19: Cognitive Functions of the Premotor Systems

• Direct Connections Between the Cerebral Cortex and Spinal Cord Play a Fundamental Role in the Organization of Voluntary Movements

• The Four Premotor Areas of the Primate Brain Also Have Direct Connections in the Spinal Cord

• Motor Circuits Involved in Voluntary Actions Are Organized to Achieve Specific Goals

• The Hand Has a Critical Role in Primate Behavior

• The Joint Activity of Neurons in the Parietal and Premotor Cortex Encodes Potential Motor Acts

  • Some Neurons Encode the Possibilities for Interaction with an Object

  • Mirror Neurons Respond to the Motor Actions of Others

  • Potential Motor Acts Are Suppressed or Released by Motor Planning Centers

• An Overall View

In Chapter 18 we surveyed the higher-order organization of sensory systems. In this chapter we turn to the higher-order functioning of the motor systems, by examining how the brain represents behavioral goals and how voluntary actions are planned to achieve those goals. To illustrate how the motor systems generate goal-oriented behavior, we focus on reaching and grasping, actions that are possible because of the prehensile hand.

The evolution of the prehensile hand greatly enriched the development of cognitive capacities in primates. Indeed, the two are interdependent. As the German philosopher Friedrich Engels wrote: "Man alone has succeeded in impressing his stamp on nature. He has accomplished this primarily and essentially by means of the hand. But step by step with the development of the hand went that of the brain."

The prehensile hand radically changed the way in which primates relate to the external world; but the change occurred slowly, and required the evolution of cortical circuits for a variety of new specialized movements adapted to different objects. In his book The Sensory Hand, Vernon Mountcastle, one of the pioneers in the study of the connection between sensation and action, quotes Herbert Spencer (Principles of Psychology, 1885) on why the hand is so critical to understanding action:

The final neural pathway for all bodily actions, including movement of the hand, is through motor neurons in the ventral horn of the spinal cord. These motor neurons are not simply responding to independently generated sensory information, however. The sensory information needed for action is the product of interaction between the motor systems and sensory systems. Under many circumstances, therefore, action and perception are inseparable. Indeed, many sensory functions serve only to allow for the planning of motor acts. As we focus on and reach for a cup of coffee, our arm is controlled in a manner that is independent of conscious experience—we do not think about which movements to perform and which muscles to contract.

Perception of space, and even more complex cognitive acts, were once thought to be represented only in higher-order sensory and association areas of cerebral cortex. In a radical departure from previous thinking, we now believe that the premotor areas in the cortex may also have cognitive functions.

At the highest levels of sensory-motor interaction, neurons do not simply encode the physical features of the sensory stimulus or the force or direction of movement. Rather, they encode something more abstract that includes features of both the object and the movement: They encode the relationship between the body and the object with respect to a particular goal. For example, in anticipation of drinking they may represent a configuration of the hand in relation to graspable features of a cup.

Although picking up a cup appears to be a simple mechanical action, the neural machinery underlying it is surprisingly complex, requiring a number of preparatory steps in the parietal and frontal premotor and motor cortex.

As discussed in Chapter 1, the discovery that electrical stimulation of different parts of the frontal lobe produces movements of the opposite side of the body had a major effect on thinking about localization of function in the brain. Brodmann's area 4, the area in the frontal lobe in which the lowest-intensity stimulation elicited movement, was designated the primary motor cortex (Figure 19–1). By systematically stimulating the primary motor cortex and attributing the movement elicited with each stimulus to the activation of neurons near the electrodes, researchers identified groups of neurons that controlled movement of specific body parts and learned that these functional groups were distributed on the cortex in the form of a somatotopic map.

In recent years our understanding of the functional organization of the motor areas of the cerebral cortex has changed dramatically, and a new picture of the cortical control of movement is emerging. The functional organization of the primary motor cortex is not simply an isomorphic map of the body in which adjacent peripheral sites are represented in adjacent cortical sites. Instead, individual muscles and joints are represented in the cortex multiple times in a complex mosaic. This makes it possible for the cortex to organize combinations of elemental movements suitable to specific tasks.

Each muscle and joint is represented by a column of neurons whose axons branch and terminate in several functionally related spinal motor nuclei (this branching is minimal for cortical cells that control distal muscles because these muscles require more independent control). The cortical neurons also form synapses with interneurons in the spinal cord. These connections allow voluntary movements to switch on entire spinal circuits—the motor neurons, interneurons, and central pattern generators that execute reflex actions. These circuits are then able to integrate and convert local sensory input to motor output without further direction from cortical centers.

In the 1930s physiologists discovered that movement could also be elicited by stimulation of premotor areas. Brodmann's area 6 contains four main premotor areas that project directly to the spinal cord. Two areas lie on the lateral surface and two on the medial surface in Brodmann's area 6 (Figure 19–1.) Each of these four cortical areas may be viewed as a relay in a densely interconnected network that controls reaching and grasping by activating spinal motor circuits.

In contrast to neurons in the primary motor cortex, movement-related neurons in the premotor areas fire in connection with a variety of movements because these neurons encode a general goal-directed command such as "grasp the cup" or "pick up the raisin." Neurons called set-related neurons, common in premotor areas but relatively rare in the primary motor cortex, are more active in the absence of any overt behavior, such as during the delay between a behavioral cue and the behavior. Other neurons encode global sensorimotor transformations, such as "always move at 180 degrees from the visual stimulus." Thus, just as there is a hierarchy of spinal and supraspinal motor control, there is a hierarchy of representations of movement features within the different motor and premotor areas of the cortex.

To produce movement, signals from premotor and motor areas of the cortex must ultimately reach motor neurons in the spinal cord. The Dutch anatomist Hans Kuypers identified three motor pathways: a direct corticospinal pathway and two indirect pathways, the medial and lateral brain stem systems (Figure 19–2).

The historically well-known corticospinal system is involved in the control of all aspects of body and limb movement but has a special role in the fractionated movements necessary for skilled motor acts such as playing the piano or typing. Much of the control of fractionated movements is exercised by the primary motor cortex. Thus, a lesion of the primary motor cortex destroys the ability to oppose the thumb and first finger so as to pick up a raisin or grasp a cup. A patient with such damage is unable to move the fingers independently and can only grasp a cup clumsily.

In humans the corticospinal tract consists of approximately one million axons, of which 30% to 40% originate from neurons in the primary motor cortex. The rest of the axons have their origins mainly in the premotor and supplementary motor cortices, and in the parietal areas lying posterior to the precentral sulcus. Together, corticospinal axons from these various areas descend through the subcortical white matter, internal capsule, and cerebral peduncle. As the fibers of the corticospinal tract descend they form the medullary pyramids, prominent protuberances on the ventral surface of the medulla. Consequently the entire projection is sometimes called the pyramidal tract. Like the ascending somatosensory system, most fibers of the descending corticospinal tract cross the midline in the medulla, at the pyramidal decussation, to terminate in the spinal cord of the opposite side.

The motor information carried in the corticospinal tract is significantly modulated by a continuous stream of information from other motor regions as well as tactile, visual, and proprioceptive information needed to make voluntary movement both accurate and properly sequenced.

The medial brain stem system originates in portions of the reticular formation, vestibular nuclei, and superior colliculus. This system receives information from the cortex and other motor centers for the control of posture and locomotion. The lateral brain stem system originates from the red nucleus. It receives input from the cortex as well but is involved in the control of arm and hand movements.

Spinal motor circuits are not regulated solely by descending commands. Reflex circuits and pattern generators within the spinal cord can coordinate stereotyped movements such as stepping without descending signals (see Chapter 35). A newborn infant, whose descending pathways cannot yet control the spinal cord, is able to execute stepping movements when lifted into the air. Descending systems coordinate reflex and patterned movements generated by spinal motor circuits and can even create new patterns of muscle activation through direct action on motor neurons. This cortical control enables greater flexibility of movements than is possible through exclusively local coordination among the spinal motor circuits.

The cortical motor areas and brain stem in turn receive input from two major subcortical structures: the cerebellum and basal ganglia (Figure 19–3, and see Figure 16–9). These two structures provide feedback essential for the smooth execution of skilled movements and thus are important for motor learning, the improvement in motor skills through practice. The cerebellum and the basal ganglia store memory for unconscious motor skills through pathways that are separate from those used to store factual memories of events that can be recalled consciously (see Chapter 66).

The cerebellum receives somatosensory information directly from primary afferent fibers arising in the spinal cord as well as information about movement from corticospinal axons descending from the neocortex. The basal ganglia receive direct projections from much of the neocortex, which supply both sensory information and information about movement (see Figure 16–9).

In primates four functionally distinct premotor areas also send direct connections to the spinal cord (see Figure 19–1).

The two areas on the lateral surface are the lateral ventral premotor area and lateral dorsal premotor area. As we shall see later, the ventral premotor cortex mostly controls mouth and hand movements. Most of its neurons do not discharge in association with simple movements toward an object. They only become active during goal-directed actions such as grasping, holding, or manipulating an object. The two areas on the medial surface are the supplementary motor area, which lies in the medial wall of Brodmann's area 6, and the cingulate motor areas, a group of motor areas buried in the cingulate sulcus. Similar premotor areas also exist in humans, but differences in size and sulcal patterns make it difficult to identify homologous areas with precision.

These four premotor areas are connected to the primary motor cortex. In addition, like the primary motor cortex, each premotor area has neurons that project to the brain stem as well as neurons that project directly to the spinal cord. Thus voluntary movements are controlled by descending signals from several cortical areas. For this reason, the task of generating limb movements is thought to be broken up into multiple subtasks, each managed in parallel by one of the several cortical motor areas.

These premotor areas also have dense reciprocal connections with the association areas in the posterior parietal cortex (Figure 19–4). These reciprocal connections constitute the visuomotor circuits that mediate different types of visually guided motor behavior such as mouth movements, arm reaching, and hand grasping.

Primates have remarkable visuomotor capacities. We can link the sight of an object with quite different actions. Seeing a cup of coffee, we can pick it up, drink from it, or throw it against the wall. As we shall learn in greater detail in the chapters on vision, some visual pathways are concerned only with perception, whereas some are concerned with planning motor acts. In fact, these two kinds of visual information are processed in separate pathways originating in different areas of cortex: the ventral (what) and dorsal (where) visual streams (Figure 19–5).

The ventral stream terminates ventrally in the inferotemporal lobe. It carries information that allows us to distinguish the visual properties of objects, an orange from a tennis ball, for example. Because the visual properties of the orange and the tennis ball are different, they are represented differently in the inferotemporal cortex, even though they may be the same size and occupy the same spatial location.

Charles Gross first demonstrated that the neurons in the inferotemporal cortex are highly selective for specific and complex visual stimuli. For example, some neurons are selective for individual faces, an extreme example of visual discrimination. Other neurons in the inferotemporal lobe near the superior temporal sulcus react selectively to the sight of movements of other individuals.

The dorsal stream terminates in the posterior parietal lobe and carries information that allows us to locate objects in space and to act on them. The dorsal stream has two branches: a dorsal (dorso-dorsal) branch and a ventral (ventro-dorsal) branch (Figure19–5B). The dorso-dorsal stream is involved in the control of movements. Damage of this pathway produces optic ataxia, a clinical syndrome in which the patient knows the location of objects but is unable to reach them properly. The ventro-dorsal stream is involved in visuomotor transformations necessary for interacting with objects, as well as space perception. The most common deficit following damage of this pathway is spatial neglect (see Chapter 17).

The idea that the dorsal stream is crucial not only for space perception, as traditionally thought, but also for the organization of action, derives from experiments by David Milner and Melvin Goodale on a patient with large lesions of the ventral visual stream. This patient had a dramatic dissociation between the capacity to distinguish geometric shapes conceptually and the capacity to act on objects. When asked to describe the orientation of a slot, the patient would respond virtually at chance. However, when asked to insert a card inside the slot, the patient performed the action well, moving the card toward the slot in the correct orientation and inserting it accurately (Figure 19–6). Thus the visuomotor circuits of the dorsal stream alone are capable of guiding behavior.

Voluntary initiation of movement is one of the defining characteristics of animal behavior. Reflex actions are more or less stereotypic responses to external or internal stimuli. Voluntary actions are manifestations of centrally generated intentions to move; the goal—reaching for a cup of hot coffee—determines a series of actions that will lead to its achievement.

Voluntary movements allow an animal to explore the world around it to satisfy not only immediate but also future needs. Unlike reflex behavior, the stimuli for voluntary behavior do not determine the response; they only set the occasion for it. Therefore, an animal may or may not respond to a stimulus, or it may respond to the same stimulus in different ways depending upon its needs. Under different conditions an animal may approach, avoid, or ignore the same stimulus.

The primate brain, particularly the human brain, is characterized by an enormous expansion of the cerebral cortex. This expansion correlates with an extraordinary increase of sensory, motor, and cognitive capacities. A large part of the expansion has occurred in three regions: the rostral part of the frontal lobe (the prefrontal cortex) and two posterior regions of cortex, the posterior part of the parietal lobe and the inferotemporal cortex.

As discussed in the preceding chapters, the posterior part of the parietal lobe and the inferotemporal cortex are association areas that integrate information from different sensory modalities. This association is essential for the formation of percepts, like space and objects. Indeed, patients with lesions in the parietal cortex typically have deficits in spatial perception, whereas patients with damage in the inferotemporal cortex have problems recognizing objects and faces.

Prior to 1970 deficits in spatial perception following parietal damage were attributed to the destruction of an area that was thought to encode a single internal representation of the external world. However, a series of anatomical and functional studies have revealed that the parietal lobe contains more than one representation of space and each of these representations is strikingly dependent on motor activity.

Jaana Hyvärinen assessed the responses of individual neurons in the inferior parietal cortex to sensory stimuli and during motor behaviors. He found that most neurons respond to sensory stimuli that occur in specific locations and often to stimuli of different modalities, both visual and somatic stimuli for example. Other neurons discharge in association with motor acts. Actions carried out with different parts of the body are represented in different but overlapping zones.

This orderly representation of movements is similar to the somatotopic arrangement of somatosensory inputs in the postcentral gyrus (see Chapter 17). Neurons in the rostral part of the inferior parietal lobe discharge in relation to movements of the mouth; neurons in the central part of this lobe fire in association with hand and arm movements; those located more caudally fire in relation to eye movements. Some neurons in this region respond to objects that lie close to the subject, in the space within arm's reach (peripersonal space), without necessarily requiring that the movement actually occur. Other neurons prefer stimuli located far from an individual's body. Neurons that respond to objects in peripersonal space are located mostly in the part of the inferior parietal lobe where mouth and hand movements are represented, whereas neurons that respond preferentially to objects further away from the body are found mostly in the part of the parietal lobe where eye movements are represented.

What are the critical steps involved in grasping? The neural control of grasping has been studied by Hideo Sakata and Giacomo Rizzolatti. Sakata investigated the anterior intraparietal area (AIP) (see Figure 19–4), the region where previous studies had identified neurons that discharge in association with grasping movements. Sakata and his colleagues recorded the activity of individual neurons in alert monkeys trained to grasp different types of objects, each requiring a specific type of grip. The monkeys carried out these tasks under normal light, using visual guidance, but also in the dark, using memory. In this way Sakata and his colleagues discovered that the neurons in this area fall into three main classes: motor-dominant, visual-dominant, and visual-and-motor neurons.

Motor-dominant neurons discharge equally well during movements performed in light and in the dark. Visual-dominant neurons discharge only during movement performed in light. Visual-and-motor neurons fire during movements performed in both light and dark but their discharge is stronger in the light. Most visual-dominant and visual-and-motor neurons also respond when the monkey looks at an object but does not reach for it. Neurons that discharge selectively during manipulation of an object also discharge selectively during visual fixation of that object without movement. Based on this finding Sakata proposed that neurons in the anterior intraparietal area are involved in transforming sensory representations of objects into motor representations of how to grasp them.

Rizzolatti and co-workers next recorded the activity of neurons in a sector of the ventral premotor cortex of monkeys—area F5 within Brodmann's area 6 (see Figure 19–4)—while the animals performed a variety of actions. They found sets of neurons that discharged in association with different types of hand actions: grasping, holding, and manipulating. Some were active regardless of how the object was grasped. Most, however, discharged only when the monkey used a specific type of grip. The most commonly represented grip types were the precision grip (grasping with the index finger and thumb, typically used for small objects), whole-hand grasp (clutching, used for large objects), and finger grasp.

Surprisingly, a considerable fraction of neurons in area F5, called canonical neurons, also discharged when the monkey simply observed an object, whether or not the object was subsequently grasped (Figure 19–7). Thus these neurons discharged in response to presentation of an object even though the neuron did not have anything to do with preparation to grasp the object. Most canonical neurons were selectively activated by the presentation of objects of a certain size, shape, and orientation. Neurons that discharged during a precision grip (performed with the index finger and thumb) also fired in response to the presentation of small objects, whereas neurons that discharged during whole-hand grasping also fired in response to the presentation of large objects.

What explains the behavior of canonical neurons? The fact that their activity is not related to motor planning and their responses are selective for certain objects and not others rules out nonspecific factors such as attention or intention, because nonspecific factors would have the same effect for all presented objects. Rather the presentation of an object seems to trigger the translation of the object's physical properties into a potential motor act.

Some Neurons Encode the Possibilities for Interaction with an Object

The studies of the parietal and premotor canonical neurons led Michael Arbib, and subsequently Rizzolatti and Giuseppe Luppino, to formulate a new model of how sensory representations of objects are transformed into hand movements.

Their model is based on the functional significance of the responses of visual-dominant and visual-and-motor neurons in the anterior intraparietal area to the presentation of three-dimensional objects. Their thinking about these responses was influenced by the notion of affordance introduced several years ago by the psychologist James J. Gibson. According to Gibson the sight of an object triggers an immediate and automatic selection of those properties of the object that allow one to interact with it. These properties, or affordances, are not the visual aspects of the object (shape, mass, color, etc.) but the pragmatic opportunities that the object affords the observer.

As we have seen, aspects of visual stimuli that are useful for action are analyzed in the dorsal stream of the visual system. Based on the extensive elaboration of an object's properties in the extrastriate visual areas of the dorsal stream beginning in V₂, the visual-dominant and visual-and-motor neurons in the anterior intraparietal cortex are able to encode the object's affordances. This information is then sent to F5 neurons that encode potential motor acts. An F5 neuron can transform a given affordance into an appropriate potential motor act because of the congruence of its response to the affordance and the motor act it controls. Object becomes action.

In real life objects usually have more than one affordance and may be grasped in several ways. How does the brain determine which is optimal? Behavioral analysis of grasping reveals that nonvisual factors determine the choice of affordance and thus how an object will actually be grasped. These factors relate both to what the object is for and the individual's intent at that moment. A cup is a simple example. The cup has three major affordances: the handle, the top, and the body. If the cup is recognized as a cup and the person wants to use it in the way a cup is commonly used, the person will grasp it by the handle. However, if the person wants to move the cup or give it to someone, the cup may be taken not by the handle but by its body or its rim (Figure 19–8).

Thus a more complete (and realistic) model of the grasping circuit involving the anterior intraparietal area and the premotor area F5 must assume that the circuit extracts automatically not one but all the affordances of an object. A specific affordance will be selected according to the information the circuit receives about the meaning of the object and the individual's intention.

As far as the meaning of the object is concerned, the ventral visual stream (Figure 19–8) is specifically dedicated to description of objects. This stream ends in the inferotemporal lobe, which is richly connected reciprocally with the inferior parietal lobe, including the anterior intraparietal area. It is likely that these connections convey to the parietal lobe the semantic properties of an object; this information, as well as the motor use of the object specified in the premotor circuits, is the basis for the selection of an appropriate affordance consistent with the standard uses of the object.

The anatomical substrate for selecting affordances on the basis of the individual's current goal could be the input to the inferior parietal lobe from the prefrontal lobe, where long-term motor planning occurs. Thus when an unconventional use of an object is intended, the prefrontal input could override the selection of standard affordances and select those affordances that are congruent with the individual's intention. For example, if an individual wants to throw a cup instead of drinking from it, the affordances presented by the body of the cup or its top will be selected rather than the affordance presented by the handle.

Mirror Neurons Respond to the Motor Actions of Others

Canonical neurons discharge both when the monkey acts on an object and when it simply looks at the object without acting. In both cases these neurons fire in much the same manner and send the same signal to other neural centers. The firing of the neuron in the absence of an overt movement represents therefore a potential motor act: This activity occurs in a circuit that plans a movement but does not trigger a motor command. Potential motor acts afford an individual the freedom to choose whether or not to respond to a stimulus or simply hold it in memory.

Potential motor acts have an entirely different significance in a set of extremely interesting visuomotor neurons, the mirror neurons, in area F5. These neurons, like all neurons in F5, discharge during specific motor acts such as grasping, tearing, or holding. In addition, they also fire when the monkey observes another individual (human or monkey) performing the same motor act. They do not discharge in response to mere object presentation. In other words, these neurons represent the action done by another individual as a potential motor act.

What may be the function of these neurons? One attractive idea is that individuals know the outcome of the motor acts they plan. Thus, when the mirror neurons discharge in response to a motor act done by another individual, the observer understands what the other individual is doing because the observed action elicits in his premotor cortex a motor plan whose outcome is known to him.

At first glance it may seem strange that our motor faculties might be involved in understanding what others are doing, what their intentions are. But how else could we obtain this information? Although our visual system provides us with a description of the overt aspects of an action, it does not tell us what the action means, what its purpose is. Mirror neurons could in principle provide us with an experience-based understanding of observed actions, a basis for understanding the intention of others (see Chapter 38).

An essential step in the biological evolution of social cognition is our ability to interact with each other in a meaningful and constructive way. When you and I talk, you not only know the content of your own mind but also have a sense of what I am thinking and how I am reacting. A defect in social cognition may contribute to autism, a serious developmental disorder in which a child's ability to communicate socially is impaired. Normal communication requires, in addition to familiarity with language and the ability to express oneself, certain sensitivity to the thoughts and feelings of the person with whom one is communicating. One of the central features of autism is difficulty in understanding the perspectives, thoughts, ideas, and intentions of another person.

The mirror neurons are probably the most basic system the brain has for understanding others' intentions. Other cortical regions, such as the region near the superior temporal sulcus and some rostral medial cortical areas, also play a role in understanding another's intentions, especially when complex reasoning is required.

Potential Motor Acts Are Suppressed or Released by Motor Planning Centers

The representation of potential motor acts by the nervous system raises a further question. What prevents a potential motor act from being executed? Are there control mechanisms inhibiting or facilitating implementation? Damage to certain premotor cortical areas, or the motor cortex, results in neurological syndromes that strongly suggest that there are such controls. Some of these behavioral disorders include difficulty in initiating movements or making movements that are not consciously intended.

A particularly telling example is the utilization behavior syndrome. Individuals with this syndrome pick up objects in an almost compulsive way. Once they observe an object, they immediately grasp it, even if it belongs to another person or to the doctor examining them. This syndrome may result from impaired restraint of the potential motor acts elicited by objects.

A key feature of voluntary motor behavior is that certain motor acts are executed while others are restrained. Voluntary action depends on sequencing elementary movements to form purposeful action. This ability is the prerequisite for many of our daily actions such as typing, using a computer, playing a musical instrument, and even speech. Karl Lashley called the task of sequencing motor actions the "serial order problem" of motor behavior. The sequence of motor actions is thought to involve parallel computations in multiple cortical areas and subcortical nuclei, including the basal ganglia and the supplementary motor area.

Neurons in the supplementary motor area are involved in the planning, generation, and control of sequential motor actions. Thus, when monkeys were trained to perform different sequences of three simple elemental arm movements—push, pull, and turn—some neurons in the supplementary motor area were active before any movement occurred but only when a specific sequence was planned. For example, a neuron could be active prior to the performance of a pull-turn-push sequence but not before a pull-push-turn sequence. Other neurons were active while a particular movement was performed but only if the movement was preceded or followed by another specific movement.

The brain recognizes objects and carries out actions in ways no existing computer can even begin to approach. Recognizing a face and appreciating a landscape are amazing computational achievements that require sophisticated processing of complex information. Indeed, all our perceptions are analytical triumphs. However, even more amazing is how all this perceptual analysis is integrated with motor circuits for even the simplest voluntary actions such as picking up a cup of coffee.

The final neural pathway for any action that determines the force exerted by individual muscles is through the motor neurons in the ventral horn of the spinal cord. Nevertheless, in sophisticated animals like monkeys or humans, actions are not produced by the spinal cord alone. Several sensory and motor regions in the brain are also involved.

The planning and execution of voluntary movement relies on sensorimotor transformations in which representations of the external environment are integrated with intentions and motor programs. This integration is the product of premotor and primary motor areas operating in conjunction with sensory and association areas of the cerebral cortex. An example of this is the communication between parietal and motor areas during visually guided reaching.

In our everyday experience it seems we perceive an object before interacting with it; and thus intuitively we might expect the brain to work in this sequential way. According to this model perceptual mechanisms first generate a unified representation of the external world, cognitive processes use this replica of the world to decide on a course of action, and finally an action plan is relayed to the motor systems for implementation.

As we have seen, this intuitive view does not capture the reality of how the brain decides and executes movement. In fact, a novel behavior requires simultaneous processing in multiple motor and sensory areas because the behavioral action is continuously monitored for errors and modified. As the behavior becomes more accurate, the need for sampling of the sensory inflow and updating of the motor program decreases—the need for the computational power of large networks lessens. Thus, for example, the presupplementary motor area is active during the learning of a behavior but becomes less active as learning progresses. After long periods of practice, when the behavior becomes automatic, activity in the presupplementary motor area ceases.