36: Locomotion

• A Complex Sequence of Muscle Contractions Is Required for Stepping

• The Motor Pattern for Stepping Is Organized at the Spinal Level

  • Contraction in Flexor and Extensor Muscles of the Hind Legs Is Controlled by Mutually Inhibiting Networks

  • Central Pattern Generators Are Not Driven by Sensory Input

  • Spinal Networks Can Generate Complex Locomotor Patterns

• Sensory Input from Moving Limbs Regulates Stepping

  • Proprioception Regulates the Timing and Amplitude of Stepping

  • Sensory Input from the Skin Allows Stepping to Adjust to Unexpected Obstacles

• Descending Pathways Are Necessary for Initiation and Adaptive Control of Stepping

  • Pathways from the Brain Stem Initiate Walking and Control Its Speed

  • The Cerebellum Fine-Tunes Locomotor Patterns by Regulating the Timing and Intensity of Descending Signals

  • The Motor Cortex Uses Visual Information to Control Precise Stepping Movements

  • Planning and Coordination of Visually Guided Movements Involves the Posterior Parietal Cortex

• Human Walking May Involve Spinal Pattern Generators

• An Overall View

The ability to move is essential for the survival of animals. Although many forms of locomotion have evolved—swimming, flying, crawling, and walking—all use rhythmic and alternating movements of the body or appendages. This rhythmicity makes locomotion appear to be repetitive and stereotyped. Indeed, locomotion is controlled automatically at relatively low levels of the central nervous system without intervention by higher centers. Nevertheless, locomotion often takes place in environments that are either unfamiliar or present unpredictable conditions. Locomotor movements must therefore be continually modified, usually in a subtle fashion, to adapt otherwise stereotyped movement patterns to the immediate surroundings.

The study of the neural control of locomotion must address two fundamental questions. First, how do assemblies of nerve cells generate the rhythmic motor patterns associated with locomotor movements? Second, how does sensory information adjust locomotion to both anticipated and unexpected events in the environment? In this chapter we address both of these questions by examining the neural mechanisms controlling walking.

Although most information on neural control of walking has come from studying the cat's stepping movements, important insights have also come from studies of other animals as well as rhythmic behaviors other than locomotion. Therefore, we shall also consider the more general question of how rhythmic motor activity can be generated and sustained by networks of neurons.

Several critical insights into the neural mechanisms controlling quadrupedal stepping were obtained nearly a century ago when it was found that removing the cerebral hemispheres in dogs did not abolish walking—decerebrate animals are still able to walk spontaneously. One animal was observed to rear itself up in order to rest its forepaws on a gate at feeding time. It was soon discovered that stepping of the hind legs could be induced in cats and dogs after complete transection of the spinal cord. The stepping movements in these spinal preparations (Box 36–1) are similar to normal stepping. Nonrhythmic electrical stimulation of the cut cord elicits stepping at a rate related to the intensity of the stimulating current. Another important early observation was that passive movement of a limb by the experimenter could initiate stepping movements in spinal cats and dogs, suggesting that proprioceptive reflexes are crucial in regulating the movements.

Finally, in 1911 Thomas Graham Brown discovered that rhythmic, alternating contractions could be evoked in deafferented hind leg muscles immediately after transection of the spinal cord. He therefore proposed the concept of the half-center, whereby flexors and extensors inhibit each other reciprocally, giving rise to alternating stepping movements. Four conclusions can be drawn from these early studies.

1. Supraspinal commands are not necessary for producing the basic motor pattern for stepping.

2. The basic rhythmicity of stepping is produced by neuronal circuits contained entirely within the spinal cord.

3. The spinal circuits can be modulated by tonic descending signals from the brain.

4. The spinal pattern-generating networks do not require sensory input but nevertheless are strongly regulated by input from limb proprioceptors.

For almost half a century following these early studies few investigations were aimed at establishing the neural mechanisms for walking. Instead, research on motor systems focused on the organization of spinal reflex pathways and the mechanisms of synaptic integration within the spinal cord (see Chapter 35). Modern research on the neural control of locomotion dates from the 1960s and two major experimental successes. First, rhythmic patterns of motor activity were elicited in spinal animals by the application of adrenergic drugs. Second, walking on a treadmill was evoked in decerebrate cats by electrical stimulation of a small region in the brain stem.

