86:2125-2143, 2001. J NeurophysiolMarc H. Schieber

You might find this additional information useful...

120 articles, 58 of which you can access free at: This article cites http://jn.physiology.org/cgi/content/full/86/5/2125#BIBL

27 other HighWire hosted articles, the first 5 are: This article has been cited by

  [PDF]  [Abstract]

, April 1, 2008; 14 (2): 195-202. NeuroscientistK. T. Reilly and A. Sirigu

The Motor Cortex and Its Role in Phantom Limb Phenomena 

[PDF]  [Full Text]  [Abstract], September 1, 2008; 18 (9): 1999-2009. Cereb Cortex

Glahn and P. T. Fox M. D. Cykowski, O. Coulon, P. V. Kochunov, K. Amunts, J. L. Lancaster, A. R. Laird, D. C.

Depth AsymmetryThe Central Sulcus: an Observer-Independent Characterization of Sulcal Landmarks and 

[PDF]  [Full Text]  [Abstract], September 1, 2008; 100 (3): 1433-1454. J Neurophysiol

A. d'Avella, L. Fernandez, A. Portone and F. Lacquaniti Modulation of Phasic and Tonic Muscle Synergies With Reaching Direction and Speed

  [PDF]  [Full Text]  [Abstract]

, February 11, 2009; 29 (6): 1915-1927. J. Neurosci.D. M. Griffin, H. M. Hudson, A. Belhaj-Saif and P. D. Cheney

Stability of Output Effects from Motor Cortex to Forelimb Muscles in Primates 

[PDF]  [Full Text]  [Abstract], March 11, 2009; 29 (10): 3132-3137. J. Neurosci.

K. J. Miller, S. Zanos, E. E. Fetz, M. den Nijs and J. G. Ojemann Finger Movements in Humans

Decoupling the Cortical Power Spectrum Reveals Real-Time Representation of Individual

on the following topics: http://highwire.stanford.edu/lists/artbytopic.dtlcan be found at Medline items on this article's topics

Veterinary Science .. Motor Cortex

including high-resolution figures, can be found at: Updated information and services http://jn.physiology.org/cgi/content/full/86/5/2125

can be found at: Journal of Neurophysiologyabout Additional material and information http://www.the-aps.org/publications/jn

This information is current as of March 26, 2009 .  

http://www.the-aps.org/.American Physiological Society. ISSN: 0022-3077, ESSN: 1522-1598. Visit our website at (monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright © 2005 by the

publishes original articles on the function of the nervous system. It is published 12 times a yearJournal of Neurophysiology

on March 26, 2009

jn.physiology.orgD

ownloaded from

Constraints on Somatotopic Organization in the PrimaryMotor Cortex

MARC H. SCHIEBERDepartments of Neurology, Neurobiology and Anatomy, Brain and Cognitive Science, and Physical Medicine andRehabilitation, the Center for Visual Science, and the Brain Injury Rehabilitation Program at St. Mary’s Hospital,University of Rochester School of Medicine and Dentistry, Rochester, New York 14642

Received 9 January 2001; accepted in final form 5 July 2001

Schieber, Marc H. Constraints on somatotopic organization in theprimary motor cortex.J Neurophysiol86: 2125–2143, 2001. Since the1870s, the primary motor cortex (M1) has been known to have asomatotopic organization, with different regions of cortex participat-ing in control of face, arm, and leg movements. Through the middleof the 20th century, it seemed possible that the principle of somato-topic organization extended to the detailed representation of differentbody parts within each of the three major representations. The armregion of M1, for example, was thought to contain a well-ordered,point-to-point representation of the movements or muscles of thethumb, index, middle, ring, and little fingers, the wrist, elbow, andshoulder, as conveyed by the iconic homunculus and simiusculus. Inthe last quarter of the 20th century, however, experimental evidencehas accumulated indicating that within-limb somatotopy in M1 is notspatially discrete nor sequentially ordered. Rather, beneath gradualsomatotopic gradients of representation, the representations of differ-ent smaller body parts or muscles each are distributed widely withinthe face, arm, or leg representation, such that the representations ofany two smaller parts overlap extensively. Appreciation of this un-derlying organization will be essential to further understanding of thecontribution to control of movement made by M1. Because no singleexperiment disproves a well-ordered within-limb somatotopic organi-zation in M1, here I review the accumulated evidence, using a frame-work of six major features that constrain the somatotopic organizationof M1: convergence of output, divergence of output, horizontal inter-connections, distributed activation, effects of lesions, and ability toreorganize. Review of the classic experiments that led to developmentof the homunculus and simiusculus shows that these data too wereconsistent with distributed within-limb somatotopy. I conclude withspeculations on what the constrained somatotopy of M1 might tell usabout its contribution to control of movement.

I N T R O D U C T I O N

Somatotopic organization long has been the hallmark of theprimary motor cortex (M1). The concept of a cortical regionsystematically organized to control movements of differentbody parts was first hypothesized by Hughlings Jackson in the1870s, based on his observations of certain epileptic patients inwhom convulsive movements systematically marched fromone part of the body to adjacent parts (Jackson 1958). Theexistence of such a cortical region was demonstrated contem-poraneously by Fritsch and Hitzig using electrical stimulation

of the canine cortex, one of the earliest demonstrations of aspecific function of a particular cortical region (Walshe 1948).As techniques for electrical stimulation improved, increasinglydetailed maps of body part representation in M1 became avail-able, culminating in the well-known summary diagrams ofPenfield’s homunculus (Penfield and Rasmussen 1950) andWoolsey’s simiusculus (Woolsey et al. 1952). These icons ofneuroscience commonly are interpreted as showing a system-atic, spatially organized, point-to-point mapping of control ofdifferent body parts by different pieces of M1 cortex (Schott1993). Indeed, in its ultimate form, Penfield’s homunculusincluded a line representing the mediolateral ribbon of M1,broken into sequential line segments representing differentbody parts, down to different segments for the thumb, index,middle, ring, and little fingers.

In the last quarter of the 20th century, however, experimen-tal evidence has accumulated indicating that the control ofdifferent body parts from M1 is not nearly so somatotopicallyorganized as the homunculus and simiusculus seem to suggest.While it remains clear that the head, upper extremity, andlower extremity have sequential and largely separate represen-tations, the representations of smaller body parts are widelydistributed within these major regions. In retrospect, data ob-tained from the 1870s to the present can be seen to be consis-tent with this distributed organization as well. Consequently,the territory controlling one body part overlaps extensivelywith the territory controlling adjacent body parts. For example,the M1 territory controlling the thumb overlaps extensivelywith the territories controlling the fingers.

