust acting on the system. The state of the system is then calculated as a function of time. Additional complexity of this model is compensated for by including terms corresponding to known properties of muscles and ligaments and by combining redundant structures (eg, synergistic muscles) into equivalent structures in the model. The complexity of the physical system make these models more difficult to use in dealing with actual data, however, especially if the kinematics of the situation require that motion be modeled about more than one axis of rotation.

Studies of whiplash [27,28] have provided more complicated, three-dimensional kinemativ models, as have recent efforts using the tensorial model [8] for the vestibulocollic reflex that involves the use of matrices to describe the transformation of head rotations into motor commands. Models of the response to sudden impact have used the lumped-parameter approach, treating the cervical vertebrae and heas as rigid bodies interconnected by deformable elements. Discrepancies between the rigid model and physical data have been attributed to the significant effect of muscular contraction on the response of the neck. [27,28] The tensorial model operates on the assumption that a redundant system such as the head-neck motor system requires the use of some optimization creterion (eg, fatigue [29] or a weighted sum of muscle forces [30])to determine activation patterns. The tensorial model treats the muscles as simple linear force generators, equally excitable under all conditions. The predictions of this model have been compared with data from the vestibulocollic reflex in the decerebrate cat and shown to predict responses more closely than a simple pulling-direction model, which proposes that muscles are maximally excited in their direction of action. [8]

In posture, a rigid model of stability has been suggested. [17,18] Application of this rigid model to functional stability has been questioned, [13,20,31] mostly because the model of the body as an inverted pendulum relies only on EMG measurements from the ankle joint and is too simple to explain the behaviors of a multisegmental, multisensory systems. Many experimenters [17, 18, 23, 24, 32, 33] have relied on the foot's center of pressure to indicate body sway on the basis of rigid body mechanics. Validity of rigid body approximation has been challenged, however, because of observation of accelerations of center of gravity [34] and of individual body segments [35] during quiet standing. Even Nashner and colleagues have begun to question thier original hypothesis and to suggest that body posture is a function of the position of all of the different joints [36] and multiple sensory inputs. [37] Stockwell and colleagues [31] developed a four-link model composed of a foot, shank, thigh, torso, and head. They found significant movement at all measured body joints, indicating that at least four degrees of freedom would be required to adequately describe postural sway in the sagittal plane.

Implications for Clinical

Practice-Research

Normal movement is a confluence of forces produces at multiple joints and influencing the action of each other joint. Analysis of complex motor behaviors by assessing the available range of motion, sensory integrity, or strength available at each individual body segment does not assist us in determining how that patient will perform when presented with a multisegmental movement such as walking. In order to more functionally assess the ability of a patient to perform normal motor patterns, we must take into account and control as many of th e factors that can modify movement production as possible.

First, present environmental circumstances that require multiple adaptive responses as part of the therapeutic intervention. As the studies discussed previosly have demostrated, motor executio is not a direct result of fixed programs, but is flexibly modified for each repetition. Second, measure the patient's response to both novel and predictable events. The ability to predict and actively plan for the oncoming event should enhance the participation of voluntary mechanisms, whereas unexpected events will more strongly bias the system to automatic and reflex reactions. Third, control the sensory inputs. For example, when testing for postural control, it is necessary to be as concerned with the perturbation parameter (eg, direction, force, duration, site of perturbation) as with the observed response patterns. Pushing someone beyond his or her base of support or tilting someone when he or she is seated is not an equivalent task to the subtle, segmental changes required when stepping over an obstacle or reaching for a heavy book. Finally, remember that overlearning can inhibit adaptation. Do not make the patient so comfortable with a single response pattern that he or she will attempt to use that response evene when it interferes with the generation of a more appropriate response. It is tempting to believe that we can teach a specific pattern of movement that will resolve all of the postural problems presented to our patients. The key to success, however, is to develop adaptable and flexible strategies that will meet the multiple demands presented in a normal enviroment.

