Walking is a complicated motor task that requires progression of the body toward the intended destination while also keeping the body upright. In humans, this task is made more challenging by the precarious placement of most of the body's mass high above a bipedal support structure. With each step, this precariously perched mass, which includes the head housing the brain, proceeds forward, transferring from one supporting foot to the next. In addition, the legs alternately swing forward to place the feet accordingly to receive and support the moving mass. Over the past century, there has been considerable effort put toward understanding the nervous system control of walking. From this effort, we have gained substantial understanding of the mechanisms involved in generating and regulating the rhythmic alternating pattern of flexion and extension that is required to propel the body forward, swing the legs to the next foot location, and regulate the transitions between these states. In contrast, our understanding of the neural control of balance while walking is far less understood. However, an inability to maintain balance limits the capacity for forward progression. For example, cats with a complete spinal cord transection can be trained to generate the rhythmic movements of the limbs important for forward progression, but are incapable of walking without assistance to maintain stability. Similarly, humans with spinal cord injuries show improvement in the locomotor-like pattern of muscle activity during treadmill walking after an extensive program of weight-supported treadmill training; however, this typically does not lead to improvements in independent functional locomotor capabilities, as stability and balance remain limiting factors.
Stability during locomotion is achieved through reactive and proactive strategies to control the motion of the center of mass and formation of the next base of support. As the name implies, reactive strategies are involved in correcting for unexpected disturbances to balance, whereas proactive strategies involve the predicted and anticipated needs of forthcoming challenges and prediction of expected disturbances to balance. Proactive strategies incorporate sensory cues with prior experience to establish and execute a plan of movement that will meet the needs of forthcoming movements. The selection and planning of the appropriate movements therefore involve many higher executive functions. Similar planning and selection of adaptive movements to the locomotor rhythm are required for proactive changes in stepping that occur when an obstacle is encountered or a change in direction is required. Our understanding of the neural control of these anticipatory or predictive adaptations to locomotion and balance is beginning to emerge, with the role of specific cortical and sensory inputs being identified. In contrast, reactive strategies must occur rapidly after an encounter with an unexpected disturbance; and therefore, higher executive functions cannot be directly involved in the release of the appropriate corrective strategy. In this sense, corrective balance reactions require a more "automated" control system, akin to the more automated control of the basic stepping rhythm that is seen in locomotion. What is striking about corrective reactions during locomotion is that despite the many options available, stereotypical responses are repeatedly evoked across subjects. The aim of this review is to put forward a model to explain how the nervous system is able to rapidly select and execute an appropriate correction to balance during locomotion. From this model, predictions can then be made as to the specific neural mechanisms that are involved in determining and executing these responses.
THE MANY OPTIONS FOR REGAINING STABILITY DURING WALKING
In the study of standing balance control, a number of corrective responses have been characterized (6). Perhaps the most well-documented responses are the ankle strategy, during which rotation about the ankle joint is used to control the motion of the center of mass, and the hip strategy, during which bending of the torso at the hips is used to quickly reposition the center of mass. These two strategies are only a small sampling of the repertoire of behaviors that can be evoked when balance is disturbed. Other strategies might include taking a step, grabbing a handrail, or raising the arms, to name a few. Similarly, during walking, several strategies have been identified including raising the arms, taking multiple stumbling steps, and grasping handrails. Any number or combination of responses might be adequate to return the person to a stable posture. So what then determines the behavior that is released to regain balance?
Obviously, the sensory input generated by the disturbance is going to play a large part in determining the behavior that is released. However, it is now well documented that the same disturbance and sensory input can generate different behaviors to regain balance. For example, when a person is pulled backward at the hips at the time of heel strike while walking on a treadmill, a robust activation of several leg muscles is evoked (9). However, we recently showed that if the person experiences the same perturbation while walking but holding handrails, the response in the leg muscles is virtually absent. This switch in the response to a diminished or even absent correction through the legs is consistent in all subjects. Remarkably, not a single subject studied continued to invoke the corrective responses with the leg muscles, although presumably this response would have been a reasonable choice. Therefore, the specific sensory conditions and mechanical state of the legs do not require that the same behavioral response is produced. Rather, the context within which the disturbance is experienced also plays an important role in determining the evoked behavior. This is perhaps best exemplified by the evocation of the ankle, hip, or change in support strategies during standing. With the instruction to stand with the feet in place, subjects will use an ankle strategy to overcome a small perturbation. However, if the subject experiences the same disturbance standing on a narrow surface, they will typically use a hip strategy. If the instruction to stand with the feet in place is not given, then the subject will typically take a corrective step. Thus, there is a strong context-dependency of the evoked behavioral response.
