The innate capacity for the generation of rhythmic movement patterns is found throughout the animal kingdom. Breathing, chewing, and locomotion are examples of rhythmic motor patterns seen in vertebrate and invertebrate species. These motor patterns are produced in large part by the activity of collections of neurons whose membrane potentials oscillate rhythmically and are found in the spinal cord. These neurons are collectively termed central pattern generators (CPGs) (8). Humans produce rhythmic motor patterns during all forms of locomotor activity, whether the task is walking, running, cycling, crawling, creeping, or swimming. To appreciate fully the control of rhythmic human movement typified by locomotion, the traditional hierarchical motor control model of brain to spinal cord to muscle needs to be reappraised. The control properties can be best understood as the interrelationship of a tripartite system consisting of supraspinal input (e.g., brain), spinal (e.g., CPG and interneurons mediating reflex pathways), and sensory feedback (e.g., primarily from muscle and skin receptors activated during movement; see Fig. 1). In Figure 1, activation of rhythmic limb movement can be triggered by descending supraspinal commands related to the decision to begin walking, for example, which in turn delegate the motor coordination for rhythmic movement to the CPG networks controlling the arms and legs. In certain circumstances, peripheral feedback may be sufficiently strong to activate the CPG mechanisms themselves. As soon as locomotion is initiated, peripheral feedback from the moving limbs arrives at the spinal cord to inform the central nervous system of local conditions, and assists in sculpting the output of the CPGs. There is thus extensive and interrelated communication between all three levels (CPG, supraspinal input, and peripheral feedback), and the CPGs can be seen as elementary building blocks on which rhythmic movement is based. Somatosensory feedback contributes to the regulation of the basic pattern, which is monitored by supraspinal centers, and ensemble activity in the whole system produces smooth locomotor movement.
The aim of this review is to develop the hypothesis that all forms of rhythmic human movement share a similar neural control, which can be thought of as a common core composed of oscillatory neurons that drive the basic motor pattern (e.g., locomotor CPG). As described above, a CPG for locomotion consists of interneurons that generate the pattern or locomotor drive to motoneurons, thus yielding rhythmic movement. The interneurons that comprise the CPG can be considered as being those involved in the timing and oscillatory behavior (i.e., the “rhythm” in rhythmic movement), and those that are related to facilitating the activation of the specific muscles necessary to produce different behaviors (i.e., interneurons involved in mediating somatosensory feedback and reflex pathways). In the strictest sense, the CPG can generate rhythmic activity without any sensory feedback (e.g., see Fig. 2). However, to generate appropriate patterns of muscle activity across all joints and in all muscles, peripheral feedback is used for motor sculpting by interneurons involved in afferent pathways. In this article, we discuss the concept of a common oscillatory core that, along with pattern-sculpting interneurons, collectively comprises the CPG for different rhythmic human motor outputs. In the human (and other mammalian preparations), it is quite difficult to separate the pattern-sculpting interneurons that were part of the CPG from those that may be segregated from the CPG in separate reflex pathways. Indeed, it is possible that many of the interneurons involved in pattern formation also are those activated in discrete reflexes. In this review, CPG is meant to convey the oscillatory neurons as well as an unknown portion of the pattern-sculpting interneurons. Activity in the CPG common core sets the background timing of rhythmic alternating muscle activation for general rhythmic activity, whereas the actual patterns of muscle activation required for each specific locomotor task are mediated by interneuronal reflex pathways. These pathways in turn are sculpted appropriately to the environmental needs of each motor task by peripheral sensory feedback and supraspinal inputs. This common core hypothesis is a modification of general operational principles defined in many other species, and can be explored usefully as a theoretical backdrop for rehabilitation and exercise interventions.