Here I review this evidence in a framework of six factorsthat constrain the somatotopic organization of M1.1) Conver-gent output from a large M1 territory controls any particularbody part, joint or muscle.2) Divergentoutput of many singleM1 neurons reaches multiple spinal motoneuron pools.3)Horizontal connectionsinterlink the cortex throughout a majorbody part region.4) Widely distributed activityappears in amajor body part region whenever any smaller body part ismoved.5) Partial inactivationof a major region affects mul-tiple smaller body parts simultaneously.6) Plasticity limits thedegree to which control of a specific body part can be assignedto a particular piece of cortex. Although I will deal mainly with

Address for reprint requests: University of Rochester Medical Center, Dept.of Neurology, 601 Elmwood Ave., Box 673, Rochester, NY 14642 (E-mail:mhs@cvs.rochester.edu).

The costs of publication of this article were defrayed in part by the paymentof page charges. The article must therefore be hereby marked ‘‘advertisement’’in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

INVITED REVIEW

21250022-3077/01 $5.00 Copyright © 2001 The American Physiological Societywww.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

the upper extremity region (from which the most experimentalevidence is available), these six factors appear to apply as wellto the representations of the face and lower extremity. No onefactor alone unequivocally disproves a detailed within-limbsomatotopy, nor can any single experiment. Yet consideredaltogether, they compel us to conclude that control of each partof the upper limb, lower limb, or face is widely distributedwithin the overall representation. To progress in understandingM1’s contribution to motor control, we must consider theimplications of these constraints on the somatotopic organiza-tion of M1.

C O N V E R G E N C E

Outputs from large territories of M1 converge on the spinalmotoneuron pool of any given muscle. The cortical territory foreach muscle is so large as to preclude spatially separate terri-tories for each muscle. Instead, the M1 territories from whichoutputs converge on two upper extremity muscles overlapextensively. This principle of convergence was articulatedmost precisely by the work of Charles Phillips and his collab-orators (described in the following sub-sections), but all studiesof movements evoked by stimulation of M1 have been consis-tent with such convergence and overlap, from the classicalstudies with cortical surface stimulation that led to the homun-culus and simiusculus, to more recent studies using intracorti-cal microstimulation (ICMS).

Classical studies employing stimulation of the corticalsurface

Because the classical studies that employed stimulation ofthe cortical surface commonly are assumed to have demon-strated a detailed within-limb somatotopic organization, I be-gin by reviewing exactly what was demonstrated in thesestudies. By modern standards, the electrical stimuli employedin these studies were intense and prolonged, exciting relativelylarge regions of cortex, and evoked overt movements ratherthan the brief flicks and twitches evoked by ICMS. Penfieldand Boldrey (1937) published a map of 77 precentral locationsfrom which cortical surface stimulation elicited movements ofthe different digits of the hand in studies of 126 human subjects(Fig. 1A). The overall region from which stimulation producedfinger movements extended 55 mm along the central sulcus.Inspection of their figure shows that thumb movements wereelicited at both the lateral and medial limits of this region, aswere movements of the little finger, and of the other digits aswell. Furthermore, comparing the region from which fingermovements were evoked with the region from which armmovements were evoked showed large, extensively overlap-ping territories representing different proximodistal parts of theupper extremity (Fig. 1B).

Because data from multiple subjects were compiled in thesemaps, inter-individual variation might have accounted for thelarge and overlapping territories of different digits and more

FIG. 1. Convergence and overlap in Penfield’s data. Enlargements are shown from Figs. 12A and 25B of Penfield and Boldrey(1937). The region enlarged is indicated by the rectangle drawn on theinsettaken from their standardized map of the hemisphere,but note that whereas the region shown inB extends laterally to the Sylvian fissure and therefore includes the face representation,the region shown inA does not.A: locations from which finger movements were elicited in data compiled from 126 patients. If onlycertain digits moved, they are indicated with Roman numerals: I5 thumb through V5 little finger. Black dots indicate locationswhere stimulation elicited movement of all the digits. Note that, contrary to the discrete order implied by the homunculus, thumbmovements were elicited medially as well as laterally, and little finger movements were elicited laterally as well as medially.B:outlines encompass the total territory from which movements of the fingers (E E E), entire hand (¦ ¦ ¦), or more proximal arm(1 – 1 – 1) were evoked. Note the overlap of distal and proximal representations. (Reproduced with permission of the LiteraryExecutors of Wilder Penfield)

2126 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

proximal parts of the upper extremity. In single subjects, anorderly, segregated somatotopic arrangement might have beenapparent. Inspection of records from single patients reveals,however, that such was not the case. Figure 2 shows, forexample, detailed results of intraoperative stimulation in onepatient studied by Penfield and colleagues. Although an overallsomatotopic trend was apparent in this single case, with move-ments of the digits being evoked more often laterally along theRolandic fissure and movements of more proximal parts of theupper extremity being evoked more often medially, move-ments of different parts were not elicited from discrete loca-tions arrayed in simple somatotopic order. The thumb, forexample, was involved in the movements produced by stimu-lation at three points along the central sulcus, and at each ofthese points stimulation evoked movements of other digits as

well (points marked by upside-down L, N, M). Finger move-ments were elicited from two more medial points as well (Zand O), with the most medial of these (O) surrounded by otherpoints from which more proximal arm movements wereevoked (7, R, X). Thus a well-ordered somatotopic represen-tation of the upper extremity was not evident in the details ofsingle cases such as this.

Penfield and Rasmussen (1950, p. 56), commenting on theirhomunculus, noted: “A figurine of this sort cannot give anaccurate indication of the specific joints in which movementtakes place, for in most cases movement appears at more thanone joint simultaneously. . . . Themotor homunculus may beused as an aid to memory in regard to movement sequence andthe relative extent of cortex in which such movement findsrepresentation. It is a cartoon of representation in which sci-

FIG. 2. Details of intraoperative stimulation in an individual human patient. The sketch reproduced here from case recordsshows the borders of the craniotomy (cross-hatching) and locations of paper markers placed at stimulated locations (upside-downletters and numbers), both drawn by Penfield and colleagues on their standard map of the hemispheric surface based on theirintraoperative photograph. To this reproduced sketch, I have added selected details from the transcribed intraoperative notesrecording the results of stimulation at each location, linked to each marker by a straight line. No response to stimulation wasobtained at point B, 10, or 11; and no transcribed note was available for point H or 2. Note that, although movements of the digitswere generally obtained laterally and more proximal movements medially, movements of different digits or more proximal partsof the extremity were not obtained from separate, somatotopically ordered points. This particular case was chosen here because ofthe relatively large number of points stimulated along the precentral gyrus (in many other cases, because of the possible risk ofsetting off seizures by stimulation of the motor cortex, many points along the postcentral gyrus were stimulated, and only a fewalong the precentral gyrus; personal communication, Dr. William Feindel), and because no cortical lesion was evident at surgery.The only lesion found was a meningeal cicatrix close to the midline (in the sketch reproduced here, the wandering dashed line closeto the midline delimited a territory of “whitening of the arachnoid”). Although a detailed description of the patient’s typical seizureswas not available, he had “Jacksonian epilepsy” with an epigastric aura. This patient (MO) was described by Penfield andRasmussen (1950, their Fig. 32) in reference to eye turning. Transcribed operative notes, an intraoperative photograph of the brainwith paper markers placed on the cortical surface, and the sketch reproduced here, all were provided courtesy of the PenfieldArchive (Montreal Neurological Institute, Dr. William Feindel, Curator).