CONSTRAINTS IMPOSED ON CORRECTIVE REACTIONS
This preceding example also points to another factor that determines in part the evoked response, namely, imposed constraints. That is, when standing on a narrow surface, the production of torque about the ankle would not be useful in correcting for balance, as the forces would not be appropriately transmitted to the ground. Therefore, the ankle strategy is of no practical value and the hip strategy is used. Similarly, if the person were standing on a small elevated surface, then even without instruction they would not be able to execute a corrective step, as there would be nowhere to place the foot. These environmental constraints are obvious and speak to the context-dependent nature of the responses discussed above. However, other less obvious constraints will also dictate to some degree the necessity for the use of a different corrective strategy. For example, a medial-lateral displacement of the support surface will typically produce a lateral placement of the foot during a corrective step (7). This often involves a cross-over step of the leg unloaded by the perturbation. The leg that crosses over must avoid collision with the other leg, which is still in place on the ground. The complexity of this type of internal constraint on the corrective behavior becomes somewhat more challenging when a person is disturbed during walking, as now the corrective response must be incorporated within the ongoing movements associated with locomotion. Thus, the constraints change on a moment-by-moment basis. For example, the position of the swing leg with respect to the stance leg starts out as the trailing leg and then becomes the leading leg at midswing. Therefore, a cross-over corrective step of the swinging leg would need to avoid the stance leg if the disturbance is experienced during the first half of the swing phase, but not if the disturbance occurred in the later half of swing. Surprisingly, the nervous system is capable of selecting and executing the appropriate response in this ever-changing environment.
How the nervous system selects the appropriate response and then executes it within the relatively short timeframe that is available is not well understood. Corrective responses to perturbations during walking can be seen in the electromyogram as early as 70-100 ms after the onset of the disturbance. Responses of this latency are unlikely to be volitional and suggest that more automated fast-acting reflex responses are involved. These fast-acting responses may be mediated by spinal, brainstem, or long-loop pathways. Moreover, these responses are not limited to muscles acting at joints specifically disturbed by the perturbation. Rather, the responses tend to involve the whole body, with responses in the arm muscles occurring at comparable latencies to the responses seen in the leg muscles (9). Therefore, simple local reflexes cannot account for these early onset corrective responses. In addition, the functional corrective responses are not generic reactions to the presentation of the disturbance. Instead, they show distinctly phase-specific adaptations in both the pattern of muscle activation and the magnitude of the motor response elicited. Therefore, although the responses occur sooner than is expected from volitional control, the responses are complicated and suggestive of a high degree of control from the central nervous system.
FINITE STATE CONTROL AND DYNAMIC STABILITY
The specificity and phase-dependent nature of the selection of the appropriate corrective response during walking is reminiscent of finite state control systems. Finite state systems operate by using IF, AND, THEN arguments. Prochazka (13) argued that finite state control was a reasonable model by which to explain the control of the transitions in the gait cycle. As an example, the transition from stance to swing phase in the walking cat is argued to be dependent in part on the position of the hip, the load in the extensor muscles, and the state of the contralateral leg. The IF-THEN arguments that might be used to describe the transition from stance to swing might look something like this: IF extensor force is low AND hip is extended AND contralateral leg is loaded, THEN flex (13). The key concept in this model is that multiple sensory inputs are evaluated continuously to judge the state of any of the rules, and then multiple motor outputs can be generated based upon these sensory inputs. In this way, the appropriate motor response to any medley of sensory inputs is predetermined by the rules that exist for that motor behavior. The need for volitional decision making is eliminated. This type of finite state control is useful for determining events during the step cycle, such as the transition from stance to swing. Recently, a neuromechanical model of the walking cat was developed using similar rules to those above to describe this transition (Fig. 1) (2). This computer simulation was able to walk with a steady rhythm over a level surface and was also able to adapt to walking up and down slopes. In addition, the simulation was able to correct for a momentary loss of ground friction, mimicking a slip, within a few step cycles. The inherent properties of the finite state system were sufficient to make these adaptations and return the simulation to a stable gait pattern without the advent of volitional intervention.