BASIC OPERATIONAL PRINCIPLES OF RHYTHMIC MOVEMENT CONTROL
Many years ago, it was demonstrated that the basic circuitry for generating coordinated patterns of rhythmic movement could be found in the isolated mammalian spinal cord. Brown (2), in his experiments with the cat, provided compelling evidence that oscillating neural circuitry (half-centers) resided in the lumbar spinal cord. This half-center (one half for flexor activation, one half for extensor) model suggests that discrete rhythm or pattern generating (i.e., CPG) networks are responsible for producing the basic locomotor rhythm and muscle activity seen in locomotion (see Fig. 2). Shown in Figure 2A is a schematic illustration of the half-center model with the flexor half-center shown at top in black and the extensor half-center at bottom in gray. Note that the tonic input signal (shown by dotted lines at left) is transformed into a rhythmic output by the reciprocal inhibitory connections (shown as black interneurons) between the half-center interneurons. That is, activity in the flexor half-center inhibits activity in the extensor half-center, and vice versa. This mutually inhibitory system is essentially self-regulating. This rhythmic patterning then is shown activating the α motoneurons (at right in Fig. 2A) for flexor and extensor muscles. Interactions within and between flexor and extensor half-centers of a given limb underlie the fine coordination of muscle activity during locomotion (7). An example of the outcome of this general organizational principle applied to human locomotion is shown in Figure 2B. Here, muscle activity during human walking recorded by surface electromyography (EMG) is plotted for an ankle flexor (tibialis anterior, TA) and an ankle extensor muscle (medial gastrocnemius, MG). The far left begins at heel contact (e.g., start of stance phase) and proceeds across the step cycle to end swing. Note the reciprocal activation seen between the two antagonistic muscles. Clearly, this is a very simple model, and this basic circuitry taken in isolation cannot account for the wide variety of motor patterns requiring differentiated activity of muscles across multiple joints that can be seen every day. Specifically, the important role for peripheral somatosensory feedback associated with movement (e.g., see Fig. 1) and the organization of bifunctional muscles that have more than one period of peak activity during the locomotor cycle are not shown in this basic model. Basic models of this circuitry have been expanded to account for the locomotor repertoire that we see routinely used in behavior (see below).
There is evidence for considerable “sharing” of neurons and reorganization of synaptic activity to produce different motor patterns with the same neurons in invertebrate preparations such as the crayfish. Extensive reorganization of neuronal circuits involved in the generation of many rhythmic motor patterns (e.g., chewing, gastric mill, pylorus, etc.) has been well documented, and is altered extensively by various neuromodulators (6). That is, neuromodulators alter the activation and synaptic efficacy in various interneuronal pathways, and allow for the expression of different motor patterns with essentially the same neurons. Furthermore, membrane properties (e.g., leading to the expression of plateau potentials and membrane “bistability”; see (4) for review) also can be regulated by these same neuromodulators. The resulting motor patterns are thus a function of both alterations in network connectivity and intrinsic membrane properties. An outstanding question is: How would we recognize similar organization in humans where intracellular recordings are impractical and ethically impossible? Indirect approaches to address this are detailed in the next section.
Reflexes As Neural Probes to Reveal Mechanisms of Movement Control
In contrast to the situation in other animal models such as the crayfish, lamprey, and cat, where direct intracellular recordings can be taken, we must rely on indirect evidence and inference to assess the contributions of CPGs to human movements. It has been suggested that CPG activity also is involved in the regulation of rhythmic arm and leg movement (13), by generating the basic pattern of muscle activity and controlling how afferent feedback helps to sculpt the basic locomotor pattern. Although it is not possible to evaluate directly the exact contribution of CPG output to the rhythmic motor pattern in humans, we can estimate the probable contributions of CPG activity to the regulation of afferent feedback during rhythmic movement by using reflex studies. If shared circuitry for various rhythmic movements is shared also within the human spinal cord, it could be observed as characteristic modulation of reflexes during rhythmic movement. That is, we can approximate the input–output properties of neural control during movement by applying a given sensory input and recording the pattern of modulation of motor output. This approach has been used to great effect in the quadrupedal locomotor system (3), and also can be extended to humans (11,13). Reflexes arising because of activation of afferent projections from receptors in skin (i.e., evoked by stimulating nerves containing afferents innervating tactile mechanoreceptors to generate cutaneous reflexes) and muscle (i.e., evoked by stimulating nerves containing afferents innervating muscle spindles to generate the Hoffmann reflex, or H-reflex) have been studied widely during rhythmic human movement. From these analyses, extensive modulation according to task (e.g., between standing and walking) and within a portion or phase of a rhythmic activity (e.g., swing vs stance in walking) has been identified during human locomotion. It should be noted that the efficacy of afferent feedback itself can be modulated by afferent feedback. For example, it is now well known that group I feedback from muscle spindles can be modulated presynaptically (i.e., before the synaptic activation of motoneurons) by other feedback generated by movement (reviewed in (1)). This is particularly the case for muscle afferent feedback, but may have an influence as well on other methods. Furthermore, although the central focus of this review is on the central regulation of rhythmic human movement, a large role for somatosensory feedback (as indicated in all diagrams) is considered to be crucial.
