Motor Control

Motor ControlSTEVEN P. WISE

National Institute of Mental Health

REZA SHADMEHR

Johns Hopkins University

I. What Controls Movement

II. What the Motor System Controls

III. Mechanisms of Motor Control

IV. Motor Memory

V. Flexibility in Motor Control

VI. Evolution of the Motor System

GLOSSARY

agonist A muscle that generates movement toward a goal.

antagonist Amuscle that generates force in a direction opposite to

that of an agonist.

central pattern generator Neural network in the central nervous

system that produces rhythmic motor commands.

forward dynamics Computation of movements that result when

muscles generate a given pattern of force.

forward kinematics Computation of limb positions for a given

pattern of joint rotations.

internal model A central nervous system representation of the

kinematics and dynamics of a motor task.

inverse dynamics Computation of muscle activations and forces

needed to reach a goal.

inverse kinematics Computation of joint rotations or limb

movements needed to reach a goal.

synergists Muscles that work together to produce a given move-

ment.

The human motor system controls goal-directed movementby selecting the targets of action, generating a motorplan, and coordinating the forces needed to achieve

those objectives. Genes encode much of the informa-tion required for both the development and theoperation of the motor systemFespecially for actionsinvolving locomotion, orientation and reproduc-tionFbut every individual must acquire and storemotor memories during his or her lifetime. Some ofthat information reaches conscious awareness, butmuch of it does not. This article focuses on the humanmotor system but draws heavily on knowledge aboutthe motor system of other mammals, especiallyprimates.

I. WHAT CONTROLS MOVEMENT

The motor system consists of two interacting parts:peripheral and central. The peripheral motor systemincludes muscles and both motor and sensory nervefibers. The central motor system has componentsthroughout the central nervous system (CNS), includ-ing the cerebral cortex, basal ganglia, cerebellum,brain stem, and spinal cord.

A. Peripheral Motor System

1. Muscles

Skeletal muscles consist of specialized cells, which fuseduring development to form fibers (technically, asyncitium). There are two types of muscle fibers:

EHB.2001.0216 No. of pages: 1–21 Pgn:: GNS

Encyclopedia of the Human Brain Copyright 2002, Elsevier Science (USA).Volume 3 All rights reserved.1

extrafusal and intrafusal. Extrafusal fibers, whichattach to tendons and then to the skeleton, produceforce and movement. Intrafusal fibers, which containmuscle spindles, attach to muscles and serve a sensoryfunction.Force and movement depend on muscle proteins,

principally myosin and actin, both of which formstrands within the muscle fibers. Molecules of myosinstore kinetic energy as a result of metabolizingadenosine triphosphate (ATP), and muscle activationconverts this chemical energy into mechanical forceandwork.Muscles generate force through a cascade ofelectrical and biochemical events, beginning with therelease of acetylcholine by motor neuron synapses atthe neuromuscular junction. This excitatory neuro-transmitter binds temporarily with the muscle’s cho-lineregic receptors, leading to depolarization of thepostsynaptic membrane and mobilization of intracel-lular calcium ions. High intracellular calcium levelsexpose a site on the actin filaments towhichmyosin canattach. Once attached, the myosin molecules thenreconfigure to force the actin and myosin filaments toslide relative to each other, which either shortens themuscle or generates force.Whether these biomechanical events cause a force

leading to movement or, alternatively, force withoutmovement depends on the interaction of those muscleswith their tendons, the skeleton, and the environment.As one consequence of the properties of actin andmyosin, the length of a muscle affects the force (ortension) that it generates, a property known as thelength–force (or length–tension) relation. As a muscleelongates, a given amount of activity generates aproportionately larger force. In addition, the proper-ties of actin,myosin, and other structural elements alsocause muscle fibers to behave approximately like aspring (technically termed viscoelasticity). Like metalsprings, muscles have varying degrees of stiffness. Forexample, when pulled with a given amount of force, aspring made of thick, inflexible metal will increase itslength much less than one made of thin, pliable metal.The former kind of spring is called stiff, and the latter iscalled compliant. Stiffness is defined as the ratio offorce change to length change.

a. Rigor mortis Muscle activation increases mus-cle stiffness, but so does death, which leads to rigormortis. Death causes depolarization of the musclefibers because the eventual depletion of ATP stops thesodium–potassium pump, upon which normal cellpolarization depends, as well as the hydrolysis of ATPby myosin, upon which detachment from actin

depends. In both rigor mortis and movement, attach-ment of myosin to actin causes stiffening of the musclefibers.

2. Motor Neurons and Motor Units

In the ventral part of the spinal cord, motor neuronsare organized into segregated motor pools, whichinnervate particular muscles. Alpha motor neuronssend their axons from the spinal gray matter toterminate on extrafusal muscle fibers. Gamma motorneurons send their axons to intrafusal muscle fibers.Motor pools extend over two to four spinal segments,with medially situated motor pools innervating axialmuscles (e.g., those of the neck and spine). Laterallysituated motor pools project to limb muscles, withthose contacting distal muscles located most laterally.The term motor unit applies to a motor neuron and

the muscle fibers it controls. Each motor neuronbranches to innervate many muscle fibers, whichreceive input from only one motor neuron (exceptvery early in development and in some disease states).The number of fibers in a motor unit varies accordingto function and within each muscle. Motor units thatcontribute to fine movements, such as those of the eyeor the fingers, usually have a small number of musclefibers. For example, motor units in the eye musclesconsist of three to six muscle fibers. However, gastro-cnemius, which forms the belly of the calf muscle, hasthousands of muscle fibers per motor unit. Largemotor neurons typically innervate more muscle fibers.There are three basic types of motor units, each

categorized by the speed with which it contracts uponelectrical stimulation and its fatigability upon repeatedstimulation. Fast, quickly fatiguing (FF) motor unitshave a short contraction time and produce a hightwitch tension. However, with repeated stimulationthe force they generate dissipates rapidly. Fast,fatigue-resistant (FR) motor units have an intermedi-ate contraction time and can maintain force longer,whereas slow, nonfatiguing (S)motor units have a longcontraction time and show little or no loss of force withrepeated stimulation. FFmotor units have largemotorneurons, fast conducting, large-diameter axons, andmuscle fibers of relatively large diameter. Smotor unitshave the opposite characteristics. Muscles have var-ious proportions ofmotor units: For example, Smotorunits make up nearly the entire diaphragm (one wouldnot want to get tired of breathing), whereas gastro-cnemius has a large proportion of FR and FF units(one certainly can get tired after a few strenuousjumps).

MOTOR CONTROL2


a. Poliomyelitis: A Disorder of Motor Neurons and

Motor Units The poliovirus invades motor neurons,leaving adjacent nerve cells intact. Poliovirus recep-tors, located at the neuromuscular junction, allow theviruses to enter the motor neuron’s axon, after whichthey migrate to its cell body. The infected cell eitherovercomes the virus or dies. If it dies, that motor unit islost, and if the entire motor pool dies permanentparalysis results. However, some motor neuronsusually survive. They develop new terminal axons thatsprout to reinnervate ‘‘orphaned’’ muscle fibers. Asingle motor neuron may innervate up to 10 times thenormal number of muscle fibers, restoring motorfunction. However, in a process called remodeling,which occurs in both healthy people and poliopatients, motor units continually lose old sprouts andgrow new ones. Decades after the onset of poliomye-litis, the enlarged motor units begin to break down,causing renewed weakness. According to one hypoth-esis, intense use of the relatively few remaining motorneurons causes this cell death.

3. Muscle Afferents

Certain sensory neurons innervate muscles and pro-vide the CNS with information about muscle lengthand force. These and other sensory neurons have cellbodies in a dorsal root ganglion, with one axonprojecting to sensory receptors in the periphery andanother terminating in the CNS.The diameter of muscle-afferent axons determines

whether they belong to group I or group II. Group I,subdivided into groups Ia and Ib, has the larger fiberdiameter and therefore faster transmission rates.Group Ia and II fibers innervate muscle spindles andare therefore called muscle spindle afferents. The termspindle refers to fine intrafusal muscle fibers that taperat the end and contain a fluid-filled capsule at thecenter. Muscle-afferent fibers wrap around muscularelementswithin the capsule.Group Ia fibers are termedprimary muscle spindle afferents; group II fibers arecalled secondary muscle spindle afferents. Group Ibfibers innervate golgi tendon organs (GTOs), whichare located in the transitional region between extra-fusal muscle fibers and tendons. The role of muscleafferents in reflex responses is discussed in Section II.B.

B. Central Motor System

All levels of the CNS contribute to motor control,including the spinal cord, medulla, pons, midbrain,

diencephalon, and telencephalon. The following sec-tions survey the major components of the centralmotor system, beginning with the spinal cord andprogressing up the neural axis to the telencephalon.Figure 1 shows some of the major components andprojections of the central motor system. Notwith-standing the impression that this component-by-component description might convey, the variouscomponents of themotor systemwork as an integratedneural network, not as isolated motor ‘‘centers.’’

1. Spinal Cord

In addition to motor neurons, spinal components ofthe motor system include sensory pathways, theproprioceptive system, and central pattern generators(CPGs). Sensory afferents bring information to theCNS from the skin, joints, and muscles, and both cellsand fiber tracts in the spinal cord relay that informa-tion to structures involved in motor control. Forexample, primary afferent neurons terminate on thedorsal column nuclei in the medulla, which relay thatinformation to the thalamus through a fiber pathwaycalled the medial lemniscus. In addition, an extensiveand intricate system of intrinsic spinal cord neuronsunderlies a group of miscellaneous functions collec-tively termed proprioceptive. Proprioceptors are sen-sory transducers in muscles, tendons, and otherinternal tissues. However, the concept of the proprio-ceptive system extends beyond this definition toinclude a wide variety of interneurons that relaysomatosensory signals locally and between segmentsin the spinal cord as well as carry descending motorcommands, largely within the spinal cord. CPGs areneural networks that generate patterned, rhythmicmovements, such as those involved in walking andrunning.