Another important concept of the finite state control system is that the rules adapt based upon the expected or anticipated state of the system. Therefore, the relevance or weighting of any sensory input can change to meet the specific behavioral task being performed. In this way, the specific group of rules and motor outputs can be scaled, deleted, or added according to the "behavioral set." As a practical example, we recently showed that a person who is pulled backward at the waist while walking produced a whole-body corrective reaction involving responses in the muscles of the legs and arms (9). If the subjects are constrained from using their arms as part of the response by folding their arms across their chest, the magnitude of the response in the leg muscles is amplified (10), whereas if the subjects hold a stable set of handles, the response in the leg muscles is eliminated and the arms assume the primary role for correcting for the disturbance. By changing the availability of the arms to participate in the corrective reaction, the behavioral set was changed. In so doing, the sensory inputs and motor outputs were scaled to meet the task constraints.
The benefit of the finite state control model is not in identifying specific neural mechanisms, rather the strength is in describing the necessary rules for the expression of the known set of behaviors. Consequently, this will lead to the suggestion of the specific sensory inputs that are most important in certain motor tasks. These can then be specifically tested. For example, Gorassini et al. (3) observed that when a walking cat steps into a hole in the support surface, the leg is quickly withdrawn and placed in a more forward position, whereas the stance leg remains supporting the animal for a prolonged time. The stance leg delays transition to swing in part because the rule "IF leg is unloaded" was not met. This classic example of the importance of sensory states to the control of the step cycle also has important implications for dynamic stability during walking. When the cat's paw enters the hole, the step cycle of both legs needs to adapt. In so doing, dynamic stability becomes challenged, in particular the medial-lateral stability of the animal. While the one leg executes a complicated corrective placement of the paw, the other leg must provide additional stabilization, as the prolonged period of gravitational pull will tend to cause the animal to fall toward the side of the swing leg. Therefore, we can predict that muscles important for medial-lateral stability, such as hip abductors, must also be involved in the evoked corrective reaction. One prediction might also be that sensory signals regarding the state of the step cycle, that is, degree of hip extension and amount of load carried by the leg, would be sufficient cues to also mediate the corrective responses in the hip muscles. Alternatively, signals regarding the frontal plane motion of the hip (extent of abduction/adduction) may be additional crucial cues to ensure the appropriate response is evoked.
Identification of the required sensory inputs in turn would suggest certain sensorimotor pathways and mechanisms within the nervous system that are most relevant to the motor task, which can also be tested. Expressing sensorimotor transformations and motor behaviors as rule-based logical expressions has the potential to point to the neural mechanisms that mediate the behaviors. It is suggested that these types of conditional logic sequences in the control of locomotion are executed by collections of interneurons (13). Presumably, if we understand the rules of behavior and the sensory inputs that are important, we will then be better positioned to target and identify the specific groups of interneurons involved.
NEURAL MECHANISMS GOVERNING THE RULES
There is growing evidence of how the integration of reactive balance control and locomotor control is achieved by the nervous system. In general terms, there are broadly two categories of controls: 1) those that regulate the rules on a moment-to-moment basis, and 2) those that select the rule sets needed for the current task. The former category of controls will require continuous adaptation of the rules, with rapid decisions and quick execution of behavior, and thus suggests mechanisms within the proximity of the motoneurons are important, therefore, groups of spinal interneurons. The latter category will require incorporation of anticipated need and will thus need to call upon prior knowledge and experience, therefore involving supraspinal and cortical structures to select suitable rule sets and appropriately adjust the weightings of the rules.
Moment-to-Moment Regulation of the Rules
Locomotion is a rhythmic motor act with predictable, repeating patterns of muscle activation and predictable, repeating mechanical states. It has long been argued that the basic rhythmic activity of the muscles during locomotion is controlled by a collection of spinal interneurons, the locomotor central pattern generator (CPG). In general, a CPG is argued to be important as a timing clock for the rhythm as well as shaping the pattern of motor output, or muscle synergies (5). Consequently, a CPG is inherently well situated to act in a supervisory role for a rule-based control system. That is, the timing properties of a CPG would be well suited to enable the rules needed for any point in the walking cycle. Therefore, the activity of the CPG can be viewed as not only regulating the timing of muscle activation, but also the timing of rules activation.