Mathematical models and robotic systems, although beyond the scope of most physical therapists, do offer an opportunity to quantify and measure variations in complex behaviors that are not easily detected in the clinic. These approaches to movement analysis may eventually present a useful tool to the therapist trying to determine the range of functional parameters that should be presented in a clinical situation. Quantified clinical data also are extremely valuable to the experimentalist trying to test hypotheses about how control mechanisms modify a response as a result of a disordered CNS or musculoskeletal system.

E Keshner, EdD, PT, is Assistant Professor, Department of Physical Therapy, University of Illinois at Chicago, and Adjunct Assistant Professor, Department of Physiology, Northwestern University, Chicago, IL 60612. Address all correspondence to Dr Keshner at Department of Physical Therapy (M/C 898), College of Associated Health Professions, University of Illinois at Chicago, 1919 W Taylor St, Chicago, IL 60612 (USA).

References

[1] Sherk HH, Parke WW. Normal adult anatomy. In The Cervical Spine Research Society: The Cervical Spine. New York, NY: JB Lippincott Co; 1983:8-22.

[2] Outerbridge JS, Melvill Jones G. Reflex vestibular control of head movements in man. Aetospace Med. 1971; 42:935-940.

[3] Goldberg J, Peterson BW. Reflex and mechanical contributions to head stabilization in alert cats. J Neurophysiol. 1986;56:857-875.

[4] Vidal PP, Graf W, Berthoz A. The orientation of the cervical vertebral column in unrestrained awake animals, I: resting position. Exp Brain Res. 1986;61:549-559.

[5] Abbs JH, Cole KJ. Neural mechanisms of motor equivalence and goal achievement. In: Wise SP, ed. Higher Brain Functions: Recent Explorations of the Brain's Emergent Properties. New York, NY: John Wiley & Sons Inc; 1985:15-43.

[6] Keshber EA, Peterson BW. Motor control strategies underlying head stabilization and voluntary head movements in humans and cats. In: Pompeiano O, Allum JHJ, eds. Vestibulospinal Control of Posture and Movement: Progress in Brain Research, Amsterdam, the Netherlands: Elsevier Science Publishers BV; 1988:329-339.

[7] Keshner EA, Baker J, Banovetz J, et al. Neck muscles demostrate preferential activation during voluntary and reflex head movements in the cat. Society for Neuroscience Abstracts. 1986;12:684. Abstract.

[8] Peterson BW, Pellionisz AJ, Baker JF, Keshner EA. Functional morphology and neural control of neck muscles in mammals. American Zoologist. 1989;29:129-149.

[9] Buchanan TS, Rovai GP, Rymer WZ. Strategies for muscle activation during isometric torque generation at the human elbow. J Neurophysiol. 1989; 62:1201-1212.

[10] Keshner EA, Campbell D, Katz R, Peterson BW. Neck muscle activation patterns in humans during isometric head stabilization. Exp Brain Res. 1989;75:335-364.

[11] Guitton D, Kearney RE, Wereley N, Peterson BW. Visual, vestibular and voluntary contributions to human head stabilization. Exp Brain Res. 1986;64-59-69.

[12] Viviani P, Berthoz A. Dynamics of the headneck system in response to small perturbations: analysis and modelling in the frequency domain. Biol Cybern. 1985;19:19-37.

[13] Keshner EA, Allum JHJ, Pfaltz CR. Postural coactivation and adaptation in the sway stabilizing responses of normals and patients with bilateral peripheral vestibular deficit. Exp Brain Res. 1987;69:66-72.

[14] Grossman GE, Leigh RJ, Abel LA, et al. Frequency and velocity of rotational head perturbations during locomotion. Exp Brain Res. 1988;70:470-476.

[15] Keshner EA, Peterson BW. Frequency and velocity characteristics of head, neck, and trunk during normal locomotion. Society for Neuroscience Abstracts. 1989;15:1200. Abstract.

[16] Schor RH, Kearney RE, Dieringer N. Reflex stabilization of the head. In: Peterson BW, Richmond FJ, eds. Control of Head Movement New York, NY: Oxford University Press Inc; 1988:141-166.

[17] Nashner LM. Adapting reflexes controlling human posture. Exp Brain Res. 1976;26:59-72.