At about the same time electromyographic recordings from numerous hind leg muscles in intact cats during unrestrained walking revealed the complexity of the locomotor pattern and brought to prominence the question of how spinal reflexes are integrated with intrinsic spinal circuits to produce the locomotor pattern. Soon thereafter, investigations of stepping in spinal cats demonstrated the similarity of locomotor patterns in spinal preparations and intact animals, thus firmly establishing the idea that the motor output for locomotion is produced primarily by a neuronal system in the spinal cord.

For the purpose of examining the patterns of muscle contraction during locomotion, the step cycle in cats and humans can be divided into four distinct phases: flexion (F), first extension (E₁) , second extension (E₂) , and third extension (E₃) (Figure 36–2A). The F and E₁ phases occur during the time the foot is off the ground (swing) , whereas E₂ and E₃ occur when the foot is in contact with the ground (stance) .

Swing commences with flexion at the hip, knee, and ankle (the F phase). Approximately midway through swing the knee and ankle begin to extend while the hip continues to flex (the E₁ phase). Extension at the knee and ankle during E₁ moves the foot ahead of the body and prepares the leg to accept weight in anticipation of foot contact at the onset of stance. During early stance (the E₂ phase) the knee and ankle joints flex, even though extensor muscles are contracting strongly. A lengthening contraction of ankle and knee extensor muscles occurs because weight is being transferred to the leg. The spring-like yielding of these muscles as weight is accepted allows the body to move smoothly over the foot, and is essential for establishing an efficient gait. During late stance (the E₃ phase) the hip, knee, and ankle all extend to provide a propulsive force to move the body forward.

The rhythmic movements of the legs during stepping are produced by contractions of many muscles. In general, contractions of flexor muscles occur during the F phase, whereas contractions of extensor muscles occur during one or more of the E phases. However, the timing and amounts of activity are different in different muscles (Figure 36–2B). For example, a hip flexor muscle (iliopsoas) contracts continuously during the F and E₁ phases, whereas a knee flexor muscle (semitendinosus) contracts briefly at the beginning of the F and E₂ phases. Another complexity is that some muscles contract during both swing and stance. Thus the motor pattern for stepping is not merely alternating flexion and extension at each joint, but a complex sequence of muscle contractions, each precisely timed and scaled to achieve a specific task in the act of locomotion.

Transection of the spinal cord of quadrupeds initially causes complete paralysis of the hind legs. It does not, however, permanently abolish the capacity of hind legs to make stepping movements: Hind leg stepping often recovers spontaneously over a period of a few weeks, particularly if the transection is made in young animals. Recovery of stepping in adult cats can be facilitated by daily training on a treadmill evoked by nonspecific cutaneous stimulation of the perineal region.

Electromyographic records of the hind leg muscles of chronic spinal cats during stepping are quite similar to those of normally walking animals. Many of the reflex responses that occur in normal animals can also be evoked in spinal animals. Spinal animals are not, however, able to maintain balance on the treadmill. Adequate control of balance requires descending signals from brain stem centers, such as the vestibular nuclei.

Contraction in Flexor and Extensor Muscles of the Hind Legs Is Controlled by Mutually Inhibiting Networks

From the studies by Graham Brown early in the 20th century we know that the isolated spinal cord can generate rhythmic bursts of reciprocal activity in flexor and extensor motor neurons of the hind legs even in the absence of sensory input (Figure 36–3). Graham Brown proposed that contractions in the flexor and extensor muscles are controlled by two systems of neurons, or half-centers, that mutually inhibit each other (Figure 36–4B).

According to Graham Brown, activity alternates between half-centers because of fatigue of the inhibitory connections. For example, if two half-centers receive tonic excitatory input, and the flexor half-center receives the stronger input, the flexor muscles will contract while the extensor half-center is inhibited. Then, as the inhibitory output fatigues, the extensor half-center's output will increase, causing inhibition of the flexor half-center and contraction of the extensor muscles until its inhibitory output fatigues. Thus the flexor and extensor muscles controlled by the two half-centers will alternately contract and relax as long as the half-centers receive sufficient tonic excitatory input.