2127CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

entific accuracy is impossible.” Although the overlapping rep-resentations of adjacent body parts observed by Penfield andcolleagues might have resulted from current spread across anunderlying discrete and orderly somatotopic representation, theoverlap also could have been a genuine feature of the under-lying representation in M1.

The detailed results of similar studies on a rhesus monkeyand on a human from the work of Woolsey and colleagues areillustrated in Fig. 3. In both species, evoked movements typi-cally involved more than one digit and/or more proximal joint.In both species, movements involving the thumb were elicitedby stimuli delivered at different locations scattered over muchof the upper extremity representation. Similarly, stimulation atmany different locations elicited movements involving thelittle finger. Although the thumb appears more heavily repre-sented in the lateral aspect of the upper extremity representa-tion and the little finger appears more heavily representedmedially, the territory in which evoked movements involvedthe thumb overlaps considerably with the territory in whichevoked movements involved the little finger. As with Pen-field’s studies, given the possible spread of stimulating current,Woolsey’s data would be consistent either with discrete, so-matotopically segregated representations of the thumb andlittle finger, or with overlapping representations in which out-puts to the muscles serving each digit converge from largecortical territories. The same argument would apply to otherpairings of digits, or pairings of digit and wrist, wrist andelbow, and elbow and shoulder. In the text bracketing the

motor simiusculi (their Fig. 131), Woolsey and colleagueswrote, “It must be emphasized . . . that this diagram is aninadequate representation of the localization pattern, since in aline drawing one cannot indicate the successive overlap whichis so characteristic a feature of cortical representation. . . .”(Woolsey et al. 1952, p. 252).

While the examples illustrated above come from the work ofPenfield’s group and Woolsey’s group, similar evidence con-sistent with convergence and overlap was present in the de-tailed results of other investigators who employed corticalsurface stimulation in systematic exploration of M1. The num-ber of such studies is too large for each to be mentioned here,but some additional examples may illustrate two general fea-tures of this literature. First, the impression of discrete soma-totopic order versus convergence and overlap varied with thenumber of points stimulated. Stimulating a limited number ofwidely spaced points along the central sulcus often demon-strated a progression from shoulder movements medially tofinger and thumb movements laterally (Bucy 1949; Fulton andKeller 1932). Even in these studies, however, some pointsfailed to follow a strict somatotopic order. In studies samplinga larger number of points, convergence and overlap becamemore apparent. In studies of anthropoid apes, for example,Leyton and Sherrington (1917) stimulated a relatively largenumber of points in each animal studied, and listed 135 dif-ferent combinations of primary, secondary, tertiary, and qua-ternary evoked upper extremity movements; these studies showconsiderable overlap of the representation of different joints

FIG. 3. Convergence and overlap in Woolsey’s data. Maps of evoked movements obtained by Woolsey and colleagues (A) ina monkey (Macaca irus) (Woolsey et al. 1952), and (B) in a human (Woolsey et al. 1979). Inset figures of an entire brain indicatethe areas enlarged. InA, dashed lines indicate areas in the anterior bank of the central sulcus (right) and posterior bank of the arcuatesulcus (left) exposed for stimulation. In bothA andB, figurines show which parts of the body moved on stimulation at each location,with black shading indicating the most vigorous, cross-hatching intermediate and stippling the least movement. In both species, theterritories from which movements of the thumb and different fingers were evoked were large and extensively overlapping.(Modified from Schieber 1990)

2128 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

and movements. Here, then, is a second general feature: focus-ing on only the initial or most prominent elicited movementwas more revealing of somatotopic order, whereas attending toall the movements elicited by stimulation at each point sug-gested more extensive convergence and overlap (Beevor andHorsley 1887; Ferrier 1873; Hines 1940; Murphy and Gellhorn1945). For example, by focusing only on the primary move-ment evoked by stimulation at each site, and comparing non-adjacent joints (e.g., shoulder vs. fingers), Leyton and Sher-rington demonstrated a gradual somatotopic progressionconsistent with the homunculus and simiusculus (e.g., theirFigs. 16 and 17), even though their data are consistent withextensive overlap when all movements of all joints were con-sidered.

Studies in which muscle contraction was measured duringcortical surface stimulation, instead of observing evokedmovement, also were consistent with convergence and overlap.Recording the tension developed by a number of monkeyhindlimb muscles, for example, revealed that cortical surfacestimulation only occasionally evoked contraction of one of therecorded muscles alone; much more often, multiple musclescontracted simultaneously, although the cortical locations fromwhich maximal contraction was evoked differed from muscleto muscle (Chang et al. 1947). Similar results in humans havebeen obtained in recent years by recording compound muscleactions potentials in response to transcranial magnetic stimu-lation (Krings et al. 1998; Wassermann et al. 1992).

Were convergence and overlap artifactual?

Until the 1970s, much if not all of this evidence of conver-gence from large and overlapping territories moving differentbody parts could have been attributed to the spread of stimuluseffects, for two reasons. First, relatively large stimulating cur-rents (on the order of 0.5–1.5 mA) had to be used at the corticalsurface to evoke movements; in comparison, currents an orderof magnitude smaller evoked movements when applied to aperipheral nerve. The large currents applied to a point at thecortical surface inevitably spread through a considerable vol-ume of tissue. At threshold for evoking simple flick move-ments (e.g., 10-ms pulses of 0.5–1.5 mA), for example, surfacestimulation evoked repetitive discharge in Betz cells up to 4mm horizontally distant from the surface point stimulated(Phillips 1956; see also Asanuma et al. 1976; Jankowska et al.1975a). Direct excitation of corticospinal neurons by surfacestimulation thus occurred within a rather large area around thestimulated point, but even a 4-mm horizontal spread of directBetz cell excitation could not account for the observed overlapof up to 55 mm.

Second, corticospinal neurons at an even greater horizontaldistance in theory could be excited indirectly. Single electricalstimuli delivered at the cortical surface evoked multiple de-scending volleys in the corticospinal tract. The earliest volley(D-wave) was produced bydirect excitation of corticospinalneurons; later descending volleys (I-waves) resulted from ex-citation of intracortical neurons whichindirectly (trans-synap-tically) excited the corticospinal neurons (Patton and Amassian1954). D-waves were evoked by current spread from the pointof surface stimulation through the superficial cortical layers,exciting the corticospinal neuron somata in layer V (Patton andAmassian 1954), or their axons still deeper (Landau et al.

1965). At threshold for direct activation of corticospinal neu-rons, then, more superficial cortical interneurons were excitedas well. These interneurons could excite corticospinal neuronsnot only directly beneath the stimulating electrode, but alsolateral to the electrode (seeHorizontal interconnections, be-low). Horizontal spread through transynaptic excitation of cor-ticospinal neurons might artifactually enlarge the cortical ter-ritories from which a given movement was evoked, producingeven more overlap. To limit such horizontal spread of excita-tion, vertical incisions in the cortex could be made to isolatesmall (33 5 mm) islands of cortex; however, this experimentalmanipulation failed to eliminate the extensively overlappingterritories (Murphy and Gellhorn 1945).