As mentioned above, the model shown in Figure 2A has been shown to be inadequate to reflect accurately the strong role for fine modulation of muscle activation in bifunctional muscles and the strong role for somatosensory feedback in regulating rhythmic motor activity. Instead, models more consistent with these features have been proposed, and one such elaboration is shown in Figure 3 (inspired by (3)). In this schema, output from the CPG goes both directly to the motoneuronal pools as well as feeding through interneuronal reflex networks and then to the motoneuronal pools. Supraspinal inputs and somatosensory feedback act directly on interneurons of the CPGs, interneuronal reflex networks, and the pools of motoneurons. The outcome is rhythmic movement generated by CPGs and regulated by supraspinal input and somatosensory feedback.
Task and Phase Dependency of Reflex Amplitude
The presence of task dependency and phasic gain control of reflex amplitude that is independent of locomotor EMG level has been used to infer the activity of CPGs in humans. That is, CPG-driven modulation of afferent feedback via premotoneuronal gating yields phase- and task-dependent reflex modulation. This approach of examining the modulation of reflexes during rhythmic movement as an indirect indicator of CPG regulation of afferent input provides the main data on which this concept has been built in humans. Changes in reflex sign (e.g., cutaneous reflexes switching from excitation to inhibition) and amplitude (e.g., H-reflexes becoming larger or smaller in size) that occur between different motor tasks are a form of task-dependent reflex modulation. The main point is that rhythmic movements have distinctly different reflex patterns from static contractions, suggesting that they have different mechanisms of neural control. Some examples from studies that have examined such modulation are shown in Figure 4. The data in Figure 4A show the result when a cutaneous reflex was evoked in the plantar flexor medial gastrocnemius muscle during leg cycling and in static contraction (14). Here, the reflex switches from an excitatory one during static contraction (gray line) to inhibition during cycling (black line). In these kinds of experiments, nerve stimulation intensity, joint angle, level of muscle activity, and posture of the subjects are all controlled to be the same during static contraction and rhythmic movement. When modulation of reflexes occurs under these conditions, it means that something related to the performance of the rhythmic movement itself is the underlying neural control mechanism. The CPG has been the presumed underlying neural control mechanism. Task dependency of cutaneous reflex modulation also was studied during arm cycling (15). Figure 4B shows corresponding observations for cutaneous reflexes in a shoulder muscle, where reflex amplitude is different during static and cycling. In this example, the response also (as with the leg muscle response shown in Figure 4A) reverses in sign from excitation during static contraction (gray line) to inhibition during cycling (black line). Task-dependent differences also can be seen as dramatic reductions in H-reflex amplitude during rhythmic leg movement (reviewed in (1)) and arm movement (reviewed in (11)). The effect of rhythmic arm movement on H-reflex amplitude in forearm muscles was studied during arm cycling (12). As can be seen in Figure 4C, when comparing static contraction and rhythmic arm cycling, there is a dramatic attenuation of H-reflex amplitude during arm cycling (this is similar to what is observed in leg muscles during leg cycling; see (1) for review). In summary, then, it can be seen that there is considerable similarity and correspondence between reflex modulation in the arms and legs. This similarity has been ascribed to putative CPGs regulating rhythmic movement of the arms and legs (13).
It has been known for some time that reflex control seen in the legs during rhythmic movement has a strict dependency on the motor task that is performed (1,13). More recently, task dependency of reflexes in arm muscles also has been observed during walking. During static motor tasks, there was a direct relationship in arm muscles between background EMG amplitude and cutaneous reflex amplitude evoked by stimulation in both the foot and hand; this relationship was very weak during locomotion (reviewed in (11)). For both the arms and legs, the weight of observations suggests that reflex amplitude often is completely dissociated from the background locomotor EMG and instead is related to the phase of the movement cycle during which the reflex is evoked (i.e., phase-dependent modulation). The overall pattern of reflex control during rhythmic movement likely represents activity of CPG networks associated with generating the neural commands for rhythmic arm and leg movement. Reflex modulation during rhythmic leg movement generally has been accepted as being at least partly produced by CPG activity, and the similarity of reflex control during rhythmic arm and leg movements suggests that there may be CPG networks contributing as well to the control of rhythmic arm movement.