The half-center hypothesis was supported by studies in the 1960s on the effects in spinal cats of the drug l-dihydroxyphenylalanine (l-DOPA), a precursor of the monoamine transmitters dopamine and norepinephrine. After the cats were treated with l-DOPA, brief trains of electrical stimuli were applied to small-diameter sensory fibers from skin and muscle. These trains of stimuli evoked long-lasting bursts of activity in either flexor or extensor motor neurons, depending on whether ipsilateral or contralateral nerves were stimulated. The afferents producing these effects are collectively termed flexor reflex afferents. Additional administration of nialamide, a drug that prolongs the action of norepinephrine released in the spinal cord, often resulted in short sequences of rhythmic reciprocal activity in flexor and extensor motor neurons (Figure 36–4A).

The system of interneurons generating flexor bursts was found to inhibit the system of interneurons generating the extensor bursts, and vice versa (Figure 36–4B). This organizational feature is consistent with Graham Brown's theory that mutually inhibiting "half-centers" produce the alternating burst activity in flexor and extensor motor neurons. The interneurons mediating the burst patterns arising from stimulation of the flexor reflex afferents have not been fully identified but they may include interneurons in the intermediate region of the gray matter in the sixth lumbar segment. Interneurons in this region produce prolonged bursts of activity in response to brief stimulation of either ipsilateral or contralateral flexor reflex afferents in spinal cats treated with l-DOPA (Figure 36–4C).

Central Pattern Generators Are Not Driven by Sensory Input

Neuronal spinal networks capable of generating rhythmic motor activity in the absence of rhythmic input from peripheral receptors are termed central pattern generators (Box 36–2). Largely because the mammalian nervous system is so complex, we lack detailed information on the neuronal circuitry and mechanisms for rhythm generation by central pattern generators in the mammalian spinal cord. However, we have considerable knowledge about the cellular, synaptic, and network properties of central pattern generators in invertebrates and lower vertebrates, which have less complex nervous systems than mammals. For example, the analysis by Sten Grillner of the central pattern generator controlling swimming in the lamprey has provided considerable insight into the mechanisms of rhythm generation in vertebrate motor systems (Box 36–3).

Recent studies using the spinal cord of the neonatal rat have identified several classes of rhythmically active interneurons and demonstrated that some of these interneurons can generate sustained membrane depolarizations (plateau potentials) in response to weak synaptic input. Because this active membrane property is known to be important for rhythm generation in simpler systems, it is likely that it also contributes to rhythm generation in the mammalian spinal cord.

Spinal Networks Can Generate Complex Locomotor Patterns

The locomotor patterns generated in deafferented or immobilized spinal animals are generally much simpler than normal stepping patterns. Usually they consist of alternating bursts of activity in flexor and extensor motor neurons (Figure 36–7A). More complex locomotor patterns can be generated in immobilized spinal animals after a period of training or through the application of additional drugs such as 4-aminopyridine. Moreover, in decerebrate animals elaborate locomotor patterns can be produced in hind limb motor neurons after deafferentation (Figure 36–7B). These patterns resemble those recorded in the same animals before deafferentation. Finally, a variety of patterns can be generated in immobilized decerebrate animals, and these patterns can be altered significantly by changing the level of tonic sensory input (Figure 36–7C).

From these observations it is clear that the spinal pattern-generating network for each leg can produce a variety of motor patterns. Which pattern is generated depends on many factors, such as the supraspinal and tonic sensory inputs to the spinal pattern generators as well as the drugs used to experimentally initiate rhythmicity. This functional flexibility may be explained by a scheme in which mutually inhibiting half-centers produce the basic rhythmicity and establish a general pattern of reciprocal activity in flexor and extensor motor neurons, whereas the details of the temporal pattern are established by a network of interneurons (the patterning network) connecting the half-centers and motor neurons (Figure 36–7D).

Although normal walking is automatic, it is not necessarily stereotyped. Mammals constantly use sensory input to adjust their stepping patterns to variations in terrain and to anticipated and unexpected conditions. This input is provided predominantly by three sensory systems: the visual, vestibular, and somatosensory systems. Proprioceptors in muscles and joints provide information about body movements and are involved in the automatic regulation of stepping. Cutaneous receptors in the skin, sometimes referred to as exteroceptors, adjust stepping to external stimuli and can also provide important feedback about body movements.