Evidence suggests that this is indeed one of the roles of the CPG for locomotion. Most reflexes that have been studied in humans or other species demonstrate phase-dependent modulation in amplitude or sign during locomotion. For example, the reflexive activation of tibialis anterior by the electrical stimulation of the tibial nerve at the ankle during the swing phase of walking reverses sign to a suppression of activity during the swing to stance transition of the step cycle in humans (14). It is argued that the cutaneous afferent signals from the tibial nerve access the tibialis anterior motoneurons by two or more separate neural circuits and that the specific path taken switches depending upon the phase of the step cycle. The locomotor CPG is one plausible means by which the path of transmission might be regulated, in the form of IF clock is swing phase THEN path A, ELSE path B. With other reflexes, the amplitude of the response is modulated across the step cycle independent of the level of excitability in the motoneurons. Premotoneuronal modulation of reflex amplitude in this way has been suggested to be reflective of CPG activity (15). The basis for this argument arises from the findings in fictive locomotor animal preparations in which decerebrate cats are pharmaceutically paralyzed to eliminate the production of movements. In this type of preparation, activation of the basic locomotor rhythm (ergo CPG) leads to phase-dependent modulation of transmission in afferent pathways via phasic modulation in the level of primary afferent depolarization. Thus, there is direct evidence in this preparation for CPG regulation of afferent input, in essence adjusting the weighting of specific sensory inputs dependent upon the timing of the central rhythm. The phase-dependent modulation of cutaneous and muscle afferent reflexes is indirect evidence for similar mechanisms at play in humans. If CPG activity is capable of directly influencing afferent input, then it is reasonable to argue that CPG activity is capable of regulating the excitability of interneuronal reflex pathways and other pathways. Consequently, the oscillating activity of the locomotor CPG network is a logical means by which the rules for selecting the appropriate balance response may be regulated on a moment-to-moment basis.
Another neural mechanism aptly suited for governing the rules on a moment-to-moment basis is reafference generated by the ongoing movement. Afferent discharge is produced as a consequence of the ongoing movement. During walking, the specific afferents that are activated will be discharged rhythmically with the rhythmic movement. Brooke and his colleagues (1) showed that afferent feedback generated by the movement of the legs during locomotor activities was capable of regulating the strength of transmission in other afferent pathways. Presynaptic modulation of transmitter release was the suggested mechanism by which this type of sensori-sensory modulation occurred. This afferent-induced modulation of sensory feedback has the potential of adjusting the weighting of specific afferent inputs dependent upon the point in the step cycle, as identified by the medley of afferent signals generated by the movement. Therefore, sensory feedback should be viewed as having more than the basic role of signaling the state of any of the rules during the movement, but of also having the ability to regulate and select the rules.
Supraspinal Selection of the Rule Sets
Environmental constraints place restrictions upon the corrective reactions that are appropriate for regaining balance during walking. For example, if you are carrying bags of groceries when you slip on a patch of ice, the arms are less capable of assisting with regaining balance. We showed that when the arms are restricted from assisting in balance corrections, the responses in the legs are facilitated (Fig. 2) (10). This sort of task-dependent adaptation in the behavior produced reflects the change in task constraints. Consequently, the group of rules must also reflect this difference. The a priori selection of the rules based upon the constraints of the task and the environment most likely results from descending influences. Thus, with the arms restricted from assisting in the balance correction, the motor output in response to the disturbance is biased toward reactions that do not include the arms.
One of the important findings in balance control studies, including studies of walking balance control, is the robust learning effect of repeated exposure to a perturbation. For example, novel exposure to a slip leads to a much more vigorous response than subsequent slips (8). Prior knowledge of the risk of a slip also leads to a very different response than if the slip is unexpected. Therefore, knowledge of environmental context and prior experience influence the group of rules that is selected before experiencing the perturbation. Presumably, selecting a specific subset of rules in this way allows the response to be more efficient.
In addition to selecting appropriate groups of rules required for the task being performed, descending influences can also adjust specific aspects of the rules, such as the weighting of particular sensory inputs or spinal reflex circuits. In human locomotor studies, direct measures of the neural circuits involved in corrective balance control are not possible. However, spinal reflexes can be used as neural probes into the state of spinal circuitry. We have shown that when subjects walk with restricted use of their arms, thereby biasing balance corrections to those involving the leg muscles (see above), the amplitudes of spinal cutaneous reflexes are modified (4). The modifications we noted were not some general increase in amplitude of all reflexes or at all points in the step cycle. Rather, the modifications were isolated to very specific reflex circuits and at particular points in the step cycle. Pijnappels et al. (12) demonstrated that facilitation of cutaneous reflexes by the motor cortex can occur during locomotion in a phase-specific manner. Together, these results indicate that supraspinal influences do not simply change the state of the lower circuitry, such as by a generalized increase or decrease in excitability, but rather can very specifically modify the function of isolated circuits, in essence changing specific features of the IF-THEN rules.