Nevertheless, it remained possible that if only a few, closelypacked corticospinal neurons could be excited directly, themap of evoked movements would resolve into discrete territo-ries for different movements or muscles. This possibility di-minished, however, when Phillips and co-workers found thatbrief (0.2-ms) low-amplitude, surface-anodal stimuli directlyexcited corticospinal neurons without indirect excitation (Hernet al. 1962). Recording intracellularly from baboon cervicalmotoneurons, and accounting for current spread in the cortex,they used such stimuli to demonstrate that the colony of cor-ticospinal neurons projecting monosynaptically to a singlecervical motoneuron must, in many instances, be spread over acortical territory of at least several square millimeters (thelargest minimal territory they measured covered 20 mm2)(Landgren et al. 1962). Moreover, the minimal territories con-taining corticospinal neurons projecting to radial, ulnar, ormedian nerve motoneurons (i.e., innervating different muscles)often overlapped. Thus even single motoneurons were shownto receive converging input from relatively large cortical ter-ritories, which overlap with the territories providing input tomotoneurons of other muscles.

Studies employing ICMS

In the late 1960s and early 1970s, Asanuma and colleaguesdeveloped the technique of ICMS. Rather than stimulating witha large electrode touching the pial surface of the cortex, amicroelectrode was advanced into the M1 cortex and posi-tioned close to layer V. Here, 0.2-ms pulses of only a fewmicroamperes, delivered in trains of 10–12 pulses at approx-imately 300 Hz, could evoke visible movement or recordableelectromyographic (EMG) activity. Single 10-mA, 0.2-mscathodal current pulses delivered in layer V were estimated todirectly excite neuronal somata within a radius of only 88mm,which in cat anterior sigmoid gyrus would encompass onlyabout 28 pyramidal neurons (Stoney et al. 1968). Initial studieswith ICMS indicated that a particular movement of a part of theforelimb, or contraction of a particular muscle, was evoked bythreshold ICMS applied within a small columnar zone ofapproximately 0.5–1 mm radius (Asanuma and Rosen 1972).Multiple small efferent zones scattered in the overall forelimbrepresentation could be found for the same movement or mus-cle. Discrete efferent zones representing different movementsor muscles appeared intermingled like the different colors oftiles in a mosaic.

The initial report of Asanuma and Rosen (1972) showed thismosaic arrangement by superimposing data from 10 Cebusmonkeys (their Fig. 8). Subsequent studies from numerous

2129CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

laboratories in many species, including detailed studies ofsingle subjects, have continued to show that maps of thresholdresponses evoked by ICMS include the same features. Twoexamples are shown in Fig. 4. In anesthetized owl monkeys(Fig. 4A), although a general within-limb somatotopic gradientcould be appreciated (distal representation stronger posterolat-erally and proximal representation stronger anteromedially),movement of a given part (such as the digits) was evoked byICMS at multiple foci scattered over a considerable portion ofthe upper extremity representation, and these foci were inter-mingled with points at which stimulation elicited movementsof other parts of the limb (Gould et al. 1986). In awakestump-tailed monkeys (Macaca arctoides,Fig. 4B), althoughthe thumb had more representation laterally, movement of anygiven digit was evoked by ICMS at foci scattered over a largeportion of the upper extremity representation, and the territoryfrom which ICMS evoked movement of a given digit over-lapped with the territory from which ICMS evoked movementof any other digit (Kwan et al. 1978). A similar pattern ofscattered threshold foci for individual muscles, intermingledwith foci for other muscles, has been described by recordingevoked EMG activity during ICMS in squirrel monkeys (Do-noghue et al. 1992; Strick and Preston 1978, 1982), macaques(Humphrey 1986), and baboons (Waters et al. 1990). Thuseven with threshold ICMS, a particular movement of a givenbody part, or a contraction of a specific muscle, is evoked by

stimulation at several sites scattered in the forelimb represen-tation, and these sites are intermingled with sites where stim-ulation evokes other movements of the same body part, ormovements of adjacent body parts, or contractions of othernearby muscles. At the same time, gradual somatotopic gradi-ents—indicating gradual shifts in the part(s) most heavilyrepresented—can often be appreciated in ICMS maps of theforelimb representation. Similar features appear in ICMS mapsof the face representation as well (Huang et al. 1988).

What is revealed with ICMS?

At this point, one might come away with the interpretationthat M1’s somatotopic organization consists of a mosaic ofsmall discrete zones, with each movement or muscle repre-sented in multiple scattered zones. The large and overlappingcortical territories demonstrated by the older methods of cor-tical surface stimulation then could have resulted from stimu-lation of many of these tiny zones at the same time. The extentto which the efferent zones mapped by threshold ICMS areactually discrete, however, is called into question by two majorconsiderations.

First, even with ICMS, stimulation does not occur entirely ata single point. Many of the corticospinal neurons that dischargein response to ICMS are excited directly at the soma or axonhillock, and these neurons do indeed lie within a small zone

FIG. 4. Intracortical microstimulation maps.A: a selected portion is shown of the threshold intracortical microstimulation(ICMS) map of the left hemisphere M1 in an anesthetized owl monkey, a species in which M1 lies on the surface of a relativelylissencephalic cortex. Black dots mark stimulated points, and solid lines surround groups of points where stimulation evokedmovements of the same body part. Added colors emphasize regions where threshold ICMS evoked movements of different partsof the upper extremity. Note that movements of a given part of the upper extremity, exemplified by the digits, were evoked fromseveral scattered foci, intermingled with foci from which movements of other parts were elicited. At the same time, an overallsomatotopic gradient [with heavier representation of the more distal parts (digits, wrist, and forearm) posterolaterally; and heavierrepresentation of the more proximal parts (shoulder and elbow) anteromedially] can be appreciated. The rectangle on thehemispheric outline inset at bottom right indicates the region enlarged. ANK, ankle; CH, chin; Dig, digits; EL, elbow; FA, forearm;M, mouth; NO, nose; SH, shoulder; TR, trunk; VIB, vibrissae; W, wrist (modified from Gould et al. 1986).B: map of locationsfrom which ICMS evoked movements of different fingers in an awake stump-tailed monkey. Different digits are indicated by 15thumb through 55 little finger. Letters indicate the type of movement: f, flexion; e, extension; a, adduction; b, abduction. Flexionof digits 2 through 5, for example, is denoted “25f.” Note that representation of movements of different digits are largelyintermingled, although the thumb tended to have a heavier representation more laterally (down). The anterior bank of the left centralsulcus has been unfolded, with the depth of the sulcus represented by the solid vertical line at right (length, 10 mm), and the lipof the anterior bank represented by the dotted vertical line. Dashed vertical lines represent the borders between Brodmann’s areas,and the outer dashed curve encloses the entire upper extremity representation (redrawn from Kwan et al. 1978).