Proprioception Regulates the Timing and Amplitude of Stepping

One of the clearest indications that somatosensory signals from the limbs regulate the step cycle is that the rate of stepping in spinal and decerebrate cats matches the speed of the motorized treadmill belt on which they tread. Specifically, afferent input regulates the duration of the stance phase. As the stepping rate increases, the stance phase becomes shorter while the swing phase remains relatively constant. This observation suggests that some form of sensory input signals the end of stance and thus leads to the initiation of swing.

Sherrington was the first to propose that proprioceptors in muscles acting at the hip are primarily responsible for regulating the stance phase. He noticed that rapid extension at the hip joint, but not at the knee and ankle joints, led to contractions in the hip flexor muscles of chronic spinal cats and dogs. More recent studies have found that preventing hip extension in a limb suppresses stepping in that limb, whereas rhythmically moving the hip can entrain the locomotor rhythm, that is, cause the timing of the neural output to match the rhythm of the externally imposed movements.

During entrainment a burst of activity in hip flexor motor neurons is initiated in synchrony with hip extension (Figure 36–8A). The afferent fibers that signal hip angle for swing initiation arise from muscle spindles in the hip flexor muscles. In decerebrate animals stretching these muscles to mimic the lengthening that occurs at the end of the stance phase inhibits the extensor half-center and thus facilitates the initiation of burst activity in flexor motor neurons during walking (Figure 36–8B).

Other important proprioceptive signals for regulating the step cycle arise from proprioceptors in extensor muscles. Electrical stimulation of sensory fibers from Golgi tendon organs and muscle spindles prolongs the stance phase, often delaying the onset of swing until the stimulus has ended (Figure 36–9A). Sensory fibers from both types of receptors are active during stance.

The intensity of the signal from the Golgi tendon organs is related to the load carried by the leg. Golgi tendon organs have an excitatory action on ankle extensor motor neurons during walking but an inhibitory action when the body is at rest (see Chapter 35). The functional consequence of this reflex reversal is that the swing phase is not initiated until the extensor muscles are unloaded and the forces exerted by these muscles are low, as signaled by a decrease in activity from the Golgi tendon organs. Unloading of extensor muscles normally occurs near the end of stance, when the animal's weight is borne by the other legs and the extensor muscles are shortened and thus unable to produce high forces.

In addition to regulating the transition from stance to swing, proprioceptive information from muscle spindles and Golgi tendon organs contributes significantly to the generation of burst activity in extensor motor neurons. Reducing this sensory input in cats diminishes the level of extensor activity by more than half; in humans it has been estimated that up to 30% of the activity of ankle extensor motor neurons is caused by feedback from the extensor muscles.

At least three excitatory pathways transmit sensory information from extensor muscles to extensor motor neurons: a monosynaptic pathway from primary muscle spindles (group Ia afferents), a disynaptic pathway from primary muscle spindles and Golgi tendon organs (group Ia and Ib afferents), and a polysynaptic pathway from primary muscle spindles and Golgi tendon organs that includes interneurons in the central pattern generator (Figure 36–9B). The continuous regulation of the level of extensor activity by proprioceptive feedback presumably allows automatic adjustment of force and length in extensor muscles in response to unexpected unloading and loading of the leg.

Sensory Input from the Skin Allows Stepping to Adjust to Unexpected Obstacles

Sensory receptors in the skin have a powerful influence on the central pattern generator for walking. One important function of these receptors is to detect obstacles and adjust stepping movements to avoid them. A well-studied example is the stumbling-corrective reaction in cats. A mild mechanical stimulus applied to the dorsal part of the paw during the swing phase produces excitation of flexor motor neurons and inhibition of extensor motor neurons, leading to rapid flexion of the paw away from the stimulus and elevation of the leg in an attempt to step over the object. Because this corrective response is readily observed in spinal cats, it must be produced to a large extent by circuits entirely contained within the spinal cord.

One of the interesting features of the stumbling-corrective reaction is that corrective flexion movements are produced only if the paw is stimulated during the swing phase. An identical stimulus applied during the stance phase produces the opposite response, excitation of extensor muscles that reinforces the ongoing extensor activity. This extensor action is appropriate; if a flexion reflex were produced during the stance phase the animal might collapse because it is being supported by the limb. This is an example of a phase-dependent reflex reversal: The same stimulus excites one group of motor neurons during one phase of locomotion but activates the antagonist motor neurons during another phase.