Balancing the Multiple Demands
In Figure 3, the multiple factors that influence the selection and weighting of the rules are shown to converge. In this way, the potentially disparate influences are integrated to produce an optimum set of rules for the needs of all the task requirements. The sequence of events in the model depicted in Figure 3 is structured to suggest a temporal restriction consistent with the needs of a finite state control system. The upper half of the model relates to the preselection process, by which the rule set is selected and the rules weighted. The lower half of the model relates to the initiation and execution of a corrective balance reaction. In this model, execution refers specifically to the motor output generated to produce the mechanical correction. Sensory feedback is used in this model in both the preselection process, as well as its essential role in the evaluation of the rules (i.e., the IF of the rule-based expression) and thereby initiation of the corrective reaction. Sensory feedback is not limited to somatosensory modalities, but also includes visual and vestibular feedback (11). In a finite state control system, there will be multiple sensory cues, each associated with several rules in the group of rules selected. This means that the multiple rules and sensory cues will potentially compete for governance of the evoked response. Prochazka (13) suggested that each of the sensory cues would activate a "motor membership function," with the amplitude of the output from each of these interactions being scaled according to the task-dependent weighting of the cue, the net behavior produced being the sum of the scaled motor membership functions. Prochazka describes this decision-making process as a "vote" amongst the motor membership functions. In this model, extensor motor memberships and flexor motor memberships could both be activated, but the behavior produced would be dependent upon the scaling of the motor memberships. Prochazka's proposed model allows for coactivation of antagonist actions. Thus, if the scaling of the memberships for the motor behavior (task) is appropriate, then a coactivation of flexors and extensors could be produced. This fits with observations made in our studies in which backward pulls to the hips applied at heel strike result in coactivation of extensors and flexors throughout the leg (9). This decision-making process in the control of locomotion has been described as the "parliamentary principle"(13). It is suggested that the decision making, or conditional logic of finite state control systems, is performed by groups of spinal interneurons. Presumably, these groups of interneurons receive inputs from descending sources (including long-loop reflex pathways), the CPG, and sensory feedback before engaging the motoneurons. Lafreniere-Roula and McCrea (5) suggested a model of the spinal circuitry for locomotion with a layer of interneurons between the CPG and the motoneurons. Building upon this model, we might predict that this intervening layer of interneurons represents a key locus for the integration of balance corrective control and locomotor control (Fig. 4). The model presented here suggests that interneurons in the spinal cord represent a key location for the evaluation of the IF-THEN rules of the rule-based control system. Primarily, this argument is made based upon the strong evidence suggesting that much of the IF-THEN control associated with locomotion can be isolated within the circuitry of the spinal cord (5). This does not mean to imply that supraspinal pathways are not involved in triggering or executing the balance corrections that occur during walking. Rather, the implication is that any such long-loop responses must be integrated within the other demands of locomotion. Presumably, this occurs near or even within the motoneuron pools involved.
SUMMARY AND FUTURE DIRECTIONS
Viewing the control of balance during locomotion in terms of a rule-based finite control system has important implications for training in rehabilitation of balance disorders and development of performance athletes. As described above, the rule sets involved in corrective balance control are flexible and adapt to environmental and task constraints. This indicates that improving moving balance control will be best achieved if the balance training program is context-appropriate and diverse. For example, as noted above, the availability of the arms to assist in balance corrections changes the rules or weightings of rules in the corrective reactions to disturbances during walking (10). Therefore, to ensure that balance control is optimized for activities in which the arms will be engaged in other tasks (e.g., throwing, catching, or carrying objects), training of balance should then include restricting the use of the arms so as to "engage" the appropriate rule sets and weightings. Another way that viewing dynamic balance control in terms of a finite state control system might be used would be to purposely shift the balance or weighting of the responses generated. For example, with the arms restricted during walking, the corrective responses in the legs to a balance disturbance use a comparable pattern of muscle activity to when the arms are unrestricted. The difference is that the responses in the legs tend to be of greater amplitude when the arms are restricted (10). One prediction would be that if training with the arms restricted occurred over a long period of time, then the weightings of the rules would bias toward the arms-restricted scenario. That is, the responses of the leg muscles would become a more prominent response even if the arms are available to assist in the correction. Presumably, this would lead to improved functional balance control and improved performance of either activities of daily living or athletic pursuits.