It has been proposed that the rhythmic EMG activity and patterns of reflex modulation observed in the arms during rhythmic movements are consistent with the existence of a separate CPG for the control of each arm (reviewed in (11)). A basic schema of the interrelation between all four limbs and CPG regulation of the limbs during human rhythmic movement is shown in Figure 5. The background and evidence to support this figure has been reviewed extensively elsewhere (5,13). Note the weak interaction (relative strength is indicated by the thickness of the lines) between the arms and stronger links between the legs. The strength of these linkages is implied by the effect that contralateral limb movement has on modulation of cutaneous and H-reflex amplitudes in the arms and legs. For example, active and passive movement of the contralateral leg strongly modulates H-reflexes in the legs (see (1)). Furthermore, there is an effect of contralateral leg movement on cutaneous reflex modulation during walking. In contrast, in the upper limb using an arm-cycling paradigm, contralateral arm movement had little effect on cutaneous or H-reflexes in forearm muscles (reviewed in (11)). There are both ipsilateral and contralateral linkages shown as well in Figure 5. In the center of the four CPG networks is illustrated another overarching set of connections. This presumed controller (with superimposed question mark denoting unknown locus) could represent supraspinal output or propriospinal connections that subserve interlimb connectivity to help coordinate the four limb oscillators. It also is possible that coordination between the four limb oscillators is an emergent property of the connections between oscillators, and requires little additional control. Presently, it is not a straightforward matter to determine this clearly in humans.
A COMMON CORE FOR CONTROL OF RHYTHMIC MOVEMENT
In the foregoing, we have established that both arm and leg movements share common elements of neural control. These are presumed to be largely subserved by CPG mechanisms. Furthermore, it was shown that rhythmic activities of the arms and legs together (such as walking and running) share commonalities with rhythmic activities of the arms and legs performed in isolation (such as arm or leg cycling). An additional portion of evidence to support common mechanisms of neural control for rhythmic movement comes from studying human infants during treadmill locomotion. In a series of studies (reviewed in (10)), Yang and colleagues perturbed the walking cycle of infants and observed that human infants can respond to perturbations, can make use of sensory feedback to modify the locomotor cycle, and can produce varying patterns of walking, including forward, backward, and sideways locomotion. These observations were made while the infants were still at an age where corticospinal connectivity is weak or absent, thus implicating spinal cord CPG mechanisms. It also has been demonstrated that there is similar neural control of backwards and forwards walking (13), leg cycling (1), and arm cycling (11) in adults. These results are consistent with a CPG running in reverse to regulate reflex amplitude and movement. Last, Dietz (5) also showed a very similar coordination pattern between all four limbs during walking, creeping, and swimming motions that are suggestive of similar CPG outputs in all activities. Thus, the premise that human rhythmic motor activities are generated at least in part by a common central core of CPG-related oscillators has a principled basis. The general operational principles for the control of rhythmic movement found in lower animal preparations extend as well to the human. The schema for this common core hypothesis is outlined in Figure 6. The general principle illustrated is that rhythmic motor timing is commonly specified regardless of motor task (e.g., walking, swimming, etc.). Interneurons influenced by sensory feedback assist in regulation of excitability and locomotor drive to the specific motoneuronal pools required for each task. Supraspinal input regulates activity in the oscillating circuits and sculpts the level of activity according to behavioral needs (e.g., walk faster or slower).
The concept of a common core generating the basic oscillation for rhythmic movement with premotor interneuronal networks specifying the appropriate patterns of motor unit recruitment can be thought of as mirroring principles at play in other physiological systems. For example, for the cardiovascular system, it is quite easy to envisage and conceptualize cardiovascular control as having central (so-called “heart and lungs”) and peripheral (local microvasculature in specific muscles) components. The central components are active regardless of the specific form of exercise (e.g., walking, cycling, swimming, etc.), but the control of the microvasculature in the muscles will be specific for the movement performed and will change from task to task. Although this analogy is instructive to understand the general idea, one cannot follow the analogy too far. In cardiac physiology, it is not possible for the right atrium to prime suddenly not the right ventricle but now the left ventricle. However, it is precisely this dynamic reorganization of connectivity and function that forms a cornerstone of neural control of rhythmic movement. Neurons of the common core CPGs and spinal interneurons mediating reflex pathways can reorganize their connections dynamically to allow for the expression of dramatically different rhythmic patterns (e.g., forward and backward locomotion). This concept has implications for not just the control of rhythmic movements but also for adaptations that may occur as a result of exercise training or rehabilitation.
SUMMARY AND FUTURE DIRECTIONS
For many years the concept of specificity of training has been used in both the exercise and rehabilitation literature. Within this concept, the more closely the training activity replicates the activity that is to be performed later, the better the training. This was most clearly articulated many years ago for strength training (9). Recognition of a common core for neural control of rhythmic movement thus also would suggest that as long as activities that engage this common core are performed, the transfer from one activity to another should be high. The main application for the common core hypothesis lies in rehabilitation of locomotion after stroke or spinal cord injury. In this domain, rehabilitation interventions could focus on general rhythmic movement (e.g., cycling) combined with specific movement training (e.g., assisted treadmill locomotion) for locomotor recovery. The common core hypothesis requires further study in both neurologically intact and neurotrauma participants to define the boundaries that may separate different forms of rhythmic movement. Also, in the neurotrauma participants most importantly, the potential to access these common mechanisms remains unproven.
Supported by grants from the Natural Sciences and Engineering Research Council of Canada (NSERC), the Heart and Stroke Foundation of Canada (BC & Yukon), the Christopher Reeve Paralysis Foundation, and the Michael Smith Foundation for Health Research. The author thanks Dr. John Brooke for helpful comments on an earlier version of this manuscript and the anonymous reviewers for their insight and suggestions.
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
. 51:393–421, 1997.
2. Brown, T.G. The intrinsic factors in the act of progression in the mammal. Proc. Royal Soc. London
3. Burke, R.E., A.M. Degtyarenko, and E.S. Simon. Patterns of locomotor drive to motoneurons and last-order interneurons: clues to the structure of the CPG
. J. Neurophysiol
. 86:447–462, 2001.
4. Calabrese, R.L., and J.L. Feldman. Intrinsic membrane properties and synaptic mechanisms in motor rhythm generators. In: Neurons, Networks, and Motor Behavior
, edited by P.S.G. Stein, S. Grillner, A.I. Selverston, and D.G. Stuart. London: MIT Press, 1997, p. 119–130.
5. Dietz, V. Do human bipeds use quadrupedal coordination? Trends Neurosci
. 25:462, 2002.
6. Hooper, S.L., and R.A. Dicaprio. Crustacean motor pattern generator networks. Neurosignals
7. Orlovsky, G.N., T.G. Deliagina, and S. Grillner. Neuronal Control of Locomotion: From Mollusc to Man
. Oxford: Oxford University Press, 1999.
8. Pearson, K.G. Common principles of motor control
in vertebrates and invertebrates. Ann. Rev. Neurosci
. 16:265–297, 1993.
9. Sale, D., and D. MacDougall. Specificity in strength training: a review for the coach and athlete. Can. J. Appl. Sport Sci
. 6:87–92, 1981.
10. Yang, J.F., T. Lam, M.Y. Pang, E. Lamont, K. Musselman, and E. Seinen. Infant stepping: a window to the behaviour of the human pattern generator for walking. Can. J. Physiol. Pharmacol
. 82:662–674, 2004.
11. Zehr, E.P., T.J. Carroll, R. Chua, D.F. Collins, A. Frigon, C. Haridas, S.R. Hundza, and A. Kido. Possible contributions of spinal CPG
activity to rhythmic human arm movement. Can. J. Physiol. Pharmacol
. 82:556–568, 2004.
12. Zehr, E.P., D.F. Collins, A. Frigon, and N. Hoogenboom. Neural control
of rhythmic human arm movement: phase dependence and task modulation of Hoffmann reflexes in forearm muscles. J. Neurophysiol
. 89:12–21, 2003.
13. Zehr, E.P., and J. Duysens. Regulation of arm and leg movement during human locomotion. Neuroscientist
14. Zehr, E.P., K.L. Hesketh, and R. Chua. Differential regulation of cutaneous and H-reflexes during leg cycling in humans. J. Neurophysiol
. 85:1178–1185, 2001.
15. Zehr, E.P., and A. Kido. Neural control
of rhythmic, cyclical human arm movement: task dependency, nerve specificity and phase modulation of cutaneous reflexes. J. Physiol
. 537:1033–1045, 2001.