2. Brain Stem

The brain stem contains motor neurons that send theiraxons through certain cranial nerves, primarily tomuscles of the tongue, face, and eyes. Like the spinalmotor pools, many of these cranial motor nucleireceive direct input from sensory neurons and lessdirect influences from proprioceptive interneurons.Brain stem CPGs generate rhythmic movements, suchas those underlying breathing and chewing.Some parts of the brain stem interact with spinal

CPGs and other components of the spinal motorsystem. One such region has been termed the midbrainlocomotion region, which is thought to trigger the

MOTOR CONTROL 3


activity of spinal CPGs and thereby initiate locomo-tion. However, higher order networks, akin to CPGsbut having more complex output patterns, have animportant role in a number of instinctive behaviors,including aggressive posturing (a form of ‘‘bodylanguage’’) and inarticulate vocalization (such ascrying, laughing, and screaming). Some of thesenetworks are located in and near a midbrain structurecalled the periaqueductal gray.

a. Reticulospinal System Cells in the brain stemreticular formation that project to the spinal cordmake up the reticulospinal system, which extendsthrough medullary, pontine, and midbrain levels. Thereticulospinal system performs a diverse set of func-tions, including the regulation of muscle tone, controlof posture and locomotion, and integration of lowerorder motor signals with those emanating from thecerebellum and cerebral cortex. Different reticulosp-inal pathways exert influences on flexor versusextensor muscles and on proximal versus distal partsof the limb.Part of the reticulospinal system serves as a fast

transmission route to postural motor neurons andhelps prevent movements from destabilizing balance.

For example, when a person lifts a heavy object, the legmuscles need to stiffen before the elbow flexes. Thispostural adjustment prevents the object’s weight frompulling the person off balance. The reticulospinalsystem activates leg muscles to stiffen them and helppreserve balance. On the whole, while people areawake the reticulospinal system has a predominantlyfacilatory influence on motor pools. However, thiseffect changes dramatically during sleep. Then, re-ticulospinal neurons exert a strong inhibitory influencethat, for example, prevents the performance ofimagined actions during dreams.Important influences over the reticulospinal system

come from other systems, including vestibular affer-ents, which signal movements of the head and itsorientation with respect to the earth’s gravitationalfield, and the motor cortex, which provides informa-tion otherwise unavailable at brain stem levels.Through vestibulospinal projections, the vestibularsystem can contribute directly to various reflexes thatadjust eye position, posture, and limb movements.However, the vestibular afferents also provide inputsto the reticulospinal system. Consider the role of thereticulospinal system as a person runs through a fieldof obstacles. The signals conveyed by the reticulosp-

Figure 1 Major components of the motor system. Corticofugal projections are depicted as gray arrows. Projections to effectors are depicted

as open arrows and arrowheads. The basal ganglia are contained within the gray box. Preganglionic autonomic motor nuclei are shown as

stippled ovals. GPi, internal segment of globus pallidus; GPe, external segment of globus pallidus; STN, subthalamic nucleus; SNr, substantia

nigra pars reticulata.

MOTOR CONTROL4


inal system to CPGs and spinal motor pools adjustposture and movement based primarily on vestibularand proprioceptive inputs. However, cortical andother higher order inputs supply the informationneeded for dynamic motor adjustments that allowpeople to step over and around visible obstacles.

b. Cerebellum and Red Nucleus The largest com-ponent of the brain stem motor system is thecerebellum. The medial cerebellum controls posture,whereas the lateral cerebellum participates more involuntary movement. Accordingly, vestibular andpropriospinal inputs predominate in the medialcerebellum, and inputs to the lateral cerebellum arisemainly from the cerebral cortex, relayed throughmossy fibers originating in the basilar pontine nuclei.In addition, the cerebellum receives mossy fiber inputfrom the red nucleus via the lateral reticular nucleus(which also has major spinal inputs) and from othersources.Mossy fibers terminate on the output nuclei ofthe cerebellum (the deep cerebellar nuclei) as well as onneurons in the cerebellar cortex.Another type of input,conveyed by climbing fibers originating in the inferiorolivary complex, is thought to signal motor error ordiscoordination. These signals may play a central rolein motor learning (see Section IV).The output of the cerebellar cortex comes from

GABAergic Purkinje cells, which inhibit neurons inthe deep cerebellar nuclei and in one of the vestibularnuclei. The deep cerebellar nuclei send excitatoryoutputs to a variety of structures. Their largestprojections terminate in the thalamus (Fig. 1), butother efferents reach the reticulospinal system, rednucleus, superior colliculus, and spinal cord. Cerebel-lar outputs to many of its targets are accompanied byreturn projections through a variety of direct andindirect pathways. One example is the cerebellarprojection to the motor cortex (via the thalamus),which is returned by a cortical projection to thecerebellum (via the basilar pontine nuclei). Recurrent,excitatory circuits such as this are thought to formfunctional networks termed cortical–cerebellar mod-ules.The red nucleus plays an enigmatic role in motor

control, especially in the human brain, but appears tobe intimately related to cerebellar function. It receivesa major projection from the deep cerebellar nuclei aswell as from the motor cortex, and the largest part ofthe red nucleus (its parvocellular, or small-cell,component) projects predominantly to the inferiorolivary complex, the source of cerebellar climbingfibers. Themagnocellular (large-cell) red nucleus sends

its axons directly to the spinal cord through therubrospinal tract, which might be particularly im-portant in stabilizing the limb by coactivating agonistand antagonist muscles. However, the magnocellularred nucleus is said to be relatively small in the humanbrain, which may reflect a dominant role of corticalmotor control in our species.

c. Superior Colliculus The superior colliculus,although typically discussed in terms of eye-movementcontrol, also has an important role in the control ofhead movements. Generally stated, its functioninvolves the orientation of the retina and otherreceptors on the head, which the superior colliculusguides through its interaction with the reticulospinalsystem, premotor neurons in the brain stem reticularformation, and direct projections to the spinal cord(the tectospinal system).

3. Diencephalon

The hypothalamus and thalamus are themajor parts ofthe motor system that lie within the diencephalon. Themotor functions of the hypothalamus are discussed inSection II.A. The thalamus does not serve a primarilymotor function when viewed as a whole. However, amajor component of the thalamus receives projectionsfrom the cerebellum and basal ganglia, which play animportant role in motor control. Two general regionsof the thalamus receive these inputsFthe ventroan-terior and ventrolateral (VA/VL) nuclei. Anterior VA/VL receives basal ganglia projections; posterior VA/VL receives cerebellar input. Each part of VA/VLsends excitatory projections to the frontal cortex andreceives excitatory projections from the same corticalareas. These reciprocal connections are thought to actas recurrent cortical–thalamic modules. Although thethalamic terminals from cerebellum and basal gangliaoverlap very little, the thalamocortical components ofthese systems converge to influence most if not allmotor areas jointly.

4. Telencephalon

Two large parts of the telencephalon have importantroles in motor control: the cerebral cortex and thebasal ganglia. Of course, the function of both struc-tures extends beyond motor control, but this articleaddresses only that role.

a. Cerebral Cortex The number of functionallydistinct motor cortical fields remains unknown. One

MOTOR CONTROL 5


heuristically useful view of the frontal cortex divides itinto three main parts: the primary motor cortex, agroup of areas collectively known as nonprimarymotor cortex, and the prefrontal cortex (Fig. 2). Thefirst two components can be referred to collectively asthe motor cortex, although this term is sometimes usedas a synonym for primary motor cortex.The primary motor cortex (abbreviated M1) corre-

sponds approximately to Brodmann’s area 4. It lies inthe anterior bank of the central sulcus and contains atopographic representation of the musculature. Mosttextbooks depict this topography in the form of ahomunculus (i.e., a projection of the body onto thecortical surface). Such pictures have validity only atthemost superficial level. They correctly imply that themedial part of M1 contains the leg and footrepresentation, that a more lateral part includes thearm and hand representations, and that an even morelateral part has the face, tongue, and mouth repre-sentation. However, at any finer level of detail, thehomunculus presents an inaccurate image of M1’sorganization. Instead, M1 consists of a mosaic of

broadly overlappingmuscle representations, with eachpart of the body represented repeatedly. The functionsof M1 are discussed in Section III.By current estimates, there are about a dozen

nonprimary motor areas. Many of these fields occupyparts of Brodmann’s area 6. However, parts of areas 8and 24, the latter also known as anterior cingulatecortex, also contain nonprimary motor areas. Amedial group of areas includes the well-knownsupplementary motor area (SMA), an area immedi-ately anterior to it (the pre-SMA), and two or morecingulate motor areas. A lateral group of areas, oftentermed simply the premotor cortex (PM), can bedivided into separate dorsal and ventral components,and these two divisions can be subdivided further intorostral and caudal parts. At least two eye-movementfields, the frontal eye field, and the supplementary eyefield, have been identified. Finer subdivisions andcombinations of these regions have been proposed andmay have some validity. The functions of these areasare discussed in Sections III and V.B.The primary and nonprimarymotor areas project to

the other components of the motor system, at virtuallyall levels of the neural axis (Fig. 1). The cortical outputsystem, often called the corticofugal system, arisesexclusively from layer 5 and includes a direct projec-tion to the spinal cord (the corticospinal system). Somemotor areas, especially M1, have monosynapticexcitatory connections with alpha motor neurons.Accordingly, the oxymoron ‘‘upper motor neuron’’has been applied to M1’s corticospinal neurons.However, M1 exerts its influence over genuine motorneurons in large measure through relatively indirectpathways, and its synapses on spinal interneuronsvastly outnumber those on motor neurons. Corticosp-inal axons accumulate in large bundles traversing theinternal capsule, cerebral peduncle, pyramidal tract,and corticospinal tract on the way from the telence-phalon to the spinal cord. Because corticospinal axonsrun through the pyramidal tract, the term pyramidalmotor system is sometimes applied to it. However,many fiber systems coexist within the pyramidal tract,such as neurons projecting to the dorsal column nucleiand other targets in the brain stem. Thus, thecorticospinal and pyramidal motor systems do notcorrespond exactly. Additional outputs from themotor cortex include massive projections to the basalganglia, the red nucleus, the cerebellum (via the basilarpontine nuclei), and the reticulospinal system.

b. Basal Ganglia Long considered part of the so-called extrapyramidal motor system, much of the

Figure 2 Motor areas of the human brain plotted onto a highly

simplified sketch of the cerebral hemispheres. The lateral surface of

the left hemisphere is shown below the medial surface of the right

hemisphere. Rostral is to the left; dorsal is up. The dashed line

indicates the fundus of the central sulcus, showing that the primary

motor cortex is located mainly within the rostral (anterior) back of

that sulcus. The stippled area on the medial surface represents the

corpus callosum. Ce, central sulcus; CMAs, cingulate motor areas;

FEF, frontal eye field; M1, primary motor cortex; PF, prefrontal

cortex; PMd, dorsal premotor cortex; PMv, ventral premotor cortex;

PPC, posterior parietal cortex; SEF, supplementary eye field; pSMA,

presupplementary motor area; SMA, supplementary motor area.