[18] Nashner LM. fixed patterns of rapid postural responses among leg muscles during stance. Exp Brain Res. 1977:30:13-24.

[19h Keshner EA, Woollacott MH, Debu B. Neck and trunk muscle responses during postural perturbations in humans. Exp Brain Res. 1988;71-455-466.

[20] Allum JHJ, Keshner EA, Honegger F, Pfaltz CR. Organization of leg-trunk-head coordination in normals and patients with peripheral vestibular deficits. In: Pompeiano O, Allum JHJ, eds. Vestibulospinal Contol of Posture and Movement: Progress in Brain REsearch. Amsterdam, the Netherlands: Elsevier Science Publishers BV; 1988:277-290.

[21] Keshner EA, Allum JHJ, Honegger F. Discriminatory analyses of stabilizing reactions in young, elderly, and parkinsonian populations. Society for Neuroscience Abstracts. 1989;15:174. Abstract.

[22] Black FO, Shupert CL, Horak FB, Nashner LM. Abnormal postural control associated with peripheral vestibular disorders. In: Pompeiano O, Allum JHJ, eds. Vestibulospinal Control of Posture and Movement: Progress in Brain Research. Amsterdam, the Netherlands: Elsevier Science Publishers BV; 1988:263-275.

[23] Diener HC, Dichgans J. Long loop reflexes and posture. In: Bels W, Brandt T, eds. Disorders of Posture and Gait. Amsterdam, the Netherlands: Elsevier Science Publishers BV; 1986:41-52.

[24] Diener HC, Dichgans J, Bacher M, Guschlbauer B. Characteristic alterations of long-loop "reflexes" in patients with Friedreich's disease and late atrophy of the cerebellar anterior lobe. J Neurol Neurosurg Psychiatry. 1984;47:679-685.

[25] Bizzi E, Dev P, Morasso P, Polit A. Effect of load disturbances during centrally initiated movements. J Neurophysiol. 1978;41:542-556.

[26] Zangemeister WH, Stark L, Meienberg O, et al. Neural control of head rotation: electromyographic evidence. J Neurol Sci. 1982;55:1-14.

[27] Merrill T, Goldsmith W, Deng YC. Three-dimensional response of a lumped parameter head-neck model to impact and impulsive loading. J Biomech. 1984;17:81-95.

[28] Williams JL, Belytscho TB. A three-dimensional model of the human cervical spine for impact simulation. J Biomech Eng. 1983;105:321-331.

[29] Crowninshield R, Brand R. a physiologically based criterion of muscle force prediction in locomotion. J Biomech. 1981;14:793-801.

[30] Penrod D, Davy D, Singh D. An optimization approach to tendon force analysis. J Biomech. 1974;7:123-129.

[31] Stockwell Koozekanani SH, Barin K. A physical model of human postural dynamics. In: Cohen B, ed. Vestibular and oculomotor physiology. Ann NY Acad Sci. 1981;374:722-730.

[32] Diener HC, Bootz F, Dichgans J, Bruzek W. Variability of postural "reflexes" in humans. Exp Brain Res. 1983;52:423-428.

[33] Horak FB, Nashner LM. Central program of postural movements: adaptation to altered support-surface configurations. J Neurophysiol. 1986;55:1369-1381.

[34] Kodde L, Geursen JB, Venema EP, Massen CH. A critique of stabilograms. J Biomed Engl. 1979f1:123-124.

[35] Valk-Fai T. Analysis of the dynamical behaviour of the body whilst "standing still." J Ky Med Assoc. 1974;72:21-25.

[36] Nashner LM, McCollum G. The organization of human postural movements: a formal basis and experimental synthesis. Brain Behav Evol. 1985;8:135-172.

[37] Nashner LM, Shupert CL, Horak FB, Black FO. Organization of posture controls: an analysis of sensory and mechanical constraints. In: Allum JHJ, Hulliger M, eds. Afferent Control of Posture and Locomotion: Progress in Brain Research. Amsterdam, the Netherlands: Elsevier Science Publishers BV; 1989:411-418.