The Muscle Spindle

Muscle spindles are found in almost all skeletal muscles and are particularly concentrated in muscles that exert fine motor control (e.g., the small muscles of the hand and eye).

Structure of the Muscle Spindle.

As its name implies, a muscle spindle is a spindle or fusiform-shaped organ composed of a bundle of specialized muscle fibers richly innervated both by sensory and by motor axons (Fig. 9-1). A muscle spindle is about 100 μm in diameter and up to 10 mm long. The innervated part of the muscle spindle is encased in a connective tissue capsule. Muscle spindles lie between regular muscle fibers and are typically located near the tendinous insertion of the muscle. The distal ends of the spindle are attached to the connective tissue within the muscle (endomysium). Thus, muscle spindles lie in parallel with the regular muscle fibers. This arrangement has important functional implications, as will be made clear later.

Figure 9-1 Muscle proprioceptors. Skeletal muscles contain sensory receptors embedded within the muscle (spindles) and within their tendons (Golgi tendon organs). A, Schematic of a muscle showing the arrangement of a spindle in parallel with extrafusal muscle fibers and a tendon organ in series with muscle fibers. B, Structure and motor and sensory innervation of a muscle spindle. C, Structure and innervation of a tendon organ.

The muscle fibers within the spindle are called intrafusal fibers to distinguish them from the regular or extrafusal fibers that make up the bulk of the muscle. Individual intrafusal fibers are much narrower than extrafusal fibers and do not run the length of the muscle. Thus, they are too weak to contribute significantly to muscle tension or to directly cause changes in the overall length of the muscle by their contraction.

Morphologically, two types of intrafusal muscle fibers are found within muscle spindles: nuclear bag and nuclear chain fibers (Fig. 9-1, B). These names are derived from the arrangement of nuclei in the fibers. Nuclear bag fibers are larger than nuclear chain fibers, and their nuclei are bunched together like a bag of oranges in the central, or equatorial, region of the fiber. In nuclear chain fibers, the nuclei are arranged in a row. Functionally, nuclear bag fibers are divided into two types: bag1 and bag2. As detailed later, bag2 fibers are functionally similar to chain fibers.

Page 160

The neural innervation of an intrafusal fiber differs significantly from that of an extrafusal fiber, which is innervated by a single motor neuron. Intrafusal fibers are multiply innervated and receive both sensory and motor innervation. The sensory supply includes a single group Ia afferent and a variable number of group II afferent fibers (Fig. 9-1, B). Group Ia fibers belong to the largest-diameter class of sensory nerve fibers and conduct at 72 to 120 m/sec; group II fibers are intermediate in size and conduct at 36 to 72 m/sec. A group Ia afferent fiber forms a primary ending consisting of a spiral-shaped terminal composed of branches of the group Ia fiber on each of the intrafusal muscle fibers. Thus, terminals of primary endings are found on both types of nuclear bag fibers and on nuclear chain fibers. The group II afferent fiber forms a secondary ending, which is found on nuclear chain and bag2 fibers, but not on bag1 fibers. The primary and secondary endings have mechanosensitive channels that are sensitive to the level of tension on the intrafusal muscle fiber.

The motor supply to a muscle spindle consists of two types of γ motor axons (Fig. 9-1, B). Dynamic γ motor axons end on nuclear bag1 fibers, and static γ motor axons end on nuclear chain and bag2 fibers.

Muscle Spindles Detect Changes in Muscle Length.

Muscle spindles respond to changes in muscle length because they lie in parallel with the extrafusal fibers and therefore will also be stretched or shortened along with the extrafusal fibers. Because intrafusal fibers, like all muscle fibers, display spring-like properties, a change in their length will change the tension that they are under, and this change is sensed by mechanoreceptors of the Ia and II spindle afferents.

Figure 9-2 shows the changes in activity of the afferent fibers of a muscle spindle when the muscle is stretched. It is clear that Ia and II fibers respond differently to stretch. Group Ia fibers are sensitive both to the amount of muscle stretch and to its rate, whereas group II fibers respond chiefly to the amount of stretch. Thus, when a muscle is stretched to a new longer length, group II firing will increase in proportion to the amount of stretch (Fig. 9-2, left), and when the muscle is allowed to shorten, its firing rate will decrease proportionately (Fig. 9-2, right). Group Ia fibers show this same static-type response, and thus under steady-state conditions (i.e., constant muscle length), their firing rate will reflect the amount of muscle stretch, similar to that of group II fibers. However, while muscle length is changing, group Ia firing also reflects the rate of stretch or shortening that the muscle is undergoing. Its activity overshoots during muscle stretch and undershoots (and possibly ceases) during muscle shortening. These are called dynamic responses. This dynamic sensitivity also means that the activity of group Ia fibers is much more sensitive to transient stretches, such as shown in the middle diagrams of Figure 9-2. In particular, the tap profile is what occurs when a doctor uses a reflex hammer to hit the muscle tendon and thereby cause a brief stretching of the attached muscle. The change in muscle length is too brief for changes in group II firing to occur, but because the magnitude of the rate of change (slopes of the tap profile) is so high with this stimulus, large dynamic responses are elicited in the group Ia fibers. Thus, the functionality of reflex arcs involving Ia afferents is what is being assessed by using a reflex hammer to tap on tendons.

Figure 9-2 Responses of a primary ending (Ia) and a secondary ending (II) to changes in muscle length. Note the difference in dynamic and static responsiveness of these endings. The waveforms at the top represent the changes in muscle length. The middle and bottom rows show the discharges of a group Ia and II fiber, respectively, during the various changes in muscle length.

γ Motor Neurons Adjust the Sensitivity of the Spindle.

Up to this point we have described only how muscle spindles behave when there are no changes in γ motor neuron activity. The efferent innervation of muscle spindles is extremely important, however, because it determines the sensitivity of muscle spindles to stretch. For example, in Figure 9-3, A, the activity of a muscle spindle afferent is shown during a steady stretch. When the extrafusal portion of the muscle contracts (Fig. 9-3, B), the muscle spindle is unloaded by the resultant shortening of the muscle, and the muscle spindle afferent may stop discharging and thus become insensitive to further changes in muscle length. However, this unloading of the spindle can be counteracted if γ motor neurons are simultaneously stimulated. Such stimulation causes the intrafusal muscle fibers of the spindle to shorten along with the extrafusal muscle fibers (Fig. 9-3, C). Actually, only the two polar regions of the intrafusal muscle contract; the equatorial region, where the nuclei are located, does not contract because it has little contractile protein. As a result, when the polar regions contract, the equatorial region elongates and regains its sensitivity. Conversely, when a muscle relaxes and thus elongates, a concurrent decrease in γ motor neuron activity will allow the intrafusal fibers to relax as well and thereby prevent the tension on the central portion of the intrafusal fiber from reaching a level at which firing of the afferents is saturated. Thus, the γ motor neuron system allows the muscle spindle to operate over a wide range of muscle lengths while retaining high sensitivity to small changes in length.

Figure 9-3 The activity of γ motor neurons can counteract the effects of unloading on the discharge of a muscle spindle afferent. A, The activity of a muscle spindle afferent is shown during steady stretch. B, α Motor neuron stimulation at time t = 0 msec causes contraction of the extrafusal fibers, which leads to muscle shortening and increased muscle tension, but unloading of the tension across the muscle spindle, which in turn induces the afferent to stop firing. Upon relaxation the muscle returns to its original length and tension is restored on the intrafusal fibers, causing the return of activity in the Ia afferent. C, Coactivation of α and γ motor neurons causes shortening of both extrafusal and intrafusal fibers. Thus, there is no unloading of the spindle, and the afferent maintains its spontaneous activity.

(Redrawn from Kuffler SW, Nicholls JG: From Neuron to Brain. Sunderland, MA, Sinauer, 1976.)

Descending motor commands from the brain typically activate α and γ motor neurons simultaneously and thus cause a synchronous contraction of extrafusal and intrafusal muscle fibers. This co-contraction means that as the muscle shortens from the contraction of extrafusal fibers, the polar regions of the intrafusal fibers also shorten, thereby maintaining relatively constant tension on the equatorial portion and thus the sensitivity of the spindle apparatus.

Page 161

As mentioned earlier, there are two types of γ motor neurons—dynamic and static (Fig. 9-1). Dynamic γ motor axons end on nuclear bag1 fibers, and static γ motor axons synapse on nuclear chain and bag2 fibers. Thus, when a dynamic γ motor neuron is activated, the response of the group Ia afferent fiber is enhanced, but the activity of the group II afferents is unchanged; when a static γ motor neuron discharges, the responsiveness of the group II afferents and the static responsiveness of the group Ia afferents are increased. The effects of stimulating the static and dynamic fibers on a group Ia afferent’s response to stretch is illustrated in Figure 9-4. Descending pathways can preferentially influence dynamic or static γ motor neurons and thereby alter the nature of reflex activity in the spinal cord.

Figure 9-4 Effects of static and dynamic γ motor neurons on the responses of a primary ending to muscle stretch. The upper trace, A, is the time course of the stretch. B shows the discharge of group Ia fibers in the absence of γ motor neuron activity. In C, a static γ motor axon was stimulated, and in D, a dynamic γ motor axon was stimulated.