MOTOR CONTROL6


motor output from the basal ganglia depends,ultimately, on the pyramidal tract. Accordingly, theconcept of an extrapyramidal motor system has beenlargely abandoned. The striatum is the basal ganglia’sinput structure, encompassing the putamen, caudatenucleus, nucleus accumbens, and other regions withinthe ventral forebrain. Likewise, the pallidum is thebasal ganglia’s output structure, including not only theglobus pallidus but also the substantia nigra parsreticulata (Fig. 1) and additional parts of the ventralforebrain. The pallidum sends GABAergic inhibitoryprojections to the brain stem and thalamus. Thesethalamic nuclei connect to nearly the entire frontallobe as well as additional areas, such as the inferiortemporal contex and the posterior papietal cortex.The major excitatory inputs to the striatum come

from the cerebral cortex (including the hippocampus)and the intralaminar complex of thalamic nuclei. Amajor input arises from the dopaminergic cells of themidbrain, located in the substantia nigra pars com-pacta and the adjacent ventral tegmental area.Degeneration of the striatum results in Huntington’sdisease, whereas degeneration of the dopaminergicneurons causes Parkinson’s disease.Much attention has recently focused on themodular

organization of cortical and basal ganglia interconnec-tions. This trend accords with the recognition of otherimportantmodules in themotor system, such asCPGs,reflex circuits, and cortical–cerebellar modules. Thecortical–basal ganglionic modules, often called‘‘loops,’’ consist of cortical, striatal, pallidal, andthalamic elements that form, at least in principle, arecurrent excitatory pathway. This circuit includes theso-called ‘‘direct pathway,’’ striatal output neurons(technically, medium spiny neurons) that projectdirectly to the pallidal projection neurons (Fig. 1),including those in the internal segment of the globuspallidus (GPi). Another part of the basal ganglia, the‘‘indirect pathway,’’ begins with projections from thestriatum to the external segment of the globus pallidus(GPe). These neurons influence the subthalamicnucleus and pallidum, in turn (Fig. 1).This sketch of the basal ganglia has heuristic value in

understanding Parkinson’s disease and other conse-quences of basal ganglia dysfunction. However, thereader should recognize that it represents a coarseoversimplification.Many other features of its anatomyare important to basal ganglia physiology. Thesubthalamic nucleus, for example, excites not justGPi but also GPe; the GPe sends inhibitory inputs‘‘back’’ to the striatum; and the motor cortex sends adirect, excitatory projection to the subthalamic nu-

cleus. Furthermore, in addition to dopaminergicinputs, serotonergic inputs arise from the raphenucleus, and there are a large variety of intrinsicneurons using other neurotransmitters, includingacetylcholine.

c. Parkinson’s Disease According to one currentview, dopamine acts to support activation of the directpathway’s striatal neurons but to suppress those of theindirect pathway. Accordingly, the direct pathway’sinhibitory influence over GPi might wane in theabsence of dopamine. This decrease in inhibition fromthe striatal direct-output pathway would lead to GPineurons having greater activity and therefore greaterinhibitory output to the thalamus. The indirect path-way, may also contribute to greater inhibition of thethalamus. This increased inhibitory influence onrecurrent cortical–thalamic modules may cause theslowing of movement (technically, bradykinesia) thatis a symptom of Parkinson’s disease.

II. WHAT THE MOTOR SYSTEM CONTROLS

The motor system controls innate behavior, which isencoded genetically; a wide variety of reflex responses,defined as output generated relatively directly bysensory inputs; and voluntary movements.

A. Instinctive Action

The hypothalamus plays a central role in the control ofinstinctive behaviors, such as those involved in loco-motion, orientation, and reproduction, and in neu-roendocrine function. Pools of neuroendocrineneurons in the periventricular parts of the hypothala-mus secrete hormones into the vascular system. Theirterminals in the neurohypophysis (posterior pituitary)release oxytocin and antidiuretic hormone (vasopres-sin). Although it might seem counterintuitive toconsider such secretions as motor in function, theyserve as a mechanism through which the CNS controlsother parts of the body, just as alpha motor neuronsdo.Different cells in the periventricular hypothalamus

serve a less direct endrocrine control role. Theseneuroendocrine cells secrete higher order controlhormones, often termed releasing factors, into theportal blood supply of the pituitary. The adenohypo-physis (anterior pituitary) responds to these hormones,

MOTOR CONTROL 7


such as corticotropin-releasing factor, which increaseor decrease the secretion of pituitary hormones such asadrenocorticotropin. These hormones play crucialroles in regulating growth and in homeostatic andreproductive functions.The hypothalamus also affects the body through its

control of the autonomic nervous system (ANS). Thehypothalamus sends descending projections to theautonomic motor nuclei in the spinal cord and brainstem. There, neurons send outputs to the peripheralganglia of the two components of the ANS, thesympathetic and parasympathetic systems. Sympa-thetic motor neurons are located at thoracic andlumbar levels of the spinal cord. Parasympatheticmotor neurons are found in the sacral spinal cord aswell as in the brain stem, from which they give rise toparts of cranial nerves V (trigeminal), VII (facial), IX(glossopharyngeal), and X (vagus). Autonomic motorneurons, like alpha motor neurons, project to theperiphery and release acetylcholine. In the sympatheticnervous system, most of these cholinergic influencesterminate on ganglia relatively near the CNS. Thesesympathetic ganglia contain neurons that releasenoradrenaline, which induces a generalized arousalof the ‘‘fight, flight, and fright’’ variety. In theparasympathetic system, cholinergic motor neuronsproject to ganglia relatively far from the CNS, neartheir visceral targets. Thus, the ANS can address thetargets of the parasympathetic system more specifi-cally than it can for the sympathetic system. Mostparasympathetic postganglionic synapses use acetyl-choline as a transmitter, although many neuropep-tides, such as vasoactive intestinal peptide, are releasedalong with acetylcholine. The hypothalamus, throughits control of the ANS, exerts an important influenceover such motor functions as vasoconstriction, re-spiration, and heart rate.However, the hypothalamus does not confine its

motor functions to the ANS. Parts of the hypothala-mus play an important role in ingestive behaviors(such as eating and drinking), defensive and agonisticbehaviors (such as flight from danger and aggression),arousal and orientation, and social behaviors, includ-ing sexual and other behaviors involved in reproduc-tion (e.g., rearing of progeny and other means bywhich genes are passed to future generations). Itsmotor roles thus include initiating complex actionpatterns, which include epigenetically expressed butgenetically encoded motor programs often termedspecies-specific or species-typical behaviors. For ex-ample, aggressive displays such as snarling at orstaring down an adversary might result from fear,

intermale competition, or irritation, with each stateand response mediated by different, partially over-lapping networks in the hypothalamus. The hypotha-lamus effectuates many of these behaviors viaprojections to the thalamus and to various brain stemstructures (e.g., the periaqueductal gray).The several systems influenced by the hypothala-

musFendocrine, autonomic, and instinctiveFmayseem dissimilar and unrelated. However, the hypotha-lamus (along with components of the amygdala)coordinates these aspects into a fully coordinatedbehavior. Imagine combat soldiers engaged in astressful and dangerous situation. A variety of spe-cies-typical behaviors accompany such perilous situa-tions, including heightened states of vigilance andarousal, which require high expenditures of energy.Parts of the medial hypothalamus detect inputs thatsignal such situations and, through their projections tothe brain stem, release and intensify arousal andvigilance behaviors. Those hypothalamic regions alsoinfluence the periventricular hypothalamus to secretecorticotropin-releasing factor, which in turn inducesthe release and circulation of adrenocorticotropichormone. This hormone stimulates the adrenal cortexto produce and release larger amounts of glucocorti-coids. A parallel ANS control signal from the hypo-thalamus to the motor nucleus of the vagus promotesinsulin secretion by the pancreas. Thus, the neuroen-docrine system (through gluocorticoids) induces theliver to release more glucose at the same time as theANS (through insulin) promotes the uptake of glucoseinto cells, especially muscle fibers. Together, thehypothalamus coordinates an adaptive response to astressful state: the mobilization and utilization ofstored nutrients to support high energy expenditure aswell as the actions appropriate to that state.

B. Reflex Responses

The motor system controls a large number of reflexes,of which this article highlights only withdrawal andmuscle-afferent reflexes.

1. Withdrawal Reflexes

Activation of cutaneous and deep receptors by apotentially damaging stimulus gives rise to an ipsilat-eral flexion reflex, accompanied by contralateralextension, called the crossed-extension reflex. A poly-synaptic network in the spinal cord mediates this

MOTOR CONTROL8


response, which acts to retract the limb from thenoxious stimulus (the flexion reflex) and enhancepostural support, especially from the legs (thecrossed-extension reflex).

2. Muscle-Afferent Reflexes

Two feedback systems, both involving muscle affer-ents, regulate force andmuscle length through reflexes.Muscle spindle afferents transduce muscle length,whereas GTO afferents transduce muscle force.

a. Force Feedback GTOs lie in series with theextrafusal fibers and receive no motor innervation.They send force information to the spinal cord, whereinterneurons receive input from the brain that specifiesthe amount of force that a muscle should produce. Ifthat muscle’s force level exceeds this set point, theGTO inputs inhibit the alpha motor neurons innervat-ing that muscle, which lowers the force producedunless some other mechanism cancels that signal.

b. Length Feedback In contrast to GTOs, musclespindles lie in parallel with extrafusal fibers and havecontractile elements that are activated by gammamotor neurons. Without efferent innervation, themuscle spindles would become slack when extrafusalfibers shorten. The spindle afferents would thenbecome silent and the CNS would lose informationabout muscle length. Gamma motor neurons activatemuscle spindles during contraction to maintain thatinformation flow. During movement and steadyposture, muscle spindle afferents sense muscle lengthwith respect to a bias length set by their gamma motorneurons. When the gamma motor neurons have highlevels of activity, the bias length is relatively short. Ifthe muscle length exceeds this bias length, musclespindle afferents increase their discharge rate andexcite alpha motor neurons innervating the samemuscle. Of course, this increase in activity tends toshorten themuscle, bringing it closer to the bias length.Generally, however, gamma motor neuron activityallows the CNS to control the sensitivity of musclespindle afferents, whichmay play their most importantrole in regulating muscle stiffness.Information regarding limb position does not

depend entirely on muscle spindles. Other sources ofinput include cutaneous and joint capsule receptors,both of which contribute information about limbposition and joint angle. In addition, group III and IVafferents also innervate the limbs but receive informa-

tion primarily from deep receptors in the muscle andcutaneous receptors that appear to respond mainly topainful stimuli rather than force or limb position.Despite this diversity of receptors, the muscle spindlesappear to be especially important for sensing musclelength, as demonstrated by the following experiment:Imagine a blindfolded person, seated with his or herelbows on a table and his or her forearm held in avertical posture. If someone else moves one forearm,the blindfolded person can indicate the position of thatarm by matching it with the other, free arm. Peopleperform this task very accurately in normal circum-stances. However, if vibration is applied to the belly ofbiceps, people consistently overestimate the angle ofextension at the elbow. The explanation for thisphenomenon involves muscle spindle afferents. Vibra-tion provides a very powerful stimulus to musclespindles, and the CNS wrongly interprets their in-creased discharge as reflecting a longer biceps muscle,which translates to increased extension at the elbow.Muscle spindle afferents also mediate stretch

reflexes, among which the monosynaptic stretch reflexis also known as the myotactic reflex or the knee-jerkresponse. For example, when one taps the skin surfaceover the knee’s patellar tendon, this stretches thequadriceps muscle. Prior to the involvement of stretchreflexes, this increase in length makes the musclegenerate more force through the length–tensionrelation. However, stretch also results in an elongationof the muscle spindles in the quadriceps, which in turncauses increased firing of the primary and secondarymuscle spindle afferents. This sensory input excitesalphamotor neurons and causes a knee-jerkwithin 15–20msec. Particularly for the muscles of the arm, wrist,and fingers, a second pathway exists through whichmuscle stretch can activate motor neurons. Thispathway, called the long-loop stretch reflex, alsobegins with muscle spindle afferents. Informationfrom these afferents is transmitted to the thalamusand then to the somatosensory and motor cortexbefore returning to the spinal cord through thecorticospinal projection. It takes 40–50msec for theinformation to traverse this entire circuit. Althoughthe long-loop reflex takes longer than the short-loopone, the brain can reprogram the long-loop response ina highly flexible manner. For example, if a stretch isexpected, people can suppress the long-loop compo-nent of the reflex or choose to respond especiallyvigorously. The major role of stretch reflexes probablyinvolves responses to an unexpected perturbation. If,during a movement, something suddenly displaces thearm, this input elicits a compensatory response from

MOTOR CONTROL 9


both the short- and long-loop reflexes. These reflexestend to change the activation levels of motor neuronsin a manner that stabilizes and stiffens the limb as itmoves along the desired trajectory. Reflexes may helpthe motor system overcome impediments that havenever been experienced. If a person’s goal includesproducing the sound ‘‘pa,’’ the lips must touch eachother to achieve this goal. However, if an experimenterpulls on the lower lip, the motor system needs toproducemore force than usual tomake the lips contacteach other (technically, occlusion). This response doesnot depend on experience with such perturbations; itoccurs the first time the lip is pulled.

c. Disorders of Long-Loop Stretch Reflexes in

Parkinson’s Disease Normally, people can volunta-rily suppress the long-loop component of the stretchreflex. In Parkinson’s disease, however, patientsappear to lose this ability. Regardless of the instruc-tions given to them, stimulation of primary musclespindle afferents produces a large response. InHuntington’s disease, by contrast, there can be acomplete loss of the long-loop stretch reflex. Therelation between the basal ganglia and control of thesereflexes remains unknown, but it probably involves itsinfluence over the motor cortex. Recent researchsuggests that loss of the long-loop reflexes accountsfor the jerky movements characteristic of Hunting-ton’s disease.

3. Role of Reflexes in Voluntary Movement

To what extent are reflexes used for generation ofvoluntary movements? Studies of patients with degen-eration of large-fiber afferents (peripheral neuropathy)have led to the conclusion that reflexes play only anindirect role in volitional action. Such patients losetheir stretch reflexes, limb position sense, and theirability to detect limb motion, but they can still makevoluntary movements. In normal individuals, anelectromyograph (EMG) pattern composed of threediscrete bursts of activity characterizes the executionof a rapid, one-joint voluntary movement. First, theagonist muscle activates (AG1), followed by theantagonist muscle (ANT) and, finally, a secondactivation of the agonist muscle (AG2). AG1 accel-erates the limb, ANT brakes the limb, and AG2stabilizes the limb and dampens oscillations aroundthe final position. When peripheral neuropathy pa-tients make a rapid thumb flexion, the typical three-phase EMGpattern described previously (AG1, ANT,

and AG2) appears in the normal manner. Therefore,the EMG pattern of voluntary movements appears tooriginate from descending motor commands to alphamotor neurons and does not depend on reflex loops.The motor command signal, however, may passthrough some of the same neurons used in reflex loops,increasing the capacity of the system to integrateperipheral and central information.

C. Voluntary Movement

This article concentrates on the voluntary control offorelimb reaching movements. However, similar pro-cesses apply to other types of voluntary movements.These processes include generating torques on anarticulated chain of limb segments, promoting stabilitythrough agonist–antagonist architecture, positioningend effectors, translating goals into plans for action,computing the joint rotations and patterns of forceneeded to implement those plans, and compensatingfor other forces.

1. Generating Torques on ArticulatedLimb Segments

A limb consists of a chain of articulated segments, withthe muscles acting as the motors technically, actuatorsthat control torques around those segments. Eachsegment of a limb can rotate with respect to the moreproximal segment. The axis of rotation centers on thejoint that connects the two segments, and musclesprovide torques on that joint. For example, a personusually has to flex his or her elbow to lift a coffee cupand sip from it. Commands from the motor systemreach the biceps muscle, activating it and producingforce, which results in flexion torques on the elbowjoint. As the elbow flexes, the resulting movementstretches the triceps, which in other circumstanceswould result in increased force output from the tricepsbecause of the length–tension relation. This increasedforce would cause an extension torque on the elbowjoint. For the hand to reach the mouth, flexion torquesneed to exceed extension torques. Thus, while sendingthe activating commands to the biceps to initiate themovement, the motor system usually sends an inhibi-tory command to the triceps’ motor pool as well. Thisreduces some of the extension torque produced by thetriceps, diminishing the resistance to the voluntaryflexion that brings the cup to the mouth.

MOTOR CONTROL10


2. Producing Stability ThroughAntagonistic Architecture

To hold a coffee cup steady requires a differentapproach: The motor system sends commands toactivate both biceps and triceps. This coactivationresults in both flexion and extension torques on theelbow joint. If the net torque, which is the sum of thesetwo torques, equals zero, then the forearm will remainstill. Why not simply shut down both muscles ratherthan waste the torques (and energy) to no effect? Theanswer has to do with limb stability. Consider whathappens if an object suddenly hits a person’s hand andcauses the arm to flex at the elbow. During coactiva-tion, the muscle that gets stretched because of theimpact (triceps) will vigorously resist because intrinsicmuscle stiffness increases with activation level. With-out coactivation (i.e., with an inactive triceps), theimpact would result in a much larger flexion of theelbow. Thus, coactivation promotes limb stability.

3. Positioning End Effectors

Although the final motor commands act on muscles tomove limb segments, the goal of a movement ofteninvolves the positioning of an end-effector. Forexample, to sip from a coffee cup, most people attendto the cup and not to what the elbow is doing.Signatures provide another case in point. Everyone hasa unique signature, and its distinctive characterpersists even if different joints and muscles performthe signing movement. This principle has been calledmotor equivalence. Its basis is that the pencil serves asan end effector regardless of which muscles move it.The preservation of unique elements in a signatureindicates that the highest levels of the CNS representvoluntary movements not as a pattern of muscleactivations but as a kinematic pattern, specifically thedesired motion of an end effector.Of course, signatures consist of complicated move-

ment trajectories. In principle, an infinite number ofpossible hand trajectories can be made between twopointsFsome straight, others curved to varyingdegrees. However, unless the goal includes a curvedtrajectory, people show a remarkable similarity in themovements that they produce in reaching from a givenhand position to a target. The hand moves with aunimodal, smooth, and symmetric velocity over thetime course of the action, and it takes a straight-linepath. Even blind people show this feature in their armmovements. When an experimenter demonstrates thetarget to a blind person by moving his or her hand to

the target position (later returning it to the originallocation), he or she makes straight and smoothreaching movements just like sighted people. Thissmoothness in hand trajectory contrasts sharply withthe changes that occur in joint positions during thesame movement. For example, consider a movementof the right arm that starts with the hand at the far leftof the midline and reaches to a target at the far right.For most people, this movement would be a straightline in terms of hand position. However, examinationof joint angles shows that the elbow initially flexes andthen extends. Therefore, its velocity is not unimodal.Human armmovements generally appear simple whendescribed in terms of hand positions and velocities butare complex when described in joint coordinates. Thisregularity remains when people perform movementswith different end effectors. For example, movementsremain smooth and simple when our hands hold a longstick. In this case, the end of the stick moves smoothlyand in a straight line.