Defining the motor control of dynamic stability in terms of a finite state control system provides a logical means to express the various behaviors that are produced to regain balance. Indeed, it also suggests that to fully understand the rules that govern the control of balance corrections, a more complete cataloging of the behavioral repertoire is required. However, the expression of the rules does not provide direct knowledge of the neural mechanisms involved. Nevertheless, defining the rules and the sensory cues that are crucial for evaluation of the rules will help lead to identification of the neural mechanisms. Traditionally, the functional relevance of a neural structure has been suggested by either ablating the structure and observing the subsequent deficit, or stimulating the structure and observing the evoked response. These types of studies have the advantage of being able to define specific anatomical connections and then inferring functional relevance based upon the observed deficit or behavior. However, there are now too many examples in the literature where ablation of a neural structure leads to only a temporary deficit in motor function, which then adapts over time to produce relatively normal motor functioning. The loss of anatomy does not necessarily mean the loss of behavior. This suggests that rule-based logical expressions are not necessarily confined to a limited group of anatomical structures. Rather, the plasticity of the nervous system allows for other sensory cues and perhaps interneuronal groups to substitute for the lost structures, presumably by adjusting the weighting of these other available cues or recalibrating the interneuronal pools. In this respect, defining the rules that govern control and the sensory cues important for executing that control has the potential to provide a more holistic description of the neural structures involved.
Supported by grants from the Natural Sciences and Engineering Research Council (Canada) and the Alberta Heritage Foundation for Medical Research.
1. Brooke, J.D., J. Cheng, D.F. Collins, W.E. McIlroy, J.E. Misiaszek, and W.R. Staines. Sensori-sensory afferent conditioning with leg movement: gain control in spinal reflex and ascending paths. Prog. Neurobiol.
2. Ekeberg, O., and K. Pearson. Computer simulation of stepping in the hind legs of the cat: an examination of mechanisms regulating the stance-to-swing transition. J. Neurophysiol.
3. Gorassini, M.A., A. Prochazka, G.W. Hiebert, and M.J. Gauthier. Corrective responses to loss of ground support during walking. I. Intact cats. J. Neurophysiol.
4. Haridas, C., E.P. Zehr, and J.E. Misiaszek. Postural uncertainty leads to dynamic control of cutaneous reflexes from the foot during human
walking. Brain Res.
5. Lafreniere-Roula, M., and D.A. McCrea. Deletions of rhythmic motoneuron activity during fictive locomotion and scratch provide clues to the organization of the mammalian central pattern generator
. J. Neurophysiol.
6. Maki, B.E., and W.E. McIlroy. The role of limb movements in maintaining upright stance: the "change-in-support" strategy. Phys. Ther.
7. Maki, B.E., M.A. Edmondstone, and W.E. McIlroy. Age-related differences in laterally directed compensatory stepping behavior. J. Gerontol. A Biol. Sci. Med. Sci.
8. Marigold, D.S., A.J. Bethune, and A.E. Patla. Role of the unperturbed limb and arms in the reactive recovery response to an unexpected slip during locomotion. J. Neurophysiol.
9. Misiaszek, J.E. Early activation of arm and leg muscles following pulls to the waist during walking. Exp. Brain Res.
10. Misiaszek, J.E., and E.M. Krauss. Restricting arm use enhances compensatory reactions of leg muscles during walking. Exp. Brain Res.
11. Patla, A.E. Strategies for dynamic stability during adaptive human
locomotion. IEEE Eng. Med. Biol. Mag.
12. Pijnappels, M., B.M. Van Wezel, G. Colombo, V. Dietz, and J. Duysens. Cortical facilitation of cutaneous reflexes in leg muscles during human gait
. Brain Res
. 787:149-153, 1998.
13. Prochazka, A. The fuzzy logic of visuomotor control. Can. J. Physiol. Pharmacol.
14. Yang, J.F., and R.B. Stein. Phase-dependent reflex reversal in human
leg muscles during walking. J. Neurophysiol.
15. Zehr, E.P. Neural control of rhythmic human
movement: the common core hypothesis. Exerc. Sport Sci. Rev.
Keywords:©2006 The American College of Sports Medicine
gait; human; balance control; perturbation; automatic postural response; central pattern generator; equilibrium