(Redrawn from Crowe A, Matthews PBC: J Physiol 174:109, 1964.)

Golgi Tendon Organ

A second type of mechanosensitive receptor associated with skeletal muscle is the Golgi tendon organ (Fig. 9-1). A Golgi tendon organ is innervated from the terminals of group Ib afferent fibers. The diameter of a Golgi tendon organ is about 100 μm and its length is about 1 mm. A group Ib fiber has a large diameter and conducts in the same velocity range as a group Ia fiber. The terminals of a Ib fiber are wrapped about bundles of collagen fibers in the tendon of a muscle (or in tendinous inscriptions within the muscle). Thus, the sensory ending is arranged in series with the muscle, in contrast to the parallel arrangement of the muscle spindle.

Because of their in-series relationship to the muscle, Golgi tendon organs can be activated either by muscle stretch or by contraction of the muscle. In both cases, however, the actual stimulus sensed by the Golgi tendon organ is the force that develops in the tendon to which it is linked. Thus, the response to stretch is the result of the spring-like nature of the muscle (i.e., by Hooke’s law, the force on a spring is proportional to how much it is stretched).

The distinction between the responsiveness of the muscle spindles and Golgi tendon organs can be made clear by comparing the firing patterns of Ia and Ib fibers when a muscle is stretched and then held at a longer length (Fig. 9-5). The Ia fiber’s firing rate will maintain its increase until the stretch is reversed. In contrast, the Ib fiber will show an initial large increase in firing, reflecting the increased tension on the muscle caused by the stretch, but will then show a gradual return toward its initial firing rate as the tension on the muscle is lowered because of cross-bridge recycling and the resultant lengthening of the sarcomeres. Therefore, Golgi tendon organs signal force, whereas spindles signal muscle length. Further evidence of this distinction is that Ib firing correlates with force level during isometric contraction even though muscle length, and therefore Ia activity, are unchanged.

Figure 9-5 Changes in group Ia and Ib firing rates when muscle is stretched to a new length as indicated in the top graph (blue line). After a transient burst, the firing rate of the Ia fiber remains constant at a new higher level that is proportional to the increase in length (lower graph, blue line). In contrast, the Ib unit shows an initial rapid increase in firing followed by a slow decrease back toward its original level (bottom graph, red line) and has a firing profile that matches the tension level in the muscle caused by the stretch (top graph, red line).

The Myotatic or Stretch Reflex

The stretch reflex is key for the maintenance of posture and helps overcome unexpected impediments during a voluntary movement. Changes in the stretch reflex are involved in actions commanded by the brain, and pathological alterations in this reflex are important signs of neurological disease. The phasic stretch reflex occurs in response to rapid, transient stretches of the muscle, such as those elicited by a doctor using a reflex hammer or by an unexpected impediment to an ongoing movement. The tonic stretch reflex occurs in response to a slower or steady stretch applied to the muscle.

The Phasic (or Ia) Stretch Reflex.

The phasic stretch reflex is elicited by the primary endings of the muscle spindles. The reflex arc responsible for the phasic stretch reflex is shown in Figure 9-6. A group Ia afferent fiber from a muscle spindle in the rectus femoris muscle is shown to branch as it enters the gray matter of the spinal cord. It will form excitatory synapses directly (monosynaptically) on virtually all α motor neurons that supply the same (also known as the homonymous) muscle and with many of its synergists, such as the vastus intermedius muscle in this case, which also acts to extend the leg at the knee. If the excitation is powerful enough, the motor neurons discharge and cause a contraction of the muscle. Note that the Ia fibers do not contact the γ motor neurons, possibly to avoid a positive-feedback loop situation. This selective targeting of α motor neurons is exceptional in that most other reflex and descending pathways will target both α and γ motor neurons.

Figure 9-6 Reflex arc of the stretch reflex. The shortest pathway in this arc contains a single synapse within the CNS; hence, it is a monosynaptic reflex. The interneuron, shown in black, is a group Ia inhibitory interneuron.

Other branches of group Ia fibers end on a variety of interneurons; however, one type, the reciprocal Ia inhibitory interneuron (black cell in Fig. 9-6), is particularly important with regard to the stretch reflex. These interneurons are identifiable because they are the only inhibitory interneurons that receive input from both the Ia afferents and Renshaw cells (see Fig. 9-11). They end on α motor neurons that innervate the antagonist muscles, in this case the hamstring muscles, including the semitendinosus muscle, which act to flex the knee.

Figure 9-11 Renshaw cell (RC) connections with motor neurons and Ia inhibitory interneurons. The circuits shown mediate Ia reciprocal inhibition of antagonist muscles (in this case an extensor) and inhibition of this reciprocal inhibition by Renshaw cells. Note that there are equivalent Renshaw cells and Ia inhibitory interneurons associated with extensor motor neurons and Ia input from spindles in extensor muscles, but they are not shown for simplicity. Orange cells are inhibitory and blue and green ones are excitatory.

The organization of the stretch reflex arc guarantees that one set of α motor neurons is activated and the opposing set is inhibited. This arrangement is known as reciprocal innervation. Although many reflexes involve such reciprocal innervation, this type of innervation is not the only possible organization of a motor control system, and indeed, descending motor pathways can override such patterns.

Page 163

The stretch reflex is quite powerful, in large part because of its monosynaptic nature. The power of this reflex also derives from the essentially maximal convergence and divergence that exist in this pathway, which is not apparent from the circuit diagrams, such as Figure 9-6, that are typically used to illustrate reflex pathways. That is, each Ia fiber will contact virtually all homonymous α motor neurons, and each such α motor neuron will receive input from every spindle in that muscle. Although its monosynaptic nature makes the Ia reflex rapid and powerful, it also means that there is relatively little opportunity for direct control of activity flow through its reflex arc. The CNS overcomes this problem by controlling muscle spindle sensitivity via the γ motor neuron system.

Inverse Myotatic or Ib Reflex

Just as the stretch reflex can be thought of as a feedback system to regulate muscle length, the inverse myotactic, or Ib, reflex can be thought of as a feedback system to help maintain force levels in a muscle. Using the upper part of the leg as an example, the Ib reflex arc is shown in Figure 9-7. In this example the receptor organs are Golgi tendon organs of the rectus femoris muscle. The afferent fibers branch as they enter the spinal cord and end on interneurons. There are no monosynaptic connections to α motor neurons. Rather, the Ib afferents synapse onto two classes of interneurons: interneurons that inhibit α motor neurons that supply the homonymous muscle, in this case the rectus femoris, and excitatory interneurons that activate α motor neurons to the antagonist (semitendinosus). Because there are two synapses in series in the CNS, this is a disynaptic reflex arc. Given these connections, Ib activity should have the opposite action of the Ia stretch reflex during passive stretch of the muscle, which explains this reflex’s other name, the inverse myotactic reflex. However, functionally, the two reflex arcs can act synergistically, as the following example shows. Recall that the Golgi tendon organs monitor force levels across the tendon that they supply. If during maintained posture, such as standing at attention, knee extensors, such as the rectus femoris muscle, begin to fatigue, the force in the patellar tendon will decline. The decline in force will reduce the activity of Golgi tendon organs in this tendon. Because the Ib reflex normally inhibits the α motor neurons to the rectus femoris muscle, reduced activity of the Golgi tendon organs will enhance the excitability of (i.e., disinhibit) the α motor neurons and thereby help reverse the decrease in force caused by the fatigue. Simultaneously, bending of the knee will stretch the knee extensors and activate the Ia fibers, which will then excite the same α motor neurons. Thus, coordinated action of both muscle spindle and Golgi tendon organ afferent fibers is needed to cause greater contraction of the rectus femoris muscle and maintenance of the posture.

Figure 9-7 Reflex arc of the inverse myotatic reflex. The interneurons include both excitatory (clear) and inhibitory (black) interneurons. This is an example of a disynaptic reflex.

Flexion Reflexes and Locomotion

The flexion reflex starts with activation of one or more of a variety of sensory receptors, including nociceptors, whose signals can be carried to the spinal cord via a variety of afferents, including group II and III fibers, collectively called the flexion reflex afferents (FRAs). In flexion reflexes, afferent volleys (1) cause excitatory interneurons to activate the α motor neurons that supply the flexor muscles in the ipsilateral limb and (2) cause inhibitory interneurons to inhibit the α motor neurons that supply the antagonistic extensor muscles (Fig. 9-8). This pattern of activity causes one or more joints in the stimulated limb to flex. In addition, commissural interneurons evoke the opposite pattern of activity in the contralateral side of the spinal cord (Fig. 9-8), which results in extension of the opposite limb, the crossed extension reflex. For our lower limbs (or in quadrupeds for both forelimbs and hind limbs), the crossed extension part of the reflex helps in maintaining balance by enabling the contralateral limb to be able to support the additional load that is transferred to it when the flexed limb is lifted.

Figure 9-8 The reflex arc of the flexion reflex. Black interneurons are inhibitory and clear ones are excitatory. FRA, flexion reflex afferent.