4. Translating Goals into Action

a. Inverse Kinematics The smooth and simplehand trajectory described previously represents akinematic plan. Before that plan can be formulated,the motor system must estimate both current handposition and the direction and magnitude of themovement needed to reach the target. Estimation ofcurrent hand position is based on two sources: Visionand proprioception. Muscle afferents from the armprovide the information necessary for estimating theorientation of each limb segment relative to itsproximal joint. If the motor system has this informa-tion, it can compute the hand position with respect tothe body. The computation of limb position from aproximal, joint-coordinate-based system to a distal,hand-centered coordinate system is called forwardkinematics. If someone moves a blindfolded person’shand, he or she still has a pretty good idea of thathand’s location. This ability depends primarily on thecomputation of forward kinematics from the lengthsensors in the muscles. If the motor system knows thelength of the limb’s muscles, it knows the angles of itsjoints and, through forward kinematics, can computethe location of the hand. The inverse of this computa-tion maps hand position to the joint angles that areappropriate for it. In order to move the hand to adesired position, the motor system needs informationabout what joint angles the muscles need to achieve inorder to move the limb segments to that position. Thiscomputation is termed inverse kinematics. In other

MOTOR CONTROL 11


words, if the motor system knows the desired handposition, it can compute the joint angles needed to putthe hand in that position through the computation ofinverse kinematics.

b. Inverse Dynamics In addition to computing thepositions of the joints and the hand for a desired limbtrajectory (kinematics), the motor system must esti-mate how much torque to produce on each joint(dynamics). Accordingly, the motor system musttranslate a desired motion of the end effector into apattern of muscle activations. This does not imply thatthe brain calculates or represents the joint torques inabsolute terms, joint by joint, but rather that the neuralnetwork must solve this problem to generate motorcommands that will achieve the goal. Consider thetorques needed to lift a full cup of coffee in contrastwith those needed to lift an empty cup. Although thehand trajectories in the two cases maymatch perfectly,torques on the elbow will differ. Therefore, the motorsystem must take into account the weight of objectsbefore it sends motor commands to the muscles. Thecomputation that estimates the motion that will occuras a result of an applied force is called forwarddynamics. The mass of objects held in the hand affectsthis computation: Activation of the biceps at a certainlevel will flex the elbow by a smaller amount for a fullcup than for an empty cup. Forward dynamics consistsof predicting the elbow angle after the biceps receivesits activation command. The ability to predict thesensory consequences of motor commands relies onthis computation. The inverse of this computation,called inverse dynamics, allows the motor system totransform the desired motion of the limb into thepatterns of muscle activation that produce the torquesrequired for the task.However, in everyday life, even movements as

simple as lifting a coffee cup can encounter impedi-ments.When something disturbs armmovements (e.g.,an unexpected change in the load on the hand),movements lose their smooth and regular character.However, provided that the perturbations have highpredictability, with practice the movements againbecome straight in terms of hand trajectory. Thisconvergence toward a straight, simple trajectory inhand coordinates (rather than joint coordinates)further supports the idea that the motor system plansmovements in terms of the position of the hand andother end effectors rather than joint angles or patternsof muscle activity. In other words, the motor systemplans in terms of goals rather than the components ofmovement. Motor learning and memory often under-

lies the ability to make smooth, straight movementdespite external perturbations and the forces of eachpart of a limb acting on the others.

III. MECHANISMS OF MOTOR CONTROL

Neurophysiologists have only begun to understand themechanisms of the motor system, and this sectionshould be considered a highly provisional account ofthese mechanisms. Most of the relevant informationcomes from studies of neuronal activity in awake,behaving monkeys.In generating the plan for a simple reaching move-

ment, the initial problem involves kinematics, figuringout the current location of the end effector (often thehand), the location of the target, and perhaps the pathbetween them. Both the premotor cortex (PM) and theposterior parietal cortex (PPC) appear to have keyroles in solving this problem. The general locations ofthese cortical regions are depicted in Fig. 2.

A. Location of Targets and Initial Hand Position

PPC, in particular, plays an important role in deter-mining the location of objects that could serve as thetarget of a reaching movement. Cells in one part of thePPC, the parietooccipital area (PO; approximatelycorresponding to areas V6 andV6A), respond to visualstimuli and, unlike most visual areas, their receptivefields have no bias toward the foveal representation.This characteristic suggests that the visual informationPO processes relates to the control ofmovement ratherthan the analysis of an object’s features. Some PPCneurons signal the presence of a target in retinalcoordinates (i.e., the location of a target with respect tothe fovea). These cells could be particularly importantin controlling eye movements to that target, but theymight also function to compute reaching, head, andeye movements within a single coordinate framework.One PPC region, the lateral intraparietal cortex,appears to be particularly important for eye move-ments, whereas a nearby region, the medial intrapar-ietal cortex (MIP; also known as the partial reachregion, PRR), plays a larger role in reaching move-ments. Other cells in PPC show an influence of‘‘extraretinal’’ signals such as eye and head position.Some PO neurons, for example, appear to indicatetarget location with respect to the head (i.e., in head-centered coordinates). These signals mark the begin-ning stages of transforming visual information from a

MOTOR CONTROL12


receptor-based (retinal) coordinate frame into onemore useful for reaching movements. For thosemovements, it is more useful to represent the targetlocation relative to the end effector that will make themovement. Other parts of PPC play a larger role indetermining the initial position of the end effector. Onepart of parietal area 5, for example, encodes handposition with respect to the shoulder. These and otherparts of PPC operate in cooperation with the motorcortex, including both PM and M1.Like the PPC, cells in the dorsal part of PM respond

to a combination of signals relevant to voluntaryreaching movement, including visuospatial and pro-prioceptive input, as well as inputs reflecting gazedirection, the location of objects in the environment,the orientation of spatial attention, and nonspatialvisual information (such as color and form). Gazeeffects show that the location of the target relative toeye position has some importance in motor control,perhaps for coordinated movements of eye and handwhen people reach one place while looking at another.Attentional signals might be important when not allreaching targets can be foveated, as often is the case fora sequence of movements in a cluttered visual scene. Inaddition to these target-related signals, cells in dorsalPM (along with many inM1) are sensitive to the initialposition of the hand, possibly through both proprio-ceptive and visual inputs. Similarly, cells in the ventralpart of PM respond to both somatosensory and visualinputs as well as spatial acoustic signals. Whereas thePPC coordinate systems seem to reflect eye-, head-,and body-centered frames, ventral PM appears to bemore specialized for the particular body-centeredcoordinate frame that is most useful for a givenmovement. In ventral PM, when a body part having atactile receptive field moves, the visual receptive fieldmoves in the same way: Visual receptive fields on thehandmove with the hand, and those on the head movewith the head. Therefore, it appears that these PM cellsencode the potential targets of action relative to thebody and update thismapwhenever the pertinent bodypart moves. Thus, PPC and the motor cortex derivemuch of the information needed for formulating akinematic trajectory, including the starting position ofthe hand and the target of movement in each of severalrelevant coordinate frameworks.

B. Dynamics

At least two brain structures mediate the generation offorce profiles that move the hand smoothly to a target:

M1 and the cerebellum. The nonprimary motor cortexand basal ganglia appear to be less involved in thisfunction.It is likely that the network computing inverse

dynamics includes the cerebellum.Damage to it resultsin movements that suggest an inability to compensatefor the complex dynamics of multijoint reachingmovements, including forces that arise due to interac-tion among limb segments (interaction torques). Forexample, the intact motor system approximates theinertia of the arm in programming activations ofmuscles, whereas cerebellar damage results in move-ments that suggest a deficit in this transformation.Normal individuals produce coordinated motion ofthe joints during reaching movements, whereas cere-bellar patients produce movements that often consistof a sequence of single-joint motions. Neurons in thecerebellum have properties consistent with this view,including Purkinje cells that reflect movement velocityand the forces needed to achieve a certain trajectory.M1 plays a role in limb dynamics as well, in part

through its reciprocal interaction with the cerebellum.A century-long controversy has surrounded the ques-tion of whether M1 neurons encode primarily kine-matics or dynamics. This problem, known informallyas the muscle versus movement debate, consists of atleast two parts. One part concerns how M1 addressesmuscles. Some studies in the 1970s suggested that theM1 neurons address individual muscles, but refine-ments in research techniques have led to the under-standing that individual neurons in M1 addressmultiple motor pools, usually of synergistic muscles.This view is consistent with the branching of corti-cospinal neurons to multiple motor pools and with thefact that they terminate principally on spinal inter-neurons rather than alpha motor neurons. Theseresults support the ‘‘movement’’ side of the debate,but there is another issue. It involves themotor-controlsignal thatM1 sends to the motor pools, and it yields adifferent answerFone more consistent with the‘‘muscle’’ side of the debate.A pure force signal in M1 neurons would support

the view that they controlmuscles. Neurophysiologistshave not completely settled this question, but threefacts have been clearly established. First, when thehand generates a force that does not lead to movement(as happens when the hand pushes against an object sorigid that it does not move), cells in M1 haveapproximately the same patterns of activity as whenthe limbmoves. Second, loads exerted on the hand caneither assist a movement or oppose it. Neither kind ofload changes the kinematics ofmovement appreciably,

MOTOR CONTROL 13


but the forces involved in making the movement candiffer greatly. Many cells in M1 differentiate betweenthese two conditions, and they therefore appear toreflect limb dynamics. Third, the posture of the arm,which changes both joint angles and limb dynamics,affects the activity of M1 cells, even when the endeffector path has nearly identical kinematics. Thepredominance of signals related to limb dynamicssupports themuscle side of the debate and, therefore, arole for M1 in controlling limb dynamics. However,this view should not be taken to an extreme. There isevidence that limb kinetmatics is reflected by M1 cellactivity, as well.