Although the basic motor pattern for stepping is generated in the spinal cord, fine control of stepping movements involves many regions of the brain, including the motor cortex, cerebellum, and various sites within the brain stem. Many neurons in each of these regions are rhythmically active during locomotor activity and hence participate in the production of the normal motor pattern. Each region, however, plays a different role in the regulation of locomotor function.

Supraspinal regulation of stepping can be divided into three functional systems. One activates the spinal locomotor system, initiates walking, and controls the overall speed of locomotion; another refines the motor pattern in response to feedback from the limbs; and the third visually guides limb movement (Figure 36–10).

Pathways from the Brain Stem Initiate Walking and Control Its Speed

In their seminal studies of decerebrate cats, Mark Shik, Fidor Severin, and Grigori Orlovsky found that tonic electrical stimulation of the mesencephalic locomotor region initiates stepping when decerebrate animals are placed on a treadmill. The mesencephalic locomotor region is situated approximately 6 mm ventral to the inferior colliculus (Figure 36–11A), close to the cuneiform nucleus. The rhythm of the locomotor pattern is related only to the intensity of electrical stimulation, not to its pattern. Weak stimulation produces a slow walking gait, which accelerates as the intensity increases; progressively stronger stimulation produces trotting and finally galloping (Figure 36–11B).

The transition from trotting to galloping is especially interesting for it involves a shift from an out-of-phase relationship between left and right legs in trotting to an in-phase relationship in galloping. This shift in interlimb coordination is also observed in spinal cats walking on a motorized treadmill as the speed is increased. Therefore it is most likely implemented by local circuits in the spinal cord.

In addition to the mesencephalic locomotor region, several other motor regions of the brain can produce locomotion when stimulated electrically. These include a subthalamic locomotor region and a nucleus in the pontine reticular formation (pontine peduncular nucleus). How these different brain stem regions interact in the normal control of locomotion is not yet known.

How are signals from locomotor regions of the brain stem transmitted to patterning networks in the spinal cord? Because application of adrenergic drugs is often sufficient to initiate stepping in acute spinal animals, an early hypothesis was that the initiation and maintenance of locomotor activity depends on activity in the descending noradrenergic pathway from the locus ceruleus or the descending serotonergic pathway from the raphe nucleus. However, because locomotor activity can be evoked after depletion of norepinephrine and serotonin, neither of these aminergic pathways is essential for locomotion. The current view is that these transmitters are modulators that regulate the magnitude and timing of motor neuron activity in the locomotor networks in the spinal cord. Although adrenergic drugs can initiate stepping movements in the spinal animal, aminergic neurons may not serve this function in nature.

Clues about the identity of the descending system that initiates locomotor activity came first from studies on the neonatal rat and lamprey. Administration of glutamate receptor agonists to the isolated spinal cord was found to initiate locomotor activity. In decerebrate cats administration of agonists to the NMDA-type glutamate receptors in the spinal cord initiates locomotor activity similar to that evoked by electrical stimulation of the mesencephalic locomotor region. The application of glutamate receptor antagonists prevents this response. These observations suggest that glutamatergic pathways are involved in initiating locomotor activity.

Considerable research has been devoted to identifying the origin and course of the descending pathways that initiate locomotor activity. The central pattern generators in the spinal cord cannot be directly activated by neurons in the nuclei near the mesencephalic locomotor region, for the axons of these neurons do not descend directly to the cord. Instead, the mesencephalic motor neurons form connections with neurons in the medullary reticular formation, whose axons descend in the ventrolateral region of the spinal cord (Figure 36–11A). These neurons are excited by stimulation of the mesencephalic locomotor region, and transection of their axons in the ventrolateral funiculus of the spinal cord prevents stimulation of the mesencephalic locomotor region from initiating locomotor activity. Thus current evidence indicates that the signals that activate locomotion and control its speed are transmitted to the spinal cord by glutamatergic neurons with axons that travel in ventral reticulospinal pathways.

The Cerebellum Fine-Tunes Locomotor Patterns by Regulating the Timing and Intensity of Descending Signals

Damage to the cerebellum results in marked abnormalities in locomotor movements, including a widened base of support, impaired coordination of joints, and abnormal coupling between limbs during stepping. These symptoms are called ataxia (see Chapter 42). Ataxic walking resembles a drunken gait. Because ataxic gait is apparent in patients with cerebellar lesions even when they are walking on level surfaces, we can conclude that the cerebellum is involved in regulating all stepping movements.

The cerebellum receives sensory information about the actual stepping movements as well as information on the state of the central pattern generators through two ascending pathways. For the hind legs of the cat these are the dorsal and ventral spinocerebellar tracts. Neurons in the dorsal tract are strongly activated by numerous leg proprioceptors and thus provide the cerebellum with detailed information about the mechanical state of the hind legs. In contrast, neurons in the ventral tract are activated primarily by interneurons in the central pattern generator, thus providing the cerebellum with information about the state of the spinal locomotor network. The cerebellum also receives input from the motor cortex and other forebrain regions related to locomotor function.

It is thought that the cerebellum compares actual movements of the legs—proprioceptive signals in the dorsal spinocerebellar tract—with intended movements— the central commands conveyed in the ventral spinocerebellar tract and collaterals of corticospinal tract neurons. When these two types of information differ, representing an error, the cerebellum computes corrective signals and sends these to various brain stem nuclei.

During walking the cerebellum influences several brain stem nuclei, including the vestibular nuclei, red nucleus, and nuclei in the medullary reticular formation. The signals from the cerebellum to the vestibular nuclei during walking are likely used to regulate balance by integrating information about the position and movement of the head derived from the vestibular system with proprioceptive information about movements of the legs.

The Motor Cortex Uses Visual Information to Control Precise Stepping Movements

Normal walking is often guided by vision, and the motor cortex is essential for visually guided movement. Experimental lesions of the motor cortex do not prevent animals from walking on a smooth floor or even on smooth inclines, but they severely impair tasks requiring a high degree of visuomotor coordination, such as walking on the rungs of a horizontal ladder, stepping over a series of barriers, and stepping over single objects placed on a treadmill belt. Such "skilled walking" is associated with considerable modulation in the activity of a large number of neurons in the motor cortex (Figure 36–12).

Many of these neurons project directly to the spinal cord and thus may regulate the activity of interneurons in the central pattern generator for locomotion, thereby adapting the timing and magnitude of motor activity to a specific task. Electrical stimulation of either the motor cortex or the corticospinal tract in normal walking cats influences the timing of locomotor activity; these effects are generally stronger than those elicited by stimulation of brain stem regions.

Planning and Coordination of Visually Guided Movements Involves the Posterior Parietal Cortex

When humans and animals approach an obstacle in their pathway they must adjust their stepping to either move around the obstacle or step over it. These adjustments begin two or three steps before the obstacle is reached. Recent studies by Trevor Drew and his colleagues have shown that the posterior parietal cortex has an essential role in planning these adjustments.

Small lesions in this region cause walking cats to misplace the positioning of their paws as they approach an obstacle, and to increase the probability that one or more legs contact the obstacle as they step over it. Recordings in the parietal cortex have identified sets of neurons that could be involved in the planning and coordination of stepping movements. Some neurons increase their activity as an animal approaches an obstacle, whereas others maintain their activity when an obstacle passes under the animal (Figure 36–13).

Visual information about the size and location of an obstacle is registered as the obstacle is approached, and this information is stored in working memory, a form of short-term memory, to guide the legs. In humans obstructing vision one stride before an obstacle is encountered does not influence foot clearance over the obstacle. In quadrupeds, however, such working memory is necessary because an obstacle is no longer within the visual field by the time the hind legs are stepping over it.

Many of the features of the working memory for guiding the legs have recently been established in cats. The main features are that the forelegs must step over the obstacle in order for the memory to be established, the trajectories of the hind legs are scaled appropriately for the height of the obstacle and for the relative positions of the hind paws and the obstacle, and the memory persists for many minutes without declining.

The neurobiological mechanisms underlying this form of memory remain to be established, but the persistence of the memory appears to depend upon neuronal systems in the posterior parietal cortex. With bilateral lesions of the medial posterior parietal cortex the memory is completely abolished after a few seconds (Figure 36–14). Complementing this observation is the recent identification by Trevor Drew and colleagues of neurons in the posterior parietal cortex whose activity is elevated while the cat straddles an obstacle with the forelegs on one side and the hind legs on the other side.

Unlike spinal cats and other quadrupeds, humans with spinal cord transection generally are not able to walk spontaneously. Nevertheless, some observations of patients with spinal cord injury parallel the findings from studies of spinal cats. In one striking case a patient with nearly complete transection of the spinal cord showed uncontrollable, spontaneous, rhythmic movements of the legs when the hips were extended. This behavior closely parallels the rhythmic stepping movements in chronic spinal cats. In another study on a few patients with severe spinal cord injury, stepping on a treadmill was improved by clonidine, a drug influencing biogenic amines.

Compelling evidence for the existence of spinal rhythm-generating networks in humans comes from studies of development. Human infants make rhythmic stepping movements immediately after birth if held upright and moved over a horizontal surface. This strongly suggests that some of the basic neuronal circuits for locomotion are innate. Because stepping can occur in infants who lack cerebral hemispheres (anencephaly), these circuits must be located at or below the brain stem, perhaps entirely within the spinal cord.

During the first year of life, as automatic stepping is transformed into functional walking, these basic circuits are thought to be brought under supraspinal control in two ways. First, the infant develops voluntary control of locomotion. From what we know about the neuronal mechanisms in the cat, this ability could depend on the maturation of reticulospinal pathways and regions of the brain stem that project to reticulospinal neurons, such as the mesencephalic locomotor region. Second, the stepping pattern gradually develops from a primitive flexion-extension pattern that generates little effective forward movement to the mature pattern of complex movements. Again based on studies of cats, it is plausible that this adaptation reflects maturation of descending systems that originate in the motor cortex and brain stem nuclei and which are modulated by the cerebellum.

Parallels between human and quadrupedal walking have also been found in patients trained after spinal cord injury. Daily training restores stepping in spinal cats and improves stepping in patients with chronic spinal injuries. In patients with complete spinal cord injury motor patterns evoked in response to rhythmic leg movements can also be modified by daily therapy (Box 36–4). Thus the spinal networks for locomotion in humans and cats can adapt.

In light of these findings there is reason to believe that human walking relies on the same general principles of neuronal organization as quadrupedal walking: Intrinsic oscillatory networks are activated and modulated by other brain structures and by afferent input. Nevertheless, human bipedal locomotion differs from most animal locomotion in that it places significantly greater demands on descending systems that control balance during walking. Indeed, maturation of the systems that control balance and stepping patterns is necessary for the infant to begin walking independently at the end of the first year. In contrast, horses can stand and walk within hours after birth. It is therefore likely that the locomotor spinal networks in humans are more dependent on supraspinal centers than those in quadrupeds. This dependence may explain in part the relatively few observations of spontaneous stepping movements in humans with complete spinal cord injury.

Locomotion in mammals typically involves rhythmic movements of the body and one or more limbs. These movements depend on precise regulation of the timing and strength of contractions in numerous muscles. Local spinal circuits, known as central pattern generators, can produce the basic motor pattern for locomotion even without sensory information from peripheral receptors. Numerous examples of central pattern generators have been analyzed at the cellular level, and it is clear that a wide variety of cellular, synaptic, and network properties are involved in these local networks.

Central pattern generators are extremely flexible; their cellular and synaptic properties can be modified by modulatory signals at chemical synapses. Their functioning depends on how they are activated and the pattern of afferent input they receive.

Modern research on mammalian locomotion dates from the 1960s, when two important experimental animal preparations were introduced. In the decerebrate animal stepping can be initiated by electrical stimulation of a site in the brain stem, the mesencephalic locomotor region. In the spinal preparation centrally generated locomotor activity can be evoked after the administration of adrenergic drugs.

Using these preparations, investigators have confirmed and extended fundamental observations made at the beginning of the 20th century, namely that the basic rhythm for locomotion is generated centrally in spinal networks, that the transition from stance to swing is regulated by afferent signals from leg flexor and extensor muscles, and that descending signals from the brain regulate the intensity of locomotion and modify stepping movements according to the terrain on which the animal is walking.