Because flexion typically brings the affected limb in closer to the body and away from a painful stimulus, flexion reflexes are a type of withdrawal reflex. In Figure 9-8, the neural circuit of the flexion reflex is shown for neurons that affect only the knee joint. Actually, however, considerable divergence of the primary afferent and interneuronal pathways occurs in the flexion reflex. In fact, all the major joints of a limb (e.g., hip, knee, and ankle) may be involved in a strong flexor withdrawal reflex. Details of the flexor withdrawal reflex vary, depending on the nature and location of the stimulus. This variability in flexion reflex is called the local sign. Flexor withdrawal reflexes also occur in areas other than the limbs; for example, visceral disease may cause contractions of muscles in the chest wall or abdomen and thereby decrease the mobility of the trunk.

The interneurons subserving flexion reflexes also appear to be part of the central pattern generator (CPG) for generating locomotion and thus are an example of how the reflex circuits are used for multiple purposes. A CPG is a set of neurons and circuits capable of generating the rhythmic activity that underlies motor acts, even in the absence of sensory input. Using the FRA interneurons as an example, one can see that activation of the FRA interneurons leads to a pattern of flexor excitation and extensor inhibition on one side and the converse pattern on the opposite side and that if the FRA interneurons on each side of the spinal cord alternated in being active, a stepping pattern would emerge. That is, walking motion is the result of alternately activating flexors and extensors in each leg such that activation of the flexors (and extensors) in the two legs occurs out of phase with each other, exactly what would be produced by alternately activating the FRA interneurons on each side. Note that such a rhythmic activity pattern in the FRA circuits is not dependent on activity from the FRAs themselves (e.g., they could be activated by descending pathways from the brain).

To show that these circuits are actually involved in generating the locomotion rhythm, spinal cord preparations were made that showed spontaneous locomotion (i.e., if the brainstem is transected and weight is supported, the spinal cord circuits can generate activity that causes the limbs to generate a normal locomotion sequence.) In one such preparation, the electromyogram from the flexors and extensors of a limb were recorded and the FRAs then stimulated to see the effect on locomotion rhythm (Fig. 9-9). Before any stimulus, a spontaneous alternating pattern of flexor and extensor electromyographic (EMG) activity exists. If the FRAs were not involved in the locomotion circuit or at least were not a critical part of the circuits responsible for generating the rhythm (Fig. 9-9, B), we would expect the stimulus to produce only a transient response (i.e., a single EMG response of the flexors and brief inhibition of the extensors) and have no long-term effect on this pattern. Such a transient response is observed (Fig. 9-9, A; EMG records just after the stimulus). However, the stimulus also causes a permanent, approximately 180-degree phase shift in locomotor rhythm, as can be seen by comparing the times of contractions before and after the stimulus. The dashed vertical lines indicate the times at which a flexor EMG response would be expected if the stimulus had produced no phase shift from the EMG activity pattern; before the stimulus, each vertical line is aligned with the onset of a flexor EMG burst, whereas after the stimulus, each vertical line occurs at the end of the flexor burst. Therefore, we can conclude that the stimulus affected the locomotor CPG itself and that the FRA interneurons are a critical part of this CPG (Fig. 9-9, C).

Figure 9-9 Phase reset of locomotion rhythm by FRA stimulation helps identify neuronal components of the underlying central pattern generator (CPG). A, EMG records from knee flexor and extensor muscles. Note the rhythmic alternating pattern before application of the stimulus. The solid vertical lines below each trace indicate the times at which flexor contraction is initiated. The dashed vertical lines indicate the times at which flexor contraction would have been initiated if the stimulus caused no lasting effect on the rhythmic pattern. B and C, Two possible models for the CPG underlying the locomotor rhythm seen in A. B does not include the FRA interneurons in the CPG, whereas C does. The data shown in A support the model shown in C.

(Data from Hultborn H et al: Ann N Y Acad Sci 860:70, 1998.)

A second important point illustrated by this experiment is that the locomotion CPG (and CPGs generally) can be influenced by strong afferent activity. The afferent influence ensures that the pattern generator adapts to changes in the terrain as locomotion proceeds. Such changes may occur rapidly during running, and locomotion must then be adjusted to ensure proper coordination.

Pyramidal versus Extrapyramidal Pathways

Descending motor pathways were traditionally subdivided into pyramidal tract and extrapyramidal pathways. This terminology reflects a clinical dichotomy between pyramidal tract disease and extrapyramidal disease. In pyramidal tract disease, the corticospinal, or pyramidal, tract is interrupted. The signs of this disease were originally attributed to the loss of function of the pyramidal tract (so named because the corticospinal tract passes through the medullary pyramid). However, in many cases of pyramidal tract disease, the functions of other pathways are also altered, and most pyramidal tract signs (see the later section Motor Deficits Caused by Lesions of Descending Motor Pathways) appear to not be caused by loss of the corticospinal tract or at least require damage to additional motor pathways. The term extrapyramidal is even more problematic. Thus, this classification system is not used in this book.

Lateral versus Medial Motor Systems

Another way of classifying the motor pathways is based on their sites of termination in the spinal cord and the consequent differences in their roles in the control of movement and posture. The lateral pathways terminate in the lateral portions of the spinal cord gray matter (Fig. 9-13). The lateral pathways can excite motor neurons directly, although interneurons are their main target. They influence reflex arcs that control fine movement of the distal ends of limbs, as well as those that activate supporting musculature in the proximal ends of limbs. The medial pathways end in the medial ventral horn on the medial group of interneurons (Fig. 9-13). These interneurons connect bilaterally with motor neurons that control the axial musculature and thereby contribute to balance and posture. They also contribute to the control of proximal limb muscles. In this book we use the lateral/medial terminology to classify the descending motor pathways. However, even this scheme is not perfect, partly because although motor neuron cell bodies form localized columns, motor neuron dendritic trees are rather large and typically span most of the ventral horn. Thus, any motor neuron can potentially receive input from so-called medial or lateral system pathways.

Figure 9-13 Descending motor pathways. Major pathways connecting the cortical and brainstem motor areas to the spinal cord are shown. A, Lateral system pathways, corticospinal (red) and rubrospinal (blue) pathways. Note that the ventral corticospinal pathway is part of the medial system, but is shown in A for simplicity. B, Medial system pathways, medullary (blue) and pontine (green) reticulospinal and lateral vestibulospinal (red) pathways. C-B, corticobulbar; C-P, corticopontine; C-S, corticospinal; LCST, lateral corticospinal tract; VCST, ventral corticospinal tract.

Lateral Corticospinal and Corticobulbar Tracts

The corticospinal and corticobulbar tracts originate from a wide region of the cerebral cortex. This region includes the primary motor, premotor, supplementary, and cingulate motor areas of the frontal lobe and the somatosensory cortex of the parietal lobe. The cells of origin of these tracts include both large and small pyramidal cells of layer V of the cortex, including the giant pyramidal cells of Betz. Although Betz cells are a defining feature of the motor cortex, they represent a small minority (< 5%) of the cells that contribute to these tracts, in part because they are found only in the primary motor cortex, and even here they represent a minority of the cells contributing to the tract. These tracts leave the cortex and enter the internal capsule, then traverse the midbrain in the cerebral peduncle, pass through the basilar pons, and emerge as the pyramids on the ventral surface of the medulla (Fig. 9-13, A). The corticobulbar axons leave the tract as it descends the brainstem and terminate in the various cranial nerve motor nuclei. The corticospinal fibers continue caudally, and in the most caudal region of the medulla, about 90% of them cross to the opposite side. They then descend in the contralateral lateral funiculus as the lateral corticospinal tract. The lateral corticospinal axons terminate at all spinal cord levels, primarily on interneurons, but also on motor neurons. The remaining uncrossed axons continue caudally in the ventral funiculus on the same side as the ventral corticospinal tract, which belongs to the medial system. Many of these fibers ultimately decussate at the spinal cord level at which they terminate.

The lateral corticospinal tract is a relatively minor tract in lower mammals but becomes quantitatively and functionally very important in primates and in humans in particular, where it contains over 1 million axons. This number still represents a relatively small proportion of the outflow from the cortex because there are approximately 20 million axons in the cerebral peduncles. Nevertheless, the corticospinal pathway is critical for the fine independent control of finger movement inasmuch as isolated lesions of the corticospinal tract typically lead to a permanent loss of this ability, even though there is often recovery of other movement abilities with such lesions. Indeed, in primates, corticospinal synapses directly onto motor neurons are particularly prevalent for the motor neurons controlling finger muscles and are probably the basis of our ability to make independent, finely controlled finger movements.

The corticobulbar tract, which projects to the cranial nerve motor nuclei, has subdivisions that are comparable to the lateral and ventral corticospinal tracts. For example, part of the corticobulbar tract ends contralaterally in the portion of the facial nucleus that supplies muscles of the lower part of the face and in the hypoglossal nucleus. This component of the corticobulbar tract is organized like the lateral corticospinal tract. The remainder of the corticobulbar tract ends bilaterally.

Rubrospinal Tract

This tract originates in the magnocellular portion of the red nucleus, which is located in the midbrain tegmentum. These fibers decussate (cross) in the midbrain, descend through the pons and medulla, and then take up a position just ventral to the lateral corticospinal tract in the spinal cord. They preferentially affect motor neurons controlling distal musculature, similar to the corticospinal fibers. Red nucleus neurons receive input from the cerebellum and from the motor cortex, thus making this an area of integration of activity from these two motor systems.

Pontine and Medullary Reticulospinal Tracts

The cells that give rise to the pontine reticulospinal tract are in the medial pontine reticular formation. The tract descends in the ventral funiculus, and it ends on the ipsilateral medial group of interneurons. Its function is to excite motor neurons to the proximal extensor muscles to support posture.

The medullary reticulospinal tracts arise from neurons of the medial medulla, in particular the nucleus gigantocellularis. The tracts descend bilaterally in the ventral lateral funiculus, and they end mainly on interneurons associated with medial motor neuron cell groups. The function of the pathway is mainly inhibitory.

Lateral and Medial Vestibulospinal Tracts

The lateral vestibulospinal tract originates in the lateral vestibular nucleus, also known as Deiter’s nucleus. This tract descends ipsilaterally through the ventral funiculus of the spinal cord and ends on interneurons associated with the medial motor neuron groups. The lateral vestibulospinal tract excites motor neurons that supply extensor muscles of the proximal part of the limb that are important for postural control. In addition, this pathway inhibits flexor motor neurons because it also excites the reciprocal Ia interneurons that receive Ia input from extensor muscles, which therefore inhibits flexor motor neurons. The excitatory input to the lateral vestibular nucleus is from both the semicircular canals and the otolith organs, whereas the inhibitory input is from the Purkinje cells of the anterior vermis region of the cerebellar cortex. An important function of the lateral vestibulospinal tract is to assist in postural adjustments after angular and linear accelerations of the head.

The medial vestibulospinal tract originates from the medial vestibular nucleus. This tract descends in the ventral funiculus of the spinal cord to the cervical and midthoracic levels, and it ends on the medial group of interneurons. Sensory input to the medial vestibular nucleus from the labyrinth is chiefly from the semicircular canals. This pathway thus mediates adjustments in head position in response to angular acceleration of the head.

The Tectospinal Tract

The tectospinal tract originates in the deep layers of the superior colliculus. The axons cross to the contralateral side, just below the periaqueductal gray matter. They then descend in the ventral funiculus of the spinal cord to terminate on the medial group of interneurons in the upper cervical spinal cord. The tectospinal tract regulates head movement in response to visual, auditory, and somatic stimuli.

Primary Motor Cortex

The primary motor cortex (or just motor cortex) can be defined as the region of cortex from which movements are elicited with the least amount of electrical stimulation. It is essentially congruent with Brodmann’s cytoarchitectonic area 4 (Fig. 10-3). In humans it is located on the parts of the precentral gyrus that form the rostral wall of the central sulcus and the caudal half of the apex of the gyrus. Based on initial mapping studies, which were done with surface stimulation, the motor cortex was described as having a topographic organization that parallels that of the somatosensory cortex. The face, body, and upper limb were represented on the lateral surface with the face located inferiorly, near the lateral fissure, the torso most superiorly, and the lower extremity mostly on the medial aspect of the hemisphere. This somatotopic organization is often represented as a figurine or in a graphic form called a motor homunculus (Fig. 9-15, A). The distortion of the various body parts in the homunculus indicates approximately how much of the cortex is devoted to their motor control. This simple homunculus was likened to a piano keyboard and fit well with traditional conceptions of the motor cortex being the final cortical stage and acting as a relay for sending motor commands to the spinal cord.

Figure 9-15 Traditional and modern views of motor cortex musculotopic organization. A, Lateral view of the cerebrum showing a plane of section through the precentral gyrus (primary motor cortex) to obtain the section shown in B. B, Classic view of motor cortex musculotopy. C, Modern view of motor cortex organization in which each body part is represented multiple times across several discrete regions.

Beginning in the 1960s and 1970s, mapping studies began using microelectrodes inserted to the deep, or output, layers of the cortex to apply stimuli. With this technique, called intracortical microstimulation (ICMS), much lower stimulus intensities could be used to evoke movements and thus allowed higher-resolution mapping of the motor cortex, which revealed a much more complex topography than was previously imagined (Fig. 9-15, C). Movement about each joint was found to be evoked by many noncontiguous columns throughout wide regions of the motor cortex. Thus, cell columns related to movement about a particular joint are actually interspersed among columns that control movement about many other joints. In sum, although the motor cortex may have large subdivisions corresponding to a limb or the head, within each such area there is a complex intermingling of cell columns that control the muscles within that body part.

Such mixing of cell columns makes functional sense because most movement requires the coordinated action of muscles throughout a limb and most connectivity in the cortex is localized (i.e., axon collaterals that connect different cell columns are primarily confined to a 1- to 3-mm region surrounding the column from which they originate). Thus, by having multiple cell columns that control movement about a joint and intermixing them with columns controlling movement about other joints, multijoint movement can be generated as a whole.

Although the somatotopy of the motor cortex is in part anatomically determined by the topography of the corticospinal pathway, it is also a dynamic map. Axon collaterals link the different cell columns, so activity in one column could potentially lead to movement about multiple joints. In fact, this can happen, but these intercolumnar connections are modulated by inhibitory GABAergic interneurons. This was shown by locally blocking GABA in one region of the motor cortex and then stimulating the neighboring region. Before the block, stimuli evoked contractions of one set of muscles, but once inhibition was blocked, contractions were also evoked in muscles controlled by the region that was no longer inhibited (Fig. 9-16). Functional connections between cell columns can be controlled on a millisecond time scale, and depending on their state, the somatotopic map can be radically changed. Longer-term plastic changes are also known to occur; for example, the use (or disuse) of a body part can affect the size of its representation.

Figure 9-16 Dynamic nature of a motor cortex musculotopic map. Inhibitory GABAergic interneurons play an important role in shaping motor responses to stimulation of each region of the motor cortex. A, Schematic showing excitatory connections between two regions of primary motor cortex and local inhibitory neurons within a single region. B, Schematic of a rat brain indicating motor cortex regions where electrical stimuli were applied to evoke movements (Vib region) and bicuculline was applied to block GABAergic synapses (in the FL region). FL, forelimb; HL, hind limb; Vib, vibrissa. C, FL EMG records showing response to stimulation of the Vib region before, during the application of bicuculline, and after washing it out. Note that Vib stimulation evoked vibrissae movement in all conditions but evoked forelimb movement only when inhibitory interneurons were blocked.

(Data from Jacobs K, Donoghue J: Science 251:944, 1991.)

Supplementary Motor Area

The supplementary motor area (SMA) is located mainly on the medial surface of the hemisphere, just anterior to the primary motor cortex, and corresponds to the medial portion of Brodmann’s area 6 (Fig. 9-14). It is subdivided into two regions: the more posterior part is referred to as the SMA proper (or just SMA), and the anterior portion is called the pre-SMA. The SMA proper is similar to the other motor areas already listed: it contains a complete somatotopic map, it contributes to the corticospinal tract, and it is interconnected with the other motor areas. In contrast, the pre-SMA is not strongly connected with the other motor areas and spinal cord but rather is connected to the prefrontal cortex.

The results of stimulation studies show that as in the motor cortex, there is a complete somatotopic map in the SMA. Stimulation of the SMA can evoke isolated movement about single joints, similar to that after stimulation of the motor cortex, but higher-intensity and longer-duration stimulation is necessary; moreover, the evoked movements are often more complex than those evoked by stimulation of the motor cortex. However, longer-duration stimulation of the primary motor cortex can also evoke complex, apparently purposeful movement sequences, so the distinction is not absolute. In addition, stimulation of the SMA can produce vocalization or complex postural movements, but it can also have the opposite result, namely, a temporary arrest of movement or speech. Removal of the supplementary motor cortex retards movement of the opposite extremities and may result in forced grasping movements with the contralateral hand.

Premotor Area

This area lies rostral to the primary motor cortex and is contained in Brodmann’s area 6 on the lateral surface of the brain (Fig. 9-14). It can be distinguished from the primary motor cortex by the higher stimulus intensities needed to evoke movement. The premotor area has been divided into two functionally distinct subdivisions: dorsal and ventral. Like the motor cortex, both subdivisions are somatotopically organized and both contribute to the corticospinal tract. The dorsal division (PMd) contains a relatively complete map representing the leg, trunk, arm, and face. In contrast, the somatotopic map of the ventral division (PMv) is mostly limited to the arm and face, with only a small leg representation. Thus, the PMv appears to be specialized for control of upper limb and head movement. A second difference between the subdivisions is that PMd contains a large representation of the proximal muscles, whereas PMv has a large representation of the distal muscles.

Cingulate Motor Areas

These motor areas are located within the cingulate sulcus at approximately the same anterior-posterior level as the SMA. There are three cingulate motor areas (dorsal, ventral, and rostral) (Fig. 9-14, B). Each contains a somatotopic map and contributes to the corticospinal tract. Microstimulation in these areas evokes movement similar to that evoked by motor cortex stimulation, except that once again, higher stimulus intensities are needed. Single-cell recordings during movements have shown that the spontaneous activity of neurons in the cingulate motor areas is related to the preparation and execution of movements.

Afferent Systems

There are two major cerebellar afferent systems: mossy fibers and climbing fibers. Mossy fibers are named for their distinctive appearance in the cerebellar cortex: as a mossy fiber courses through the granule layer, on occasion it swells and sends out a bunch of short twisty branchlets. These entities are called rosettes and are points of synaptic contact between these fibers and neurons in the granule cell layer. Mossy fibers arise from many sources, including the spinal cord (the spinocerebellar pathways), dorsal column nuclei, trigeminal nucleus, nuclei in the reticular formation, primary vestibular afferents, vestibular nuclei, cerebellar nuclei, and the basilar pontine nuclei. The details of mossy fiber projection patterns are beyond the scope of the present chapter; however, several general points are worth noting:

In contrast to the diverse origins of mossy fibers, climbing fibers all originate from a single nucleus: the inferior olive, which is located in the rostral medulla, just dorsal and lateral to the pyramids. The olivary neurons are almost all projection cells whose axons leave the nucleus without giving off collaterals and then cross the brainstem to enter the cerebellum primarily via the inferior cerebellar peduncle. Like mossy fibers, olivocerebellar axons are excitatory and send collaterals to the cerebellar nuclei as they ascend through the cerebellar white matter to the cortex. In the cerebellar cortex, olivocerebellar axons may synapse with basket, stellate, and Golgi cells but form a special synaptic arrangement with Purkinje cells. Each Purkinje cell receives input from only a single climbing fiber, which “climbs” up its proximal dendrites and makes hundreds of excitatory synapses. Note, the terminal portion of the olivocerebellar axon is referred to as a climbing fiber. Conversely, each olivary axon will branch to form about 10 to 15 climbing fibers.

The inferior olive is a distinctive brain region for several reasons. As already noted, its neurons are virtually all projection cells, so there is little local chemical synaptic interaction between the cells. Instead, olivary neurons are electrically coupled to each other by gap junctions. In fact, the olive has the highest density of neuronal gap junctions in the CNS. This allows olivary neurons to have synchronized activity that gets transmitted to the cerebellum. Afferents to the olive may be divided into two main classes, excitatory input, which arises from many regions throughout the CNS, and inhibitory GABAergic input from the cerebellar nuclei and a few brainstem nuclei. Although these afferents can modulate the firing rates of olivary neurons (as is typical in most brain regions), the membrane conductance of olivary neurons limits this modulation to a range of a few hertz and endows these neurons with the potential to be intrinsic oscillators. Instead of just modulating firing rates, olivary afferent activity acts to modify the effectiveness of the electrical coupling between olivary neurons and thus changes the patterns of synchronous activity delivered to the cerebellum. Afferent activity may also modulate expression of the oscillatory potential of olivary neurons. Thus, the inferior olive appears to be organized to generate patterns of synchronous activity across the cerebellar cortex. The functional significance of these patterns remains controversial. One hypothesis is that they provide a gating signal for synchronizing motor commands to various muscle combinations.

Cellular Elements and Efferents of the Cortex

Despite its enormous expansion throughout vertebrate evolution, the basic anatomic organization of the cerebellar cortex has remained nearly invariant. The circuitry is also among the most regular and stereotyped of any brain region. The cerebellar cortex contains eight different neuronal types: Purkinje cells, Golgi cells, granule cells, Lugaro cells, basket cells, stellate cells, unipolar brush cells, and candelabrum cells. These cells are found in all regions of the cerebellar cortex, with the exception of unipolar brush cells, which are limited mainly to cerebellar areas receiving vestibular input (i.e., the flocculonodular lobe). These eight cell types are distributed among the three layers that make up the cerebellar cortex of higher vertebrates. The outer or superficial layer is the molecular layer (Fig. 9-20, A). Stellate and basket cells are found here. The deepest layer is the granule cell layer. This layer has the highest cellular density in the nervous system and contains granule, Golgi, and unipolar brush cells. Separating the molecular and granule cell layers is the Purkinje cell layer, formed by Purkinje cell somata, which are arranged as a one-cell-thick sheet of cells. Candelabrum cells are also located in this layer. Lugaro cells are situated slightly deeper at the upper border of the granule cell layer.

The sole efferent from the cortex is the Purkinje cell axon, which also has local collaterals and is GABAergic and inhibitory. Thus, the remaining seven cell types are local interneurons. Of these, the stellate, basket, Golgi, Lugaro, and candelabrum cells are also inhibitory GABAergic neurons, whereas the granule and unipolar brush cells are excitatory.

Microcircuitry of the Cortex

The dendrites, axons, and patterns of synaptic connections of most neurons within the cerebellar cortex are organized with respect to the transverse (short) and longitudinal (long) folial axes (Fig. 9-21). In the vermis, where the folia run perpendicular to the sagittal plane, these axes lie in the sagittal and coronal planes, respectively. In the hemispheres, where the folia are oriented at various angles with respect to the sagittal plane, this correspondence is lost, and the local folial axes must then serve as the reference axes.

Figure 9-21 Three-dimensional view of the cerebellar cortex showing some of the cerebellar neurons. The cut face at the left is along the long axis of the folium; the cut face at the right is at right angles to the long axis. BC, basket cell; CF, climbing fiber; CN, cerebellar nuclear cell; GC, Golgi cell; Glm, glomerulus; GrC, granule cell; MF, mossy fiber; PC, Purkinje cell; PF, parallel fiber; SC, stellate cell.

The Purkinje cell dendritic tree is the largest in the CNS. It extends from the Purkinje cell layer through the molecular layer to the surface of the cerebellar cortex and for several hundred microns along the transverse axis of the folium, but only 30 to 40 μm in the longitudinal direction. Thus, it is like a flat pancake that lies in a plane parallel to the transverse folial axis. Accordingly, a set of Purkinje cell dendritic trees can be thought of as a stack of pancakes, with the stack running along the longitudinal folial axis.

The dendritic trees of the molecular layer interneurons (stellate and basket cells) are oriented similar to the Purkinje cell dendritic tree, although they are much less extensive. The axons of stellate and basket cells run transversely across the folium and form synapses with Purkinje cells. Stellate and basket cells synapse onto Purkinje cell dendrites. In addition, basket cells make synapses on the Purkinje cell soma and form a basket-like structure around the base of the soma, which gives the basket cell its name.

Granule cells are small neurons with four to five short unbranched dendrites, each ending in a claw-like expansion that synapses with a mossy fiber rosette and with terminals from Golgi cell axons in a complex arrangement known as a glomerulus. The axons of granule cells ascend through the Purkinje cell layer to the molecular layer, where they bifurcate and form parallel fibers. The parallel fibers run parallel to the cerebellar surface along the longitudinal axis of the folium (perpendicular to the planes of the Purkinje, stellate, and basket cell dendritic trees) and form excitatory synapses with the dendrites of the Purkinje, Golgi, stellate, and basket cells.

The orthogonal relationship between the parallel fibers and the dendritic trees of the Purkinje cells and molecular layer interneurons (basket and stellate cells) has significant functional consequences. This arrangement allows maximal convergence and divergence to occur. A single parallel fiber, which can be up 6 mm long, will pass through more than 100 Purkinje cell dendritic trees (and also interneuron dendrites); however, it has the chance to make only one or two synapses with any particular cell because it crosses through the short dimension of the dendritic tree. Conversely, a given Purkinje cell receives synapses from about 200,000 parallel fibers. Thus, a beam of parallel fibers can be excited experimentally, which will excite a row of Purkinje cells and interneurons that are in line with this beam (Fig. 9-22). In addition, because the axons of the interneurons run perpendicular to the parallel fibers, this beam of excitation will be flanked by inhibition. Although this classic electrophysiological experiment clearly demonstrates the functional connectivity of the cerebellar cortex, whether such beams of excitation occur normally remains a controversial question.

Figure 9-22 Functional connectivity of the cerebellar cortex. The geometry of the cerebellar cortical circuits makes electrophysiological determination of the functional connectivity of the cellular elements possible. The figure shows a classic paradigm in which stimulation of the cerebellar cortex activates a beam of parallel fibers (brown). Recordings from the stellate and basket cells (green cells) and Purkinje cells (PCs) (orange cells) in line with this beam show that they are excited by the parallel fibers. In contrast, Purkinje cells located rostral or caudal to the beam receive only inhibition (purple areas) as a result of the perpendicular spatial relationship of the parallel fibers and the stellate and basket cell axons.

The Golgi cells are inhibitory interneurons in the granule cell layer. The geometry of their axonal and dendritic arbors is an exception to the orthogonal and planar organization of the cortex in that their dendrites and axons carve out roughly conical territories. One can think of it as two cones, tip to tip, where the soma is at the point at which the two cone tips meet. The dendritic tree forms the upper cone, which often extends into the molecular layer, and the axon forms the lower one. Golgi cells are excited by mossy and climbing fibers and by granule cell axons (parallel fibers) and inhibited by basket, stellate, and Purkinje cell axon collaterals. They in turn inhibit granule cells. Thus, they participate in both feedback (when excited by parallel fibers) and feedforward (when excited by mossy fibers) inhibitory loops that control activity in the mossy fiber—parallel fiber pathway to the Purkinje cell.

Lugaro cells have fusiform somata from which emerge two relatively unbranched dendrites, one from each side, that run along the transverse folial axis for several hundred microns, usually just under the Purkinje cell layer. Purkinje cell axon collaterals provide the main input to these neurons, with granule cell axons adding a minor input. The axon terminates mainly in the molecular layer on basket, stellate, and possibly Purkinje cells. Thus, these cells appear to sample the activity of Purkinje cells and provide both positive-feedback signals (they inhibit the interneurons that inhibit Purkinje cells) and negative-feedback signals (they directly inhibit the Purkinje cell).

Unipolar brush cells have only a single dendrite that ends as a tight bunch of branchlets that resemble a brush. These cells receive excitatory input from mossy fibers and inhibitory input from Golgi cells. It is thought that they synapse with granule and Golgi cells, which would make these cells an excitatory feedforward link in the mossy fiber—parallel fiber pathway.

Candelabrum cells are GABAergic cells located in the Purkinje layer. Their dendrites and axons terminate in the molecular layer, where the axonal arborization pattern resembles a candelabrum.

Activity of Purkinje Cells in the Cerebellar Cortex in the Context of Motor Coordination

Mossy fiber input to the cerebellar cortex, via their excitation of granule cells, causes a Purkinje cell to discharge single action potentials, referred to as simple spikes (Fig. 9-23). The spontaneous simple spike firing rate is typically around 20 to 50 Hz but can vary widely (from 0 to > 100 Hz), depending on the relative balance of excitation from parallel fiber input and inhibition from cerebellar cortex interneurons. Thus, this activity reflects the state of the cerebellar cortex.

Figure 9-23 Responses of a Purkinje cell to excitatory input recorded extracellularly. A, Granule cells, via their ascending axons and parallel fibers, excite Purkinje cells and trigger simple spikes. B, Climbing fiber activity leads to high-frequency (≈500 Hz) bursts of two to four spikes known as complex spikes in Purkinje cells.

In contrast, a climbing fiber discharge causes a high-frequency burst of action potentials, called a complex spike (Fig. 9-23), in an all-or-none manner because of the massive excitation that is provided by the single climbing fiber to a Purkinje cell. This excitation is so powerful that there is essentially a one-to-one relationship between climbing fiber discharge and a complex spike. Thus, complex spikes essentially override what is happening at the cortex level and reflect the state of the inferior olive. The average firing rate of a spontaneous complex spike is only about 1 Hz.

Because the climbing fibers generate complex spikes at such a low frequency, they do not substantially change the average firing rates of Purkinje cells, and consequently, it is commonly argued that they have no direct role in shaping the output of the cerebellar cortex and are therefore not involved in ongoing motor control. Instead, it is commonly thought that their function is to alter the responsiveness of Purkinje cells to parallel fiber input. In particular, under certain circumstances, complex spike activity produces a prolonged depression in parallel fiber synaptic efficacy, termed LTD (long-term depression). This phenomenon is the proposed mechanism by which climbing fibers act in motor-learning hypotheses. Such hypotheses typically state that the parallel fiber system, and hence simple spikes, are involved in generating ongoing movement and, when there is a mismatch between the intended and actual movement, this error activates the inferior olive and complex spikes result, which then lead to LTD of the active parallel fiber synapses. This adjustment in synaptic weight will change the motor output in the future. If this change results in a properly executed movement, activation of the inferior olive will not occur and the motor program will be unchanged, but if there is still an error, the olivocerebellar system will trigger additional complex spikes that will cause further changes in synaptic efficacy, and so on. Major challenges to this view are that motor learning can occur when LTD is chemically blocked and that learned behavior can remain after removal of portions of the cerebellum where the memory is supposedly stored.

An alternative view is that the olivocerebellar system is directly involved in motor control (note that this does not preclude a role in motor learning as well) and, in particular, helps in the timing of motor commands. This view follows from the types of motor deficits observed in cerebellar damage and makes use of the special properties of the inferior olive mentioned earlier, namely, that it can generate rhythmic synchronous complex spike discharges across populations of Purkinje cells. These complex spikes would then produce synchronized inhibitory postsynaptic currents (IPSPs) on cerebellar nuclear neurons as a result of the convergence present in the Purkinje cell axon to the cerebellar nuclear projection. Because of the membrane properties of cerebellar nuclear neurons, these synchronized IPSPs could have a qualitatively different effect on nuclear cell firing than would the IPSPs caused by more numerous, but asynchronous simple spikes. Specifically, they could trigger rebound bursts in the nuclear cells that would then be transmitted to other motor systems as a gating signal. In fact, voluntary movements appear to be composed of a series of periodic accelerations that reflect a central oscillatory process. However, whether the olivocerebellar system helps time motor commands requires further evidence.

Direct Pathway

The overall action of the direct pathway through the basal ganglia to motor areas of the cortex is to enhance motor activity. In the direct pathway, the striatum projects to the internal segment of the globus pallidus (GPi). This projection is inhibitory, and the main transmitter is GABA. The GPi projects to the VA and VL nuclei of the thalamus. These connections also use GABA and are inhibitory. The VA and VL nuclei send excitatory connections to the prefrontal, premotor, and supplementary motor cortex. This input to the cortex influences motor planning, and it also affects the discharge of corticospinal and corticobulbar neurons.

The direct pathway appears to function as follows. Neurons in the striatum have little background activity, but during movement they are activated by their input from the cortex. In contrast, neurons in the GPi have a high level of background activity. When the striatum is activated, its inhibitory projections to the globus pallidus slow the activity of pallidal neurons. However, the pallidal neurons themselves are inhibitory, and they normally provide tonic inhibition of neurons in the VA and VL nuclei of the thalamus. Therefore, activation of the striatum causes disinhibition of neurons of the VA and VL nuclei. When disinhibited, the VA/VL neurons increase their firing rates, exciting their target neurons in the motor areas of the cerebral cortex. Because the motor areas evoke movement by activating α and γ motor neurons in the spinal cord and brainstem, the basal ganglia can regulate movement by enhancing the activity of neurons in the motor cortex.

Indirect Pathway

The overall effect of the indirect pathway is to reduce the activity of neurons in motor areas of the cerebral cortex. The indirect pathway involves inhibitory connections from the striatum to the external segment of the globus pallidus (GPe), which in turn sends an inhibitory projection to the subthalamic nucleus and to GPi. The subthalamic nucleus then sends an excitatory projection back to the GPi (Fig. 9-25, A).

In this pathway, pallidal neurons in the external segment are inhibited by the GABA released from striatal terminals in the globus pallidus. The GPe normally releases GABA in the subthalamic nucleus and thereby inhibits the subthalamic neurons. Therefore, striatal inhibition of the GPe results in the disinhibition of neurons of the subthalamic nucleus. The subthalamic neurons are normally active, and they excite neurons in the GPi by releasing glutamate. When the neurons of the subthalamic nucleus become more active because of disinhibition, they release more glutamate in the GPi. This transmitter excites neurons in the GPi and consequently activates inhibitory projections that affect the VA and VL thalamic nuclei. The activity of the thalamic neurons consequently decreases, as does the activity of the cortical neurons that they influence.

The direct and indirect pathways thus have opposing actions; an increase in the activity of either one of these pathways might lead to an imbalance in motor control. Such imbalances, which are typical of basal ganglion diseases, may alter the motor output of the cortex.

Vestibuloocular Reflex

Eye movement probably first evolved to hold the eye still, in contrast to limb movement, where one typically wants to generate movement with respect to the external world. The reason is that visual acuity degrades rapidly when there is eye movement relative to the external world (i.e., the visual scene slips across the retina). A major cause of such slippage is movement of the head. The vestibuloocular reflex (VOR) is one of the main mechanisms by which head movement is compensated in order to allow a stable visual scene to be maintained on the retina.

To maintain a stable visual scene on the retina, the VOR produces movement of the eyes that is equal and opposite the movement of the head. This reflex is initiated by stimulation of the receptors (hair cells) in the vestibular system (see Chapter 8). Recall that the vestibular organs are sensitive to head acceleration, not visual cues, and thus the VOR occurs in both the light and dark. Functionally, it is what is called an open loop system in that it generates an output (eye movement) in response to a stimulus (head acceleration), but its immediate behavior is not regulated by feedback about the success or failure of its output. It is worth noting, however, that in the light at least, any failure by the VOR to match eye and head rotation will result in what is called retinal slip (i.e., slip of the visual image across the retina), and this error signal can be fed back to the VOR circuits by other neuronal pathways and over time can lead to adjustments in the strength of the VOR to eliminate the error. This adaptation of the VOR is a major model for studying plasticity in the brain.

As stated, acceleration signals initiate the VOR. The output of the VOR, however, must be a change in eye position in the orbit. Thus, the problem to be solved by the nervous system is to translate the acceleration signals sensed by the vestibular organs into correct positional signals for the eyes. Mathematically, this can be thought of as a double integration. The first integration is done by the vestibular receptor apparatus because although the hair cells respond to head acceleration, the signals in the vestibular afferents are proportional to head velocity (at least for most stimuli that are encountered physiologically). The second integration, from velocity to position, occurs in the CNS in circuits described later.

The head can move in six different ways, often referred to as six degrees of freedom: three translational and three rotational. To compensate for these different types of movement, there are both translational and angular VORs, as well as separate subsystems for handling movement about different directions (e.g., rotation about a vertical or a horizontal axis).

Optokinetic Reflex

The optokinetic reflex (OKR) is a second mechanism by which the nervous system stabilizes the visual scene on the retina, and it often works in conjunction with the VOR. Whereas the VOR is activated only by head motion, the OKR is activated by movement of the visual scene, whether caused by motion of the scene itself or by head motion. That is, the sensory stimulus for this reflex is slip of the visual scene on the retina as detected by motion-sensitive retinal ganglion cells. An example of the former is when you are sitting in a train and a train on the adjacent track begins moving, your eyes rotate to keep the image of the neighboring car stable. This often leads to a sensation that you are moving (this is not entirely surprising because OKR circuits feed into the same circuits as used by the vestibular system).

The OKR can work in conjunction with the VOR to stabilize the visual image and is particularly important for maintaining a stable image when head movements are slow because the VOR works poorly in these conditions. In addition, as mentioned earlier, the VOR circuits by themselves act in an open loop mode and thus have no way to correct errors or calibrate their performance (i.e., detect a mismatch between head and eye rotation). The OKR allows for corrections and for calibration by triggering mechanisms to adjust the sensitivity of the VOR. Such mismatches occur as the head grows during childhood or when one puts on glasses.

Saccades

In animals whose eyes have a fovea, it becomes particularly advantageous to be able to move the eye with respect to the world (i.e., the main visual scene) so that objects of importance can be focused onto the fovea and scrutinized with this high-resolution part of the retina. Two classes of eye movement underlie this ability: saccadic and smooth pursuit. Movements that bring a particular region of the visual world onto the fovea are called saccades. For example, to read this sentence you are making a series of saccades to bring successive words onto your fovea to be read. However, even afoveate animals make saccades, and thus saccades may also be used to rapidly scan the visual environment.

Saccades are extremely rapid eye movements. In humans, eye velocity during a saccade can reach 800 degrees/sec, as compared with movement velocity of less than 10 degrees/sec generated in response to typical VOR and OKR stimuli (velocities of up to ≈120 degrees/sec can be produced by OKR stimuli in humans; however, they are still much slower than the maximal saccade velocities). Saccades can be made voluntarily or reflexively. Moreover, although they are usually made in response to visual targets, they can also be made toward auditory or other sensory cues, in the dark, or toward memorized targets.

Interestingly, visual processing appears to be suppressed just before and during saccades, particularly in the magnocellular visual pathway that is concerned with visual motion. This phenomenon is known as saccadic suppression and may function to prevent sensations of sudden, rapid movement of the visual world that would result during a saccade in the absence of such suppression. The mechanisms underlying saccadic suppression are not fully known, but in areas of the cortex related to visual processing, the responsiveness of the cells to visual stimuli is reduced and altered during saccades.

Smooth Pursuit

Once a saccade has brought a moving object of interest onto the fovea, the smooth pursuit system allows us to keep it stable on the fovea despite its continued motion. This ability appears to be limited to primates and allows prolonged continuous observation of a moving object. Note that in some respects, smooth pursuit might seem similar to the OKR, and in fact there may not be an absolute dividing line because as the target size grows, the distinction between target and background is lost; however, for small moving targets, smooth pursuit requires suppression of the OKR. You can see the effect of this suppression by moving your finger back and forth in front of this text while tracking it with your eyes. Your finger will be in focus while the words on this page will be part of the background scene and will become illegible as they slip along your retina.

Nystagmus

When there is a prolonged OKR or VOR stimulus (e.g., if you keep turning in one direction), these reflexes will initially counterrotate the eyes in an attempt to maintain a stable image on the retina, as described earlier. However, with a prolonged stimulus, the eyes will reach their mechanical limit, no further compensation will be possible, and the image will begin to slip on the retina. To avoid this situation, a fast saccade-like movement of the eyes occurs in the opposite direction, essentially resetting the eyes to begin viewing the visual scene again. Then the slow OKR- or VORinduced counterrotation will start anew. This alternation of slow and fast movement in opposite directions is nystagmus and can be displayed on a nystagmogram (Fig. 9-26). Thus, nystagmus can be defined as oscillatory or rhythmic movements of the eye in which there is a fast and a slow phase. The nystagmus is named according to the direction of the fast phase, because it is more easily observed.

Figure 9-26 Nystagmogram showing eye movements that occur during nystagmus. The plot shows a left nystagmus because the fast phase is directed toward the left (downward on the graph).

In addition to being induced physiologically by VOR or OKR stimuli, nystagmus can result from damage to the vestibular circuits, either in the periphery (e.g., VIII nerve) or centrally (e.g., vestibular nuclei), and can be a useful diagnostic symptom.

Vergence

Conjugate eye movement is movement of both eyes in the same direction and in an equal amount. Such coordination allows a target to be maintained on both fovea during eye movement and is necessary to maintain binocular vision without experiencing diplopia (double vision). However, when objects are close (< 30 m), maintaining a target on both fovea requires eye movement that is no longer identical. Such disjunctive or vergence movements are also necessary for fixation of both eyes on objects that are approaching or receding. During convergence movements, accommodation of the lens for near vision and pupillary constriction also occur. In sum, the stimuli for vergence movements are diplopia and blurry images.

Motor Neurons of the Extraocular Muscles

Three cranial nerve nuclei supply the extraocular muscles: oculomotor, trochlear, and abducens nuclei. Note that we will sometimes refer to these three nuclei collectively as the oculomotor nuclei; however, the context should make clear whether we mean the specific nucleus or all three. Motor neurons for the ipsilateral medial and inferior recti, ipsilateral inferior oblique, and contralateral superior rectus muscles reside in the oculomotor nucleus; those for the contralateral superior oblique muscle reside in the trochlear nucleus; and those for the ipsilateral lateral rectus muscle are located in the abducens nucleus. These motor neurons form some of the smallest motor units (1 : 10 nerve-to-muscle ratio), consistent with the very fine control needed for precise eye movement.

An important point regarding motor neurons to the extraocular muscles is that most have spontaneous activity when the eye is in the primary position (looking straight ahead) and their firing rate correlates with eye position and velocity. This spontaneous activity allows the antagonist muscle pairs to act in a push-pull fashion, which increases the responsiveness of the system. That is, as motor neurons innervating one muscle are activated and cause increased contraction, those to its antagonist are inhibited and lead to relaxation.

In addition to motor neurons, the abducens nuclei have internuclear neurons. These neurons project, via the medial longitudinal fasciculus, to medial rectus motor neurons in the contralateral oculomotor nucleus. As we will see, this projection facilitates the coordinated action of the medial and lateral recti that is needed for conjugate movements, such as occur in the VOR.

Circuits Underlying the Vestibuloocular Reflex

The VOR acts to counter head motion by causing rotation of the eyes in the opposite direction. There are separate circuits for rotational and translational movement of the head. The sensors for the former are the semicircular canal, and the sensors for the latter are the otoliths (the utricle and saccule). The circuits for the angular VOR are more straightforward (but still complex!), so we will focus on these pathways to illustrate how this reflex works; however, the basic scheme is the same: vestibular afferents go to vestibular nuclei, the vestibular nuclei in turn project to the various oculomotor nuclei, and motor neurons in the oculomotor nuclei give rise to axons that innervate the extraocular muscles. What varies are the specific vestibular and oculomotor nuclei that are involved.

Focusing on the angular VOR pathways, the pathway for generating horizontal eye movement originates in the horizontal canals, and the analogous one for vertical movement originates in the anterior and posterior canals. Figure 9-27, A, shows the basic circuit for the horizontal VOR. Note that only the major central circuits originating in the left horizontal canal and vestibular nuclei are shown; however, mirror image pathways arise from the right canal and vestibular nuclei. Vestibular afferents involved in the horizontal VOR pathway primarily synapse in the medial vestibular nucleus, which projects to the abducens nucleus bilaterally; inhibitory neurons project ipsilaterally and excitatory ones project contralaterally. Control of the medial rectus muscle is achieved by abducens internuclear neurons that project from the abducens to the part of the oculomotor nucleus controlling the medial rectus muscle. Note the double crossing of this pathway, which results in aligning of the responses of functional synergists (e.g., the left medial rectus with the right lateral rectus).

Figure 9-27 Circuits underlying the horizontal vestibuloocular reflex (VOR). A, The vestibular nuclei receive excitatory input from the horizontal canal afferents and project to the abducens (VI) nucleus. The VI nucleus innervates the lateral rectus and projects to the contralateral oculomotor (III) nucleus, which controls the medial rectus. Excitatory neurons are shown in red, inhibitory ones in blue. Note that only the major pathways originating in the left vestibular nuclei are shown. For clarity, only the beginnings of mirror image pathways from the right vestibular nuclei are shown (dotted lines). B, Flow of activity in the VOR circuitry induced by leftward head rotation. Increased soma size and axonal thickness indicate increased activity; thinner axons indicate decreased activity in comparison to levels at rest (A). Note that leftward rotation causes both an increase in activity of the left vestibular afferents and a decrease in activity of the right ones. MLF, medial longitudinal fasciculus. Vestibular nuclei: I, inferior; L, lateral; M, medial; S, superior.

The vertical VOR pathway primarily involves the superior vestibular nucleus, which has direct bilateral projections to the oculomotor nucleus.

Consider what happens in the horizontal canal pathway when there is head rotation to the left as shown in Figure 9-27, B. Leftward head rotation would cause the visual image to slip to the right. However, compensation by the VOR will be triggered by depolarization of the hair cells of the left canal in response to the angular acceleration (Fig. 8-27). The depolarized hair cells will cause increased activity in the left vestibular afferents and thereby excite neurons of the left medial vestibular nucleus. These include excitatory neurons that project to the contralateral abducens nucleus and synapse with both motor neurons and internuclear neurons. Excitation of the motor neurons will lead to contraction of the right lateral rectus and rotation of the right eye to the right, whereas excitation of the internuclear neurons of the right abducens nucleus will lead to excitation of the medial rectus motor neurons in the left oculomotor nucleus, thus causing the left eye to rotate to the right as well.

If we now follow the pathway starting with the inhibitory vestibular neurons that project from the left medial vestibular nucleus to the ipsilateral abducens nucleus, we can see that the activity of these cells leads to inhibition of motor neurons to the left lateral rectus and motor neurons to the right medial rectus (the latter via internuclear neurons to the right oculomotor nucleus). Consequently, these muscles will relax, thereby facilitating rotation of the eyes to the right. Thus, the eye is being pulled by the increased tension of one set of muscles and “pushed” by the release of tension in the antagonist set of muscles.

Note that the mirror image pathways originating from the right canal have been left out of Figure 9-27 for clarity, but the changes in activity through them with leftward head rotation would be exactly the opposite, and thus they would function synergistically with those that are shown. As an exercise the reader should work out the resulting changes in activity through these circuits. Remember that leftward head rotation hyperpolarizes the hair cells of the right canal, thereby leading to a decrease in right vestibular afferent activity and disfacilitation of the right vestibular nuclear neurons.

Now, consider the commissural fibers that connect the two medial vestibular nuclei. These fibers are excitatory but end on local inhibitory interneurons of the contralateral vestibular nucleus and thus inhibit the projection neurons of that nucleus. This pathway reinforces the actions of the contralateral vestibular afferents on their target vestibular nuclear neurons. In our example, commissural cells in the left vestibular nucleus will be activated and therefore cause active inhibition of the right medial vestibular nuclei projection neurons, which reinforces the disfacilitation caused by the decrease in right afferent activity. In fact, this commissural pathway is powerful enough to modulate the activity of the contralateral vestibular nuclei even after unilateral labyrinthectomy, which destroys the direct vestibular afferent input to these nuclei.

Finally, it is important to note that superimposed on the brainstem circuits is the cerebellum. Parts of the vermis and flocculonodular lobe receive primary vestibular afferents or secondary vestibular afferents (axons of the vestibular nuclear neurons), or both, and in turn project back to the vestibular nuclei directly and via a disynaptic pathway involving the fastigial nucleus. The exact role of these cerebellar circuits in generating the VOR is much debated, but they are critical inasmuch as damage to them leads to abnormal eye movement, such as spontaneous nystagmus, and other symptoms of vestibular dysfunction.

Circuits Underlying the Optokinetic Reflex

The stimulus eliciting the OKR is visual (retinal slip), so photoreceptors are the start of the reflex arc. Key brainstem centers for this reflex lie in the tegmentum and pretectal region of the rostral midbrain. They are the nucleus of the optic tract (NOT) and a group of nuclei collectively known as the accessory optic nuclei (AON). Direction-selective, motion-sensitive retinal ganglion cells are a major afferent source carrying visual information to these nuclei. In addition, input comes from primary and higher-order visual cortical areas in the occipital and temporal lobes. These latter afferent sources become particularly important in primates and humans. Cells of the NOT and AON have large receptive fields, and their responses are selective for the direction and speed of movement of the visual scene. Interestingly, the preferred directions of movement of the NOT/AON cells correspond closely to motion caused by rotation about axes perpendicular to the semicircular canals, thereby facilitating coordination of the VOR and OKR to produce stable retinal images. The efferent connections of these nuclei are numerous and complex and not fully understood. There are polysynaptic pathways to the oculomotor and abducens nuclei and monosynaptic input to the vestibular nuclei, which allows interaction with the VOR. There are projections to various precerebellar nuclei, including the inferior olive and basilar pontine nuclei. These pathways then loop through the flocculus and back to the vestibular nuclei. In sum, via several pathways operating in parallel, activity ultimately arrives at the various oculomotor nuclei whose motor neurons are activated, and proper counterrotation of the eyes results.

Circuits Underlying Saccades

Saccades are generated in response to activity in the superior colliculus or the cerebral cortex (frontal eye fields and posterior parietal areas). Activity in the superior colliculus is related to computation of the direction and amplitude of the saccade. Indeed, the deep layers of the superior colliculus contain a topographic motor map of saccade locations. From the superior colliculus, information is forwarded to distinct sites for control of horizontal and vertical saccades, referred to as the horizontal and vertical gaze centers, respectively. The horizontal gaze center consists of neurons in the paramedian pontine reticular formation (PPRF), in the vicinity of the abducens nucleus (Fig. 9-28, A). The vertical gaze center is located in the reticular formation of the midbrain, specifically, the rostral interstitial nucleus of the medial longitudinal fasciculus and the interstitial nucleus of Cajal. Because the circuitry and operation of the horizontal gaze center are better understood than those of the vertical gaze center, it is discussed here in detail. However, cells showing similar activity patterns have been described in the vertical gaze center.

Figure 9-28 Horizontal saccade pathways. A, Circuit diagram of the major pathways. EBN, excitatory burst neuron; FEF, frontal eye field; IBN, inhibitory burst neuron; LBN, long lead burst neuron; OPN, omnipause neuron; PPRF, paramedian pontine reticular formation. B, Firing patterns of some of the neurons involved in making saccades. Excitation of burst neurons of the right horizontal gaze center causes abducens motor neurons on the right and medial rectus motor neurons on the left to be activated. The ascending pathway to the oculomotor nucleus is through the medial longitudinal fasciculus. The left horizontal gaze center is simultaneously inhibited.

Figure 9-28, A is an overview of the neural circuitry by which saccades are generated, and Figure 9-28, B shows the activity of certain types of neurons found in the gaze center that are responsible for horizontal saccades. Each horizontal gaze center has excitatory burst neurons that project to motor neurons in the ipsilateral abducens nucleus and to the internuclear neurons (which will excite medial rectus motor neurons in the contralateral oculomotor nucleus). It also has inhibitory burst neurons that inhibit the contralateral abducens. These burst neurons are capable of extremely high bursts of spikes (up to 1000 Hz). Moreover, the gaze center has neurons showing tonic activity and burst-tonic activity.

Normally, both inhibitory and excitatory burst neurons are inhibited by omnipause neurons located in the nucleus of the dorsal raphe. When a saccade is to be made, activity from the frontal eye fields or the superior colliculus, or both, leads to inhibition of the omnipause cells and excitation of the burst cells on the contralateral side. The resulting high-frequency bursts in the excitatory burst neurons provide a powerful drive to motor neurons of the ipsilateral lateral rectus and contralateral medial rectus (Fig. 9-28, A) while at the same time, inhibitory burst neurons permit relaxation of the antagonists. The initial bursts of these neurons allow strong contraction of the appropriate extraocular muscles, which overcomes the viscosity of the extraocular muscle, and permits rapid movement to occur.

Circuits Underlying Smooth Pursuit

Smooth pursuit involves tracking a moving target with one’s eyes (Fig. 9-29). Visual information about target velocity is processed in a series of cortical areas, including the visual cortex in the occipital lobe, several temporal lobe areas, and the frontal eye fields. It should be noted that in the past it was thought that the frontal eye fields were related only to control of saccades, but evidence has recently shown that there are distinct regions within the frontal eye fields dedicated to either saccade production or smooth pursuit. Indeed, there may be two distinct cortical networks, each specialized for one of these types of eye movement. Cortical activity from multiple cortical areas is fed to the cerebellum via parts of the pontine nuclei and nucleus reticularis tegmenti pontis. Specific areas in the cerebellum, namely, parts of the posterior lobe vermis, the flocculus, and the paraflocculus, receive this input and in turn project to the vestibular nuclei. From the vestibular nuclei, activity can then be forwarded to the oculomotor, abducens, and trochlear nuclei, as was described for the VOR earlier.

Figure 9-29 Smooth pursuit pathways. The stimulus for smooth pursuit eye movement is a moving visual target. This causes activity to flow through the circuitry diagramed in the figure and leads to maintenance of the fovea on the target. FEF, frontal eye field; LGN, lateral geniculate nucleus; MVN, medial vestibular nucleus; NRTP, nucleus reticularis tegmenti pontis; SEF, supplementary eye field; V1, primary visual cortex. MT and MST are higher-order visual association areas.

Circuits Underlying Vergence

The neural circuits underlying vergence movements are not well known. There are premotor neurons (neurons that feed onto motor neurons) located in the brainstem areas surrounding the various oculomotor nuclei. In some cortical visual areas and the frontal eye fields there are neurons whose activity is related to the disparity of the image on the two retinas or varies during vergence movements. How vergence signals in these cortical areas feed into the brainstem premotor neurons is not clear. The cerebellum also appears to play a role in vergence movements because cerebellar lesions impair this type of eye movement.