C. Kinematics

There is no evidence for a purely kinematic signal inM1. A purely kinematic signal would be invariant toloads on the limb or the initial posture of the arm. Suchinvariance has been observed in PPC (area 5), sup-porting the idea that it is mainly involved inmovementplanning in kinematic terms rather than movementexecution in terms of limb dynamics. However, in M1evidence points to a combination of kinematic anddynamic signals, as shown in the following experiment:Like people, monkeys can move their fingers upwardby bending thewrist in different ways. If the palm is up,then the wrist can be flexed (i.e., moved in the palm’sdirection) to move the fingers upward. If the palm isdown, then the wrist can be extended to achieve thesame results. Some M1 neurons reflect only limbdynamics (i.e., flexion or extension), but others showan intermediate pattern of activity that takes intoaccount both kinematics (end effector movement) anddynamics (muscle activity).Large parts of nonprimary motor cortex, especially

those on the lateral parts of the hemisphere, appear tofunction in the sensory guidance of movement at akinematic level. For example, neurons in the dorsaland ventral PM have greater activity during visuallyguided movements than during memorized sequences,but they are not much affected by loads. The formerobservation points to a specialization for sensoryguidance ofmovement, the latter to a specialization forkinematics. As noted previously, the activity of bothPPC and PM neurons reflects the location of move-ment targets. However, this is not a purely sensoryresponse in either PPC or PM. Instead, their activityreflects the motor significance of those signals atdifferent levels. A term that has been used for motorsignificance in this sense is intention. For example,

when monkeys indirectly move a spot on a videomonitor by pushing or pulling on a joystick, research-ers can distinguish neuronal signals related to thedirection of spot movement from those reflecting thedirection of limb movement. Many neurons in at leastone part of PPC, MIP, signal the direction of handmotion, not spotmotion. Also, they signal the locationof movement-guiding spots only when the spot directsa reaching movement and not when it directs an eyemovement. A population of cells in dorsal PM and thevast majority of cells in M1 do so as well. As anotherexample, when a visual cue indicates the location of thenext target of a reaching movement, but movementneeds to be delayed until some future time, cells in boththe PPC (area 5) and dorsal PM cortex signal thedirection and amplitude of the planned movement.This ‘‘delay-period’’ activity begins about 100msecafter the cue appears and can continue for severalseconds. PPC neurons do not distinguish betweenobjects that will be the target of the next reachingmovement and similar objects that indicate (e.g., bytheir color) that they will not be the subject ofimmediate action. Thus, PPC neurons signal potentialmovements or movement targets but not necessarilythose that are currently planned. PPC neurons alsoreflect, in their activity, the expected benefit to begained by moving toward a potential target. Neuronsin dorsal PM, in contrast to PPC neurons, have delay-period activity onlywhen the object will be the target ofthe upcoming movement. Thus, at the level of thenonprimary motor cortex, especially PM, neuralactivity appears to relatemainly to the implementationof near-term kinematic plans and other relatively high-level goals.Cells in the basal ganglia, also reflect the kinematics

of reaching movements, such as direction and ampli-tude. However, the basal ganglia is unlikely to have asignificant role in solving the inverse dynamics pro-blem or computing themotor plan.Most basal gangliaactivation or inactivation occurs too late to have a verylarge role in movement initiation or the planning thatprecedes it. Instead, the activity in basal gangliadevelops at about the same time as muscle activityand continues during movement. Accordingly, it hasbeen proposed that pallidal output plays a predomi-nantly modulatory role. According to one hypothesis,motor-control signals depend, in part, on recurrent,mutually supporting activity in cortical–thalamicmodules (or loops) that include motor cortex. Pallidaloutput may affect limb kinematics by facilitating orsuppressing these recurrent circuits during an ongoingmovement. It appears paradoxical that when the

MOTOR CONTROL14


motor parts of the basal ganglia’s output projectionare surgically destroyed, patients with Parkinson’sdisease can initiate movements more easily and movefaster than before the surgery. Damage to motor-control structures usually causes rather than relievesmotor dysfunction. However, the movement-modula-tion hypothesis resolves this paradox. Applying thebrakes will slow a car, even though the brakes are notpart of the movement-generation system. It is alsothought that the basal ganglia function in context-dependent movement selection as well as in sequentialand internally generated movements.

D. Goal Achievement

1. Limb Trajectory

M1 neurons reflect information about the direction,magnitude, and speed of the movement, in addition topostural signals. They have their greatest dischargerate for movements in one direction, with system-atically less activity as the direction of movementdiverges from that direction. They are thereforebroadly tuned for movement direction, although thispreferred direction can change with variations instarting hand position, the posture of the limb, andmany other factors. On the assumption that these cellscontribute tomovements in their preferred direction, itis possible to compute a single vector, termed thepopulation vector, representing the net contribution ofa neuronal population. The M1 population vectoranticipates the direction of limb movement forstraight-line reaching movements (as well as for avariety of curved trajectories) by 30–120msec.Cells in the nonprimary motor cortex have proper-

ties similar to those of M1 cells but with someimportant differences. For example, cells in M1 aregenerally specific for the limb used, usually the limbcontralateral to the hemisphere in which the cell islocated. Cells in dorsal PMhave activity broadly tunedfor reaching direction likeM1 neurons but have nearlythe same directional preference for movements of theleft hand or the right hand to the same visuospatialtarget. This finding shows that dorsal PM cells reflectthe movement in terms of either visual targets or thetrajectory of an end effector (in this case, a handle thatthe monkey moves). Evidence points to the formerexplanation. It appears that PM and M1 differ in thedegree to which the visual inputs that guide a move-ment affect the activity of its neurons. Experimentshave been done in which the monkey must follow a

visual trajectory, but this visual input is projecteddirectly to its eyes so that the relationship betweenvision and movement can be altered. Sometimes thevisual target trajectory and the movement trajectory,perhaps an oval, are the same, but sometimes the handtrajectories must be circular to match an oval visualinput or vice versa. PM populations reflect the visualtarget trajectory with more fidelity than the endeffector trajectory, whereas M1 populations moreclosely reflect the movements of the end effector (thehand). This supports the idea that the motor systemcomputes movement trajectories in terms of endeffectors, with visual target trajectory predominantin PM and limb trajectory predominant in M1.

2. Compensation for External Perturbations

A part of the long-loop reflex, M1, receives somato-sensory information that helps compensate for un-expected perturbations during movement. Forexample, when a hand movement is stopped inprogress by an external force,M1 neurons that precedeand accompany the movement with a burst of activityrenew or increase that activity in response. This signalprobably arises from the muscle spindles, whichshorten in concert with the extrafusal fibers due totheir input from gamma motor neurons. When thelimb is stopped, both alpha and gammamotor neuronscontinue to discharge according to themotor plan. Theextrafusual fibers build up force, but themuscle cannotshorten. The muscle spindle fibers continue to con-tract, which generates a signal comparable to thatevoked by amuscle stretch. This signal is relayed toM1and it causes the motor-control signal to be augmen-ted. The resulting increase in activity serves tocompensate, at least partially, for the perturbation,although it may not be adequate to overcome theimpediment. This kind of mechanism could accountfor the achievement of goals upon the initial presenta-tion of a particular perturbation.

IV. MOTOR MEMORY

A. Implicit and Explicit Memory Systems

The brain regions that store motor memories differfrom those that store conscious memories. The formercomprise an aspect of procedural memory or knowl-edge and the latter declarative memory or knowledge.Psychologists often refer to procedural knowledge asimplicit memory and to declarative knowledge as

MOTOR CONTROL 15


explicit memory. Some psychologists use the term‘‘habit’’ interchangeably for procedural knowledge,but this usage should not be confused with itsbiological meaning, which involves instinctive beha-vior.The idea that different brain structures underlie

explicit versus implicit memory comes from observingthe effects of brain damage. Damage to structures inand near the medial temporal lobe (MTL) results inloss of certain recently acquired information. Amnesicpatients with MTL lesions can learn and retain skillssuch as mirror tracing, rotary pursuit, bimanualtracking, and compensation for complex forces ap-plied to the limb during reaching movements. Despitethis motor learning, the patients may not be able torecall the training episodes.The distinction between an explicit memory system,

which depends on the MTL, and an implicit motormemory system has several implications for motorcontrol in the human brain. Voluntary actions havebeen defined as those that are learned, attended, andbased on a comparison among alternatives. Thisawareness depends on the explicit memory system.Other actions, including but by no means limited toreflex movements, proceed without conscious aware-ness. Some subconscious movements bear obviousmarkings of this unawareness, such as the stretch reflexor the vestibuloocular reflex (VOR). The latter servesas a case in point. When people move their head leftwhile looking at something, their eyes move equallyfast and equally far in the opposite direction. They areprobably aware of the object at the focus of attention.However, they cannot report anything about themotor memory that allows them to keep lookingdirectly at that object, regardless of their head move-ments. Adjusting the VOR involves motor learning inthe broadest sense but differs dramatically from thatunderlying voluntary movement. People cannot makeVOR-like eye movements voluntarily. Movementssuch as the VOR and other reflexes can only becontrolled implicitly.Other movements that can be made without con-

scious awareness closely resemble voluntary actions.They can be guided either implicitly or explicitly. Thebest studied example of this phenomenon is termedblindsight. Normally, pointing to visible targets isaccompanied by explicit knowledge of the action andthe goal. However, some people with damage to thevisual system can point to a visual stimulus whiledenying that they see it. Thus, some of the visuomotornetworks remain functional even when the networksunderlying visual perception fail. Phenomena such as

blindsight have led to a distinction between CNSsystems underlying vision for action (and thereforeimplicitly guided action) and those involved in visionfor perception (whichmay ormay not lead to explicitlyguided action). The distinction between these twoinformation processing systems does not depend onbrain damage. Normal people can make finger move-ments that accurately match the size of objects theytouch but nevertheless describe the size of those objectsincorrectly due to visual illusions. People can alsomake accurate saccadic eye movements to fixate avisible target, although they report that the targetmoved in some different direction due to differentkinds of illusions.Which brain structures underlie motor memories,

implicitly guided action, and procedural knowledge?To answer this question, it is useful to distinguishreflexes (such as for the VOR) from explicitly guidedmovements. Much of the information underlying theformer is stored at the brain stem level, including thecerebellum. The cerebellum also has an important rolein classical conditioning, throughwhich sensory inputsare linked to stimuli that trigger reflex responses.Accordingly, it is clear that the cerebellum plays animportant role in motor learning and memory at thereflex level.However, the situation ismore complex formovements that are sometimes voluntary and underconscious control (explicitly guided action) but thatalso can become automatic and unattended as theybecome routine (implicitly guided action).The oldest and most common idea holds that the

MTL subserves explicit knowledge, whereas the basalganglia underlies implicit knowledge. Patients withParkinson’s disease show particular deficits on tasksinvolving implicit estimation of event probabilities, forexample, along with a wide variety of tasks involvingmotor skills. A newer idea holds that explicit knowl-edge is stored in prefrontal cortex and an associatedpart of the basal ganglia, alongwith theMTL, whereassome aspects of implicit knowledge are stored inmotorcortex and their associated parts of the basal ganglia,along with the cerebellum. [The distinction betweenprefrontal and MTL function is that between long-term information storage of many months or years(prefrontal cortex) and intermediate-term storage ofmanyweeks ormonths (MTL)]. Brain imaging studies,especially those that have explored the concept ofattention to action, appear to be more consistent withthe newer view. In motor learning, attention to one’sactions and explicit knowledge of those actions (andoutcomes) dominate the early part of training on anovel task. As a motor task becomes more automatic,

MOTOR CONTROL16


fewer attentive resources are engaged, and eventually itmight be performed without awareness. Neuroima-ging results have consistently shown increased bloodflow, an indirect marker of synaptic activity, in theprefrontal cortex as subjects begin to perform motortasks that involve learning a new skill. This increasedactivation generally declines to baseline as the task isextensively practiced and becomes automatic. Con-versely, as a sequence or skill becomesmore automatic,cerebellum (especially posterior parts), nonprimarymotor cortex, and PPC show increases in activity.These findings support the hypothesis that prefrontalcortex–basal ganglionic modules subserve voluntarymovement, whereas themotor cortex–basal ganglionicmodules (along with the cerebellum and the PPC)underlie more automatic movements of the same kind.Besides neuroimaging experiments, other evidence

also implicates the cerebellum in motor learning. In amonkey trained to stabilize the wrist against anexternally imposed load, for example, the cerebellumhas been inactivated by cooling it. This manipulationeliminated the previously learned, predictive compo-nent of the muscle activity, which would oppose theimposed load. Furthermore, development of motormemory has been associated with an increase in thenumber of synapses onto the Purkinje cells in thecerebellum. Significant synaptic remodeling on Pur-kinje cells takes place within 1–4 hr after completion ofinitial training.

B. Internal Models

1. Acquisition

When a novice operator learns to control a novelmechanical system, the brain solves three types ofcomputational problems: optimization of the taskperformance criteria to arrive at a kinematic plan (i.e.,learning how the mechanical system should behave inorder to accomplish the goals of the task), learning amodel of the forward dynamics of the mechanicalsystem (i.e., acquiring an ability to predict how themechanical system will behave as a function of currentinput, and learning a model of the inverse dynamics ofthe controlled system (i.e., acquiring an ability topredict the inputs that should be provided to themechanical system for a given desired change in stateof that system). For example, consider the case of anoperator acting on a joystick that controls the thrustproduced by the motors of a remote underwatervehicle. The task involves moving the vehicle frompoint A to point B. The kinematic plan specifies a

smooth vehicle trajectory. The forwardmodel specifieshow the vehicle will behave as the operator moves thejoystick. The inversemodel estimates how the operatorshould move the joystick so that the vehicle movesalong the planned trajectory. An internal model (IM)is a blanket term used to describe the informationcontained in the solution to these three types ofcomputational problems, and motor memory refers tothe representation of IMs in the brain.People use IMs in nearly every voluntarymovement.

For example, consider trying to lift an empty bottle ofmilk that has been painted white on the inside so that itappears full. The motor system will generate muscleactivation patterns in the arm that provide the forcesappropriate for lifting a full bottle, resulting in aflailing motion. This fact indicates that in program-ming the motor output to the muscles of the arm, themotor system uses certain visual characteristics of theobject to predict and compensate for its mechanicaldynamics. Learning IMs has been investigated inexperiments in which a subject reaches to a target whileholding the handle of a lightweight robotic arm.Disengagement of the robot’smotors results in smoothand straight-line movements. Motor learning startswhen the investigator programs a pattern of forces forthe robotic arm to produce. These forces representnovel dynamics. When a person starts the trainingprocess, the computations that the brain performs inprogramming muscle activations do not take intoaccount the novel forces, resulting in jerky handmovements. To compensate, initially people stiffen theentire limb through general coactivation of the mus-cles. This results in improved arm stability but servesonly as a temporary and relatively ineffective strategyfor responding to the perturbations. With training,stiffness returns to normal levels at the same time as thebrain builds an IM of the novel dynamics. The motoroutput changes to specifically account for the addi-tional forces. The development and use of a new IMcan be demonstrated by turning off the robot’s motorsat the onset of movement. The resulting movement is amirror image of that observed early in the trainingprocess. Therefore, the motor command that reachesthe muscles includes a prediction by the IM of theforces required to overcome the imposed mechanicaldynamics. The motor system retains this skill formonths after the training session.

2. Consolidation

As one of its fundamental properties, the neuralsubstrate of explicit memory gradually undergoes a

MOTOR CONTROL 17


change, becoming more resistant to disruption. Newlyacquired memories are more sensitive to new experi-ences and more susceptible to interference or braininjury. Posttraining treatments, including electricshocks, removal of key anatomical sites, or inhibitionof protein synthesis, retard this progression and oftenresult in loss of the recently acquired information.These interventions, however, have little effect onrecall once a window of time has passed sinceacquisition. Consolidation refers to this time-depen-dent process. Does such a process occur in the storageof motor memories?Until recently, little evidence supported the idea that

time affected properties of motor memory. Forexample, it was found that whereas electroconvulsiveshock interfered with conscious memories, it sparedretention of a newly acquired skill (e.g., reading wordsin mirror image). Recent work, however, has foundevidence for a temporal gradient in motor memory.The results suggest that within 4–6 hr after completionof training in a reaching task, functional properties ofthe motor memory gradually change. Subjects learnedreaching movements in force pattern 1, leading to aninternal model termed IM1, and then a second forcepattern, termed force pattern 2, leading to an IM2.Their ability to learn IM2 depended on the time thathad passed since completion of practice in forcepattern 1. If only a short time had intervened (lessthan 4 hr), learning of IM2 was impaired compared tothat of naive subjects. By the time 6 hr had elapsed, theinterference caused by their training in force pattern 1had subsided. This evidence points to the limitedcapacity of the system initially engaged in learning newskills. With time, the information maintained in thissystem fades, allowing the learning of additional skills.If this limited-capacity system serves a kind ofintermediate-term memory stage for motor skillspending consolidation into a long-term memory store,then onewould predict that its disruptionmight lead toan inability in long-term recall of motor memories.Indeed, learning IM2 within 4 hr after performing themovements in force pattern 1 results in an inability torecall IM1 in tests of long-term recall.

a. Apraxia as a Disorder of Motor Memory Apra-xia occurs in numerous forms, of which at least twomay represent disorders of motor memory. Ideomotorapraxia is characterized by kinematic errors made aspatients copy movements or perform according toverbal instruction. Ideational apraxia is revealed bythe incorrect miming of an action appropriate for atool. In these patients, the ability to make the

movements is relatively intact, but the motor systemfails to produce the motor program that should becalled up by the input, be it verbal or nonverbalinstructions, as in ideomotor apraxia, or an object witha well-known use, as in ideational apraxia. The brainareas that fail in these situations have not beendefinitively established, but both PM and PPC appearto be involved. PM plays an important role in retrievalof a motor response as cued by visual or auditorystimuli, and PPC lesions cause some forms of apraxia.For example, a dentist with PPC damage suddenlycomplained that he ‘‘did not know’’ how to use a drill.This has been viewed as an example of a loss of apreviously learned motor skill.

V. FLEXIBILITY IN MOTOR CONTROL

The ability to store motor memories enables the motorsystem to select awide variety ofmovements in a highlyflexible manner. People can select actions from a largerepertoire of skills that have been previously learned,depending on context. In many contexts, achieving thegoal of the motor system depends onmovements madedirectly to targets, such as reaching to grasp an object.In others, more flexible relationships need to beestablished between objects and actions.According to current thinking, different kinds of

motor flexibility are afforded by different parts of thefrontal cortex. Lateral nonprimary motor areas (suchas dorsal PM) are thought to compute arbitrarymappings based on external (sensory) cues, whereasmedial nonprimarymotor areas (such as SMA) play ananalogous role for internally generated actions, in-cluding memorized movement sequences. M1 isthought to enable a different sort of flexibility throughthe fractionation of motor synergies. Thus, it iscommonly held that M1 functions mainly to permitmotor ‘‘fractionation’’ (i.e., the independent control ofmuscle groups that usually work in concert). Forexample, the long muscles of the arm attach to severalfingers to either flex or extend them. However, peoplecan move their fingers one at a time.

A. Sensorimotor Mapping

Many reaching movements are made directly toward atarget object. The term standard mapping applies tothese kinds of movement. People often look at anobject, orient attention toward it, and reach directly to

MOTOR CONTROL18


touch or grasp it. However, human behavior would bescarcely recognizable were it limited to standardmapping. People can also look in one direction whileattending to a different place and reaching to a third,and everyone can use both spatial and nonspatialinformation to guide action.

1. Transformational Sensorimotor Mapping

In explaining the difference between standard andnonstandard mapping, prism adaptation serves as acase in point. When a diffracting prism distorts visualinput, objects that lie directly in front of a personmightappear to be 101 off to the right. When the person triesto reach to the object, he or she will reachB101 too farto the right. However, the motor system can recognizethis error and correct it through the process of motoradaptation, similar to that described previously forstudying internal models (see Section IV.B). Thisbehavior serves as a paradigmatic example of standardmapping: The motor system achieves the goal ofdirecting the hand to the object, even though the prismdistorts the object’s location. In contrast, people candecide voluntarily to make a movement B101 to theright of an object’s location. This behavior serves as anexample of nonstandard mapping. The motor systemcan produce an output that uses the spatial informa-tion in an object and transforms it according to somealgorithm (such as 101 to the right) to produce a moreflexiblemotor output than a system limited to standardmapping. This form of nonstandard mapping can betermed transformational mapping because the motoroutput depends on some transformed function of thespatial input.

2. Arbitrary Sensorimotor Mapping

Both of these spatially guided movements contrastwith another form of nonstandard mapping termedarbitrary mapping. When nonspatial visual inputs,such as color, are used to determine the goals of action,an arbitrary mapping must be learned. Imagine abuilding with rooms of two colors: yellow rooms thatrequire doorknobs to be twisted clockwise and bluerooms that require the opposite. Reaching toward thedoorknob relies on standard mapping. Opening thedoors requires arbitrary mapping, at least to do soreliably on the first attempt. Although this example isartificial, most signal- or symbol-guided behavior,including almost all language-guided behavior, de-pends on arbitrary mapping. Examples include stop-ping at a red light or at the sound of the word ‘‘stop.’’

B. Internal versus Sensorimotor Mapping

The idea that lateral nonprimary motor areas (such asPM) underlie external control of action, whereasmedial areas (such as SMA) subserve internal control,has a link to related ideas concerning the motorfunctions of the basal ganglia and the cerebellum. Ithas been proposed that the cerebellum functionspreferentially in externally guided action, whereas thebasal ganglia controls mainly internally guided action.However, the functions of these structures are morecomplex than can be captured by such a simpledichotomy. The cerebellum participates in movementsbased on internal as well as external cues, and the basalganglia plays a role in movements modulated bysensory as well as nonsensory information. Further-more, contrary to earlier views, influences from bothcerebellum and basal ganglia converge on both medialand lateral nonprimary motor areas.How can these differing views be reconciled? Some

progress can be made by recognizing functionalsubdivisions with both the basal ganglia and thecerebellum and by eschewing all-or-none dichotomies.Cells in caudal parts of one deep cerebellar nucleus, thedentate nucleus, have a preference for movementsbased on visual inputs (externally guided action)compared to kinematically similar movements gener-ated from memory, whereas rostral parts have fewercells with such preferences (i.e., they are relativelynonselective). Likewise, cells in the dorsal parts of theGPi have a strong preference formemorized sequences(internally guided action), whereas the ventral parts ofGPi lack such selectivity. M1 and SMA receive theirlargest inputs (via the thalamus) from the less selectiveparts of the basal ganglia and cerebellum. Accord-ingly, M1 and SMA play a fairly general role in motorcontrol and lack strong specializations for eitherinternal or external control.In contrast, PM receives its predominant inputs

from the part of the dentate nucleus more selective forexternally guided movements. Accordingly, PM has alarge role in externally (often visually) guided action.For example, lesions to or inactivation of dorsal PMpreventmonkeys fromusing color (or other nonspatialvisual information) to choose an action.The internally selective (dorsal) part of GPi prefer-

entially influences the pre-SMA, which likewise ap-pears to be specialized for internally guided action.Cells in pre-SMA have their greatest activity forinternally guided movements, including movementsequences, and show other sequence- or order-specificpatterns of activity. Temporary inactivation of the pre-

MOTOR CONTROL 19


SMA disrupts the ability to produce a memorized(internally guided) sequence but not a visually trig-gered one. Brain imaging and brain lesion researchalso suggests the involvement of medial nonprimarymotor areas in the learning of both limb- and eye-movement sequences and the ‘‘spontaneous’’ genera-tion of action. In particular, pre-SMA, has beenproposed to function in changing or updating motorplans based primarily on signals internal to the CNS.Therefore, both medial and lateral parts of the

motor cortex contain generalized as well as specializedcomponents. Generalized parts include SMA (medi-ally), and M1 (laterally). Specialized elements includethe laterally situated PM, which maps sensory infor-mation ontomotor outputs in a highly flexiblemanner,and the medially situated pre-SMA, which along withSMA maps a vast array of memorized and othernonsensory information onto motor output. A medialspecialization for internal information accords withthe role recently proposed for the rostral cingulatemotor area in linking incentive magnitude to theselection of action, as occurs when one must choose anaction based on some estimation of preferred outcome.

VI. EVOLUTION OF THE MOTOR SYSTEM

The motor system was born, not made, and many ofthe characteristics of the human motor system reflectits history. The CNS’s chief function involves theacquisition of a behavioral repertoire, which can bestored both genetically and epigenetically, and theselection from that repertoire of the actionsmost likelyto enhance an individual’s fitness, in the inclusive senseof that term.It seems likely that the ability to move in a goal-

directed manner developed in an invertebrate ancestorof vertebrates. Many experts believe that the coordi-nated guidance of action represents the principalfunction of the CNS and comprises, in all likelihood,the crucial adaptive breakthrough made by thevertebrates and their ancestors. (Other animals madethe same leap, but this article focuses on the humanlineage.) The emphasis on motor control in evolutionmay seem paradoxical because, to us, complex andsophisticated cognition seems to be the principalcharacteristic of advanced animals, not motor control.However, in evolutionary history, long before ourancestors possessed the capacity for language, abstractreasoning, or highly general problem solving, theymoved in relation to objects and places in their

environment. It seems likely that the invertebrateancestors of vertebrates (technically, cephalochor-dates) evolved from filter-feeding ancestors. Theycould swim in a coordinatedmanner, and they directedthose actions with a brain and sense organs concen-trated on the head, including paired eyes. The ability todo these things allowed our very distant ancestors toadopt a hunting life, one probably dominated byolfaction and vision. According to the fossil record,this revolutionary advance occurred about 500–550million years ago. The human motor system, alongwith that of other vertebrates, has developed itscapabilities by building on the original system. Themotor system that people use to run, walk, talk, andmanipulate objects evolved from the same one thatoriginally evolved to control swimming and that wasadapted to control flying, burrowing, and galloping inother species.Thus, the motor system has a high degree of

functional plasticity, at least asmeasured over geologictime. This plasticity of function contrasts with theconservatism of this basic morphological organizationthroughout vertebrate evolution. In the past, autho-rities have speculated that the brain evolved fromcaudal to rostral. That is, early comparative anato-mists thought that the basal ganglia, for example,being located mainly in the telencephalon, evolvedrelatively recently in evolution. Anatomical namessuch as neostriatum testify to this notion. However,recent evidence from comparative morphology showsthat the basal ganglia evolved very early in vertebrateevolution, probably at about the same time as theappearance of the first true vertebrates. Most of theremainder of the motor system also reflects ourinheritance from our distant vertebrate ancestors. Inaddition to the basal ganglia, the superior colliculus,the red nucleus, the reticulospinal system, CPGs, andthe alpha motor neurons retain their basic organiza-tional patterns and functions from the earliest historyof the vertebrate brain. The evolution of the cerebel-lum remains less certain, but it clearly emergedrelatively early in the history of vertebrates as well.What has changed in the past 500 million years or

so? Of course, many of the basic components of themotor system have been elaborated and modified inmany ways. In addition, an entirely new componenthas been added. In contrast to the ancient componentsof the motor system enumerated previously, the motorareas of neocortex have evolved much more recently.In a form that people can recognize as distinctivelyneocortical, a clear-cut motor cortex appears only inmammals. Thus, a separate motor cortex appears to

MOTOR CONTROL20


have evolved only approximately 180 million yearsago, a short time by evolutionary standards, especiallycompared to the ancient lineage of the motor system’sother main components. The relatively recent devel-opment of themotor cortex suggests that an importantaspect of motor cortex function involves the modula-tion and control over subcortical motor systems,including the introduction of novel levels of flexibilityin the motor system, such as transformational andarbitrary mapping, and the fractionation of relativelyhardwired motor synergies.

See Also the Following Articles

Suggested Reading

Bhushan, N., and Shadmehr, R. (1999). Computational architecture

of the human adaptive controller during learning of reaching

movements in force fields. Biol. Cybernetics 81, 39–60.

Bloedel, J. R., Ebner, T. J., and Wise, S. P. (Eds.) (1996). The

Acquisition of Motor Behavior in Vertebrates. MIT Press,

Cambridge, MA.

Goldberg, M. E., Eggers, H. M., and Gouras, P. (1999). The ocular

motor system. InPrinciples of Neural Science (E. R.Kandel, J. H.

Schwartz and T. M. Jessel, Eds.), 4th ed. Elsevier, New York.

Joseph, A. B., and Young, R. R. (Eds.) (1992). Movement Disorders

in Neurology and Neuropsychiatry. Blackwell, Boston.

Leiguarda, R. C., and Marsden, C. D. (2000). Limb aprax-

iasFHigher-order disorders of sensorimotor integration. Brain

123, 860–879.

Loewy, A. D., and Spyer, K.M. (Eds.) (1990). Central Regulation of

Autonomic Functions. Oxford Univ. Press, Oxford.

Middleton, F. A., and Strick, P. L. (2000). A revised neuroanatomy

of frontal subcortical circuits. In Frontal–Subcortical Circuits in

Psychiatry and Neurology (D. G. Lichter and J. L. Cummings,

Eds.), Guilford, New York.

Milner, A. D., and Goodale, M. A. (1996). The Visual Brain in

Action. Oxford Univ. Press, Oxford.

Passingham, R. E. (1993). The Frontal Lobes in Voluntary Action.

Oxford Univ. Press, Oxford.

Porter, R., and Lemon, R. (1993). Corticospinal Function and

Voluntary Movement. Clarendon, Oxford.

Stein, P. S. G., Grillner, S., Selverston, A. I., and Stuart, D. G. (Eds.)

(1997). Neurons, Networks, and Motor Behavior. MIT Press,

Cambridge, MA.

MOTOR CONTROL 21


Documents

Motor Control