2130 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

close to the electrode tip. As noted above, in cat M1, low-amplitude ICMS pulses have been estimated to directly exciteon the order of 28 pyramidal neuron somata within a radius of88 mm (Stoney et al. 1968). In the baboon, a 0.2-ms3 5-mApulse delivered in layer V was estimated to directly excite90–900 small and 1–5 large pyramidal neurons within a radiusof 40–125mm, whereas at 90mA, a generous estimate of theeffective spread of stimulating current was only 0.6 mm(Andersen et al. 1975). Direct excitation of neuronal somata oraxon hillocks by ICMS thus is reasonably focal.

Shortly after ICMS came into use, however, investigatorsrealized that the same pulses could excite additional cortico-spinal neurons at greater distances through two mechanisms.One mechanism is direct excitation of the intracortical collat-erals of pyramidal tract axons, which may extend horizontallyup to 1 mm away from the soma (Asanuma et al. 1976).1 Asecond mechanism is indirect,trans-synaptic excitation. In-deed, the greatest part of the descending volley produced byICMS results from suchtrans-synaptic excitation of cortico-spinal neurons, even when the stimulating electrode is withinlayer V and currents are as low as 5mA or less (Jankowska etal. 1975b). Moreover, repetitive ICMS of the same intracorticalpoint at frequencies of 200–400 Hz, the type of stimulationneeded to evoke detectable movement or EMG activity in thestudies described above, produces powerful temporal summa-tion of this trans-synaptic excitation (Asanuma et al. 1976;Jankowska et al. 1975b). Although mosttrans-synapticallyexcited corticospinal neurons probably lie relatively close tothe ICMS microelectrode, some somata may be more distant.Horizontally extending axon collaterals can produce excitatorypostsynaptic potentials (EPSPs) in M1 pyramidal tract neuronswithin a 1- to 2-mm radius (Asanuma and Rosen 1973; Mat-sumura et al. 1996), and intracortical microstimulation canexcite some pyramidal tract neurons within this radius (Bakeret al. 1998). While most of the effects of ICMS result fromexcitation of pyramidal tract neurons quite close to the micro-electrode, a penumbra of other pyramidal tract neurons may beexcited as well. Quantitative estimates of what fraction ofobserved muscle contraction or movement results from directexcitation of local somata versus excitation of penumbral neu-rons are not available.

A modified ICMS technique recently has been developedthat avoids the temporal summation of repetitive stimulation athigh frequency (;300 Hz) used in conventional ICMS toproduce detectable output effects in quiescent animals. Instimulus-triggered averaging (StTA), single ICMS pulses aredelivered at much lower frequencies (10–20 Hz) while awakemonkeys perform active movements, and triggered averagingthen is used to extract the effects of these pulses from ongoingvoluntary EMG activity (Cheney and Fetz 1985). Because thetemporal summation of EPSPs is eliminated, StTA probablyproduces much less of its effect via indirect,trans-synapticexcitation. Yet maps made with StTA continue to show largecortical territories for individual muscles that overlap exten-sively with the cortical territories of other muscles (Park et al.2001).

This brings us to the second consideration: exactly what is

being mapped with threshold electrical stimulation? ThresholdICMS mapping in M1 entails placing the electrode tip at acertain point in or near layer V, and gradually adjusting stim-ulus strength until on half of stimulation trials the discharge ofsome motor units is just detected, either by recording EMGactivity, or by having enough motor units discharge to producean externally observable movement. With either assay, theexperimental observation means that the evoked output fromM1 to a particular muscle (or potentially a combination ofmuscles when observing movement) was greater than the out-put to other muscles, not that output occurred to that musclealone. Output may well have occurred to motoneurons of othermuscles; such output to other muscles simply was insufficientto cause them to discharge (or to discharge enough to produceobservable movement). The apparently discrete zones of ICMSmaps obtained with threshold stimuli thus represent the quan-titatively greatest outputs, not qualitatively exclusive outputs.

What happens, then, if the stimulus strength is increasedbeyond threshold? Are outputs to additional muscles revealed?Phillips and colleagues recorded simultaneously from singlemotor units in the thenar eminence (thumb muscles), in the firstdorsal interosseous muscle (FDI, an index finger muscle), andin extensor digitorum communis (EDC, which extends the 4fingers)2 while using ICMS to map the forelimb region ofbaboon M1 (Andersen et al. 1975). Threshold stimulation atmost points evoked discharge in only one of the three motorunits. Using currents up to 80mA, however, they found that thethree motor units recorded from these three muscles each couldbe brought to discharge with ICMS at many points spread overa wide cortical territory, and that the total territories fromwhich each motor unit could be discharged overlapped exten-sively (Fig. 5). Their calculations showed that spread of thehigher currents did not account for the overlap. Even at 90mA,a current larger than they routinely employed, a generousestimate of the spread of current effective for direct stimulationof somata and axons was only 0.6 mm, while the motor unitterritories overlapped several millimeters. Hence the colonyof Betz cells whose output excited each motor unit necessarilywas spread over a considerable cortical territory, largely inter-mingled with the colonies of Betz cells exciting the motorunits of the other muscles.3 Similar findings were obtainedwith intracellular recordings from hindlimb motoneurons(Jankowska et al. 1975a).

Subsequently, several investigators have confirmed thatICMS maps show multiple scattered loci from which thresholdstimulation evokes movement about a particular joint, or EMGactivity in a particular muscle. In between are scattered thresh-old loci for other movements or muscles, forming a “complexmosaic.” As stimulus intensity is increased systematicallyabove threshold, however, movements are produced at addi-tional joints (Sessle and Wiesendanger 1982), or contractions

1 Indeed, recent studies of electrical excitation in slices of rat visual cortexindicate extracellular stimulation excites axonal branches more readily thanaxon initial segments or somata (Nowak and Bullier 1998a,b).

2 The large spatial extent of the “colony” of layer V neurons projecting tothe EDC motoneuron pool of macaques recently has been demonstratedanatomically by retrograde transneuronal transport of rabies virus injected intothe EDC muscle belly (Rathelot and Strick 2000).

3 These authors did not consider excitation of additional, penumbral neuronsvia horizontal axon collaterals or indirecttrans-synaptic excitation. Even if oneconsiders a 2-mm penumbra of lesser excitation, however, the overlap theydemonstrated for cortical territories of different muscles is too extensive toattribute entirely to spread of excitation and therefore indicates interminglingof corticospinal neurons exciting different muscles.

2131CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

are evoked in more and more muscles (Donoghue et al. 1992),so that the loci for any particular muscle tend to expand andcoalesce, revealing the large total territory representing thatmuscle, which overlaps extensively with the territories repre-senting other muscles (Humphrey 1986; Sato and Tanji 1989).This expansion and coalescence into large and overlappingterritories cannot be attributed entirely to current spread andindirect, trans-synaptic excitation. Thus ICMS, like surfacestimulation, indicates convergence of M1 outputs from largeand overlapping M1 territories onto different muscles or move-ments.

Because even ICMS involves some spread of current tomultiple neurons, and because such current can excite neuronsboth directly and indirectly, the question of whether stimula-tion at a single “point” in M1 actually produces output to morethan one muscle cannot be resolved with extracellular electricalstimulation. The ideal experiment for resolving this question(recording intracellularly from the spinal motoneurons of sev-eral muscles while stimulating intracellularly in the somata ofsingle corticospinal neurons in M1) remains technically inac-cessible even now. In the 1980s, however, separate lines ofevidence developed rendering much of the previous argumentsmoot.

D I V E R G E N C E

For many years neuroscientists generally believed that agiven corticospinal neuron made monosynaptic connections tothe motoneurons of only one muscle. These specific connec-tions to particular muscles enabled the “upper motor neurons”in M1 to selectively activate the muscles needed to perform

fine, relatively independent movements (Phillips and Landau1990). Shortly after Phillips and colleagues had articulated theconcept of convergence from wide M1 territories onto spinalmotoneurons, evidence appeared demonstrating that the outputprojections from single M1 neurons often diverge to innervatethe motoneuron pools of more than one muscle.

Anatomic evidence of such divergence was obtained byfilling single corticospinal axons with horseradish peroxidase(HRP), revealing that collateral branches of a single cortico-spinal axon often ramified over several spinal segments pro-viding terminal arbors in the motoneuron pools of up to fourmuscles (Fig. 6A) (Shinoda et al. 1981). Physiologic evidenceof divergence came from the use of spike-triggered averagingof rectified EMG activity to identify functional, short-latencyconnections from M1 neurons to spinal motoneuron pools (Fig.6B) (Fetz and Cheney 1978, 1980). Many single M1 neuronsproduced postspike effects, indicative of relatively direct cor-ticomotoneuronal connections, in up to six different forearmmuscles. Spike-triggered averaging also has shown divergentoutputs from M1 neurons controlling the intrinsic muscles ofthe hand; those used in the finest of relatively independentmovements (Buys et al. 1986; Lemon et al. 1986). Further-more, a recent study indicates that the functional connectionsof single M1 neurons may diverge, not only to different mus-cles moving the fingers and wrist, but also to muscles movingthe elbow and shoulder (McKiernan et al. 1998). These diver-gent projections from single M1 neurons obviously constrainthe degree to which M1’s output can be organized in a strictwithin-limb somatotopy. The set of muscles receiving theoutput of a single M1 neuron may act on multiple fingers; on

FIG. 5. Convergence and overlap demonstrated with ICMS. Results of ICMS at currents up to 80mA in 12 electrodepenetrations (denoted A through N) down the anterior wall of a baboon’s central sulcus while recording 3 single motor units: onein extensor digitorum communis (EDC, which extends all 4 fingers, radial nerve innervation); another in the thenar muscles(Thenar, which act only on the thumb, median nerve innervation); and a 3rd in the 1st dorsal interosseous (FDI, which acts on theindex finger, ulnar nerve innervation). In each penetration, black dots indicate locations where ICMS was ineffective, whereasnumbers indicate threshold current (mA) for evoking discharge of the motor unit. Lateral is to the viewer’s left; medial to the right.With currents up to 20mA (red), multiple small zones from which each motor unit could be discharged were identified. Althoughthe zones for the different motor units were largely interdigitated, on close inspection these small zones also overlapped to somedegree. At higher currents up to 40 (orange) and then 80mA (yellow), the small zones for each motor unit expanded and coalescedinto large cortical territories, increasing their mutual overlap. Current spread could not account for this degree of convergence andoverlap in the cortical territories of these 3 motor units, which each were served by different peripheral nerves and each acted ondifferent digits (modified from Andersen et al. 1975).

2132 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

the fingers and the wrist; or even on the fingers, wrist, elbow,and shoulder.

H O R I Z O N T A L I N T E R C O N N E C T I O N S

The concept of a strict somatotopic organization implied thatdifferent sites within M1 acted on their output targets relatively

independently. By providing site-specific outputs to selectedelements, the somatotopic map in M1 was thought to act like apiano keyboard on which higher levels of the cortex could playout motor programs. This notion has been supported by ana-tomic studies demonstrating that the majority of intracorticalconnections within M1 are relatively local, spreading horizon-tally over a radius of only 1–2 mm. Lesions made by passinga microelectrode through monkey M1 cortex radially (normalto the pial surface) resulted in dense fiber degeneration spread-ing horizontally from the lesion over a radius of 200–300mm,and less densely over a radius of 2–3 mm (Gatter and Powell1978). Intracellular injections of HRP into cat pyramidal tractcells showed axon collaterals spreading horizontally in layersV and VI, densely over a radius of 0.5–0.8 mm, and lessdensely over 1.5–2.0 mm radius, with a few extending as far as2–3 mm from the soma (Landry et al. 1980). Neurobiotin-filledcat pyramidal neurons in layers II and III extend horizontalaxons collaterals within these layers for up to 1 mm (Keller andAsanuma 1993).

These anatomic findings are consistent with studies demon-strating that the strongest physiologic interactions are foundbetween M1 neurons separated by 1–2 mm. Low-amplitude (4mA) ICMS in cat M1 evokes monosynaptic postsynaptic po-tentials (PSPs) in neurons within only a 0.5-mm horizontalradius, and polysynaptic PSPs chiefly within 1-mm radius(Asanuma and Rosen 1973). Spike-triggered averaging of in-tracellular potentials in monkey M1 likewise has shown thatEPSPs and inhibitory postsynaptic potentials (IPSPs) are stron-gest and most common within 1–2 mm (direct rather thanhorizontal distance) of the triggering neuron (Matsumura et al.1996). The observations that two sequential 20-mA ICMSpulses delivered through the same electrode produce intracor-tical facilitation of evoked EMG, whereas sequential pulsesdelivered through electrodes separated horizontally by 1.5–2.0mm do not also support the notion that the strongest physio-logic interactions between M1 neurons occur within 1.0 mm,although the same ICMS pulses do affect the discharge of somepyramidal tract neurons 1.5–2.0 mm away (Baker et al. 1998).Similarly, synchronous discharges in the action potential trainsof two neurons (indicating shared or serial inputs) are foundmore commonly when the two neurons are close enough to berecorded from the same microelectrode, and the likelihood offinding synchronous discharge decreases with horizontal sep-aration until synchrony is rarely detected between neurons

FIG. 6. Divergence in the projection of single corticospinal neurons.A: ahorseradish peroxidase (HRP)–filled corticospinal axon has been reconstructedin the transverse plane of the ventral horn of the monkey spinal cord. The cordmidline and central canal are to theleft; the lateral column to theright. Thefilled corticospinal axon enters the spinal gray matter from the lateral columnand branches repeatedly, ultimately giving off terminal ramifications in theoutlined motoneuron pools of 4 different muscles (reproduced from Shinoda etal. 1981).B: averages of rectified electromyographic (EMG) activity are shownfrom 6 muscles [extensor digitorum secundi et tertii (ED23), extensor carpiulnaris (ECU), extensor digitorum quarti et quinti (ED45), extensor digitorumcommunis (EDC), extensor carpi radialis longus (ECRL), and extensor carpiradialis brevis (ECRB)] that act on the wrist and/or fingers in macaquemonkeys. Each average of rectified EMG was triggered from several thousandspikes discharged by the simultaneously recorded M1 neuron whose averagedaction potential is shown in thetop trace.The brief (;10 ms) peaks that appearin each of the 1st 4 EMG traces shortly after the neuron spike indicate thatmotoneurons innervating these 4 muscles received synaptic excitation at ashort and fixed latency following the discharge of action potentials by the M1neuron (modified from Fetz and Cheney 1980).

2133CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

separated by more than 2.0 mm (Grammont and Riehle 1999;Kwan et al. 1987; Riehle et al. 1997; Smith and Fetz 1989).Horizontal interconnections extending 1–2 mm may mediateinteractions between M1 neurons contributing to the control ofdifferent muscles acting about the same joint (Capaday et al.1998), or neighboring joints of the same extremity (Kwan et al.1987).

Although the strongest intracortical interconnections thusoccur within a 1-mm radius of a given pyramidal neuron,recent anatomical studies have shown that many M1 neuronsextend axon collaterals even further horizontally, interconnect-ing much larger regions of M1 (Fig. 7). HRP injections in theICMS-defined digit region of monkey M1 revealed that neu-rons near the injection site extended terminal arbors throughoutthe upper extremity region, including the territory wherethreshold ICMS had evoked movements of the shoulder, el-bow, or wrist (Huntley and Jones 1991). Conversely, neuronalsomata throughout the upper extremity representation werefilled retrogradely by the HRP injection in the digit region,indicating that neurons throughout the upper extremity territorysend axon collaterals into a single focus within the digit rep-resentation. When the injection was placed laterally, close tothe face representation, labeled terminals and somata still werefound in the ICMS-defined shoulder representation, 7–8 mmmedial to the injection site. Similarly, when Fast Blue (FB) wasinjected in the digit representation, and Diamidino Yellow(DY) was injected 7–8 mm away in the shoulder region ofmonkey M1, retrogradely labeled FB and DY somata werefound intermingled throughout the M1 cortex between the two

injections, including some double-labeled neurons (Tokunoand Tanji 1993).

Horizontally projecting axon collaterals interconnecting theentire upper extremity representation have been demonstratedas well in the cat and the rat, where the collaterals have beenshown to arise predominantly from pyramidal neurons in layersIII and V, and to have predominantly excitatory, glutamatergiceffects (Aroniadou and Keller 1993; Keller 1993; Weiss andKeller 1994). In monkey M1, inhibitory, GABAergic neuronshave predominantly vertically oriented projections (DeFelipeand Jones 1985), although the axons of GABAergic basketcells may project horizontally for 1–3 mm. Furthermore, theeffective range of intracortical inhibition may be much greater(Kujirai et al. 1993), in part because local inhibitory interneu-rons may receive excitatory inputs from long-range horizontalprojections within M1 (Jacobs and Donoghue 1991). Long-range horizontal interconnections within M1 thus provide asubstrate for information to be interchanged through a networkdistributed widely through the M1 upper extremity represen-tation, again limiting the degree to which a particular M1 sitecan be associated with control of a particular body part.

Besides long-range intrinsic connections within M1, afferentinputs to M1 also show considerable horizontal distribution.Given that any particular locus within M1 tends to receivesomatosensory input from the same body part moved by ICMSat the locus (Murphy et al. 1978; Rosen and Asanuma 1972),it is not surprising that somatosensory inputs to M1 also havea scattered and intermingled distribution (Lemon 1981; Wonget al. 1978). In large part, somatosensory inputs to M1 arrive

FIG. 7. Horizontal interconnections in the M1 upper extremity representation. After using ICMS to map the M1 upper extremityrepresentation in a macaque monkey’s left hemisphere, HRP was injected in the low-threshold digit representation at the siteindicated by the large, filled black circle with surrounding coarse-stippled penumbra. This injection resulted in widespread terminallabeling (fine stippling inA) and retrograde filling of neuronal somata (small black dots inB). ICMS was delivered at sites indicatedby the large black dots in bothA andB. Regions where stimulation at many contiguous sites elicited movement of the same bodypart are delimited with dashed lines, exceptional sites being indicated individually. Stars indicate points where stimulation up to40 mA failed to evoke observable movement. The solid line at theright indicates the central sulcus, and data from its anterior bankhave been represented as if projected to the hemisphere’s surface. Scale bars atbottomrepresent 1 mm. Anterior is to theleft;posterior,right; medial,up; lateral,down (modified from Huntley and Jones 1991).

2134 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

via short U-fibers from the primary somatosensory cortex,fibers that arborize over a considerable rostrocaudal distance inM1. In macaques, these corticocortical axons may give off twoto three terminal arborizations separated by up to 800mm(DeFelipe et al. 1986). Similar arborization patterns have beenfound for corticocortical afferents to M1 from area 5 in the cat(Kakei et al. 1996). Thalamocortical afferents to M1 from thecat ventroanterior and ventrolateral (VA/VL) nuclear complexdistribute their terminal fields even more extensively, somecovering areas up to 5.03 4.8 mm (almost 25 mm2) inrostrocaudal and mediolateral dimensions (Shinoda et al.1993). Both corticocortical and thalamocortical afferents thusdistribute their information widely in the M1 forelimb repre-sentation.

The functional role played by long horizontal connectionswithin M1 remains uncertain. Nevertheless, physiologic stud-ies in awake behaving animals have demonstrated a number oftypes of correlations between the discharge of M1 neurons thatmay in part be mediated by these long horizontal interconnec-tions. One form of correlation between relatively distant M1neurons has been demonstrated by averaging the intracellular(IC) potential of one neuron triggered from the extracellular(EC) action potentials of a second neuron (Matsumura et al.1996). These averages frequently reveal a broad depolarizationof the IC neuron straddling the triggering spikes of the ECneuron. Such average synchronous excitation potentials(ASEPs) indicate that the IC and EC neurons both receivesome sort of synchronous excitation. ASEPs, although mostcommon and most intense in pairs of neurons separated by,2mm, also have been found in pairs of neurons separated by upto 4.5 mm in monkey M1.

A second type of long-range correlation has been demon-strated by examining the trial-by-trial variation in the dischargeof monkey M1 neurons averaged over 600 ms during a reach-ing task (Maynard et al. 1999). The trial-by-trial variation inaverage discharge rate of two neurons was more likely to becorrelated if the two neurons had similar preferred directions,suggesting that functionally similar neurons receive sharedinputs that fluctuated from trial to trial. The strength of suchcorrelations did not depend on the horizontal distance betweenthe two neurons, however. Although interneuronal separationsof only up to 2 mm were examined, that the correlationstrength was independent of separation distance suggests thatthis type of correlation extends beyond 2 mm.

The most extensive evidence of long-range interactionswithin M1, however, comes from studies of local field potential(LFP) oscillations, which occur synchronously over regions ofthe M1 upper extremity representation extending 14 mm me-diolaterally along the central sulcus of monkeys (Donoghue etal. 1998; Murthy and Fetz 1992, 1996a). These LFP oscilla-tions are coherent with simultaneous oscillations in EMG ac-tivity (Baker et al. 1997; Hari and Salenius 1999). Neurons atsites separated by up to 10 mm have been found to haveoscillatory modulation of their discharge in phase with theseLFP oscillations, and pairs of such neurons often show peaks incross-correlograms of their spike discharges recorded duringLFP oscillations (Baker et al. 1999; Murthy and Fetz 1996b).The fact that such correlations during LFP oscillations can befound between neurons in the left and right M1 indicates thatintrinsic horizontal connections are unlikely to be the soleanatomic basis for such widespread synchronization. Never-

theless, horizontal connections intrinsic to the M1 upper ex-tremity representation may contribute to synchronous LFPoscillations, associating the widespread neurons needed to per-form a coordinated movement of the entire extremity.

D I S T R I B U T E D A C T I V A T I O N

For many years, authorities debated whether it was musclesper se, or the movements they produced, that were representedsomatotopically in M1 (Phillips 1975). The convergence of M1outputs to single motoneuron pools from wide and overlappingcortical territories, and the divergence of output from singleM1 neurons to multiple motoneuron pools, both necessarilyconstrain any somatotopic representation of individual musclesin M1. The overlapping cortical territories of different musclesraise the possibility, however, that different combinations ofactivity in multiple muscles are represented at different corticalsites. Voluntary movements, even movements of a single jointor a single finger, typically involve simultaneous contractionsof multiple muscles (Beevor 1903; Schieber 1995). The simul-taneous contractions of a such a set of muscles producingmovement of one body part might be represented at one loca-tion in M1, while the simultaneous contractions of a partiallyoverlapping set of muscles producing movement of a differentbody part might be represented at a different location. Al-though electrical stimulation mapping had also suggested over-lap of movement representations (above), electrical stimulationis unlikely to mimic accurately the natural cortical activationthat occurs during voluntary movements.

Since the 1960s, a number of techniques (including singleneuron recording, functional neuroimaging, and magnetoen-cephalography) have become available for probing corticalactivity during voluntary movements performed by awake sub-jects. A well-ordered, discrete somatotopic organization of M1would imply that movements of different body parts involveactivation of spatially distinct regions of M1, with these re-gions arrayed in somatotopic order. Somatotopically orderedactivation during voluntary movements should be demonstra-ble with these modern techniques.

Experimental studies examining M1 activity during move-ments of different parts of the upper extremity, however, haverevealed relatively little evidence of activation in spatiallydistinct regions of M1. In monkeys performing individuatedmovements of each finger and of the wrist, single neurons werefound to discharge in relation to movements of several differ-ent fingers, which obviously constrains the degree to whichmovements of different fingers could be represented in spa-tially distinct regions of M1 (Schieber and Hibbard 1993).Moreover, the M1 territories containing neurons active duringmovements of different fingers were virtually coextensive,with little evidence of a somatotopic shift in the center ofactivity from lateral to medial for movements of the thumbthrough little finger and wrist (Fig. 8). Similarly, functionalmagnetic resonance imaging (fMRI) in humans has shownextensive overlap of the cortical territories activated duringperformance of thumb, index finger, ring finger or wrist move-ments (Sanes et al. 1995). Magnetoencephalography in humanslikewise has shown that the dipole sources of the neuromag-netic fields generated during movements of different digits arenot arrayed in somatotopic order, either in a single subject oraveraged across multiple subjects (Cheyne et al. 1991; Salenius

2135CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

et al. 1997). These studies all indicate that movements ofdifferent fingers are not mediated by activity in different,spatially segregated regions, somatotopically arrayed in M1.Rather, they suggest that movement of any finger is mediatedby the activity of neurons widely distributed in the M1 upperextremity representation.

Strict somatotopic organization of M1 also would predict

that the territory active during movements of multiple fingersshould be larger than the territory activated during movementof a single digit. When this hypothesis has been tested exper-imentally, however, the extent and amplitude of activation inthe primary sensorimotor cortex has been found to be signifi-cantly larger during movement of a single finger than duringsimultaneous movement of multiple fingers (Kitamura et al.1993; Remy et al. 1994). Such results indicate that the processof moving multiple fingers is not simply the sum of activatingmultiple separate M1 territories, each controlling a differentfinger; rather, moving a single finger without the others re-quires more M1 activity than moving multiple fingers simul-taneously. Presumably, such extra activation occurs because,besides controlling the motion of the one finger, M1 activelyparticipates in stabilizing other parts of the upper extremityduring the individuated movement of a particular finger (Hum-phrey and Reed 1983; Schieber 1990).

Could the distributed activation observed in awake behavingsubjects reflect activation of somatotopically organized cortex,with one region producing the movement and other regionsstabilizing other body parts? If one considers only studies ofactivation in awake behaving subjects, this interpretation cer-tainly is possible. But ICMS should have demonstrated such anunderlying somatotopic organization (see Figs. 4 and 5), andthe divergence of output from single M1 neurons to musclesthat move all four fingers and the wrist (Fig. 6), or to musclesacting on both the digits and the shoulder (McKiernan et al.1998), would certainly limit the degree of somatotopic segre-gation of these representations. Distributed representation pro-vides a more parsimonious interpretation of all these observa-tions considered together.

Several studies have demonstrated comparatively small, so-matotopically ordered shifts in the location of activation duringmovements of different parts of the upper extremity. It must berecognized, however, that these shifts are detected by usinganalytic approaches that minimize the contribution of activa-tion common to different movements. For example, somato-topically ordered shifts may be detected in the centroids ofactivation calculated for movements of different fingers. Theseshifts are small, however, compared with the total spatialextent of the territory activated. In monkeys, the centroids ofactivation during different finger and wrist movements werefound to be spread over 2 mm along the central sulcus, whereasthe field containing active neurons extended 8–9 mm (Fig. 8)(Schieber and Hibbard 1993). In humans, centroids of fMRIactivation for movements of different fingers may be spreadover 2.46 mm (no greater than the thickness of the cortexitself!) (Beistener et al. 2001; Hlustik et al. 2001; Indovina andSanes 2001), whereas the total extent of the hand representa-tion along the central sulcus is roughly 50 mm (Hlustik et al.2001; Penfield and Boldrey 1937). In both species, comparingthe spatial separation of the centroids with the spatial extent ofthe hand representation indicates that the territories activatedduring movements of different fingers must overlap exten-sively.

Analytic techniques that make even less use of the activationcommon to different movements have demonstrated greaterapparent separation. Although the territory of fMRI activationduring thumb movement overlaps extensively with that duringlittle finger movement, difference images that subtract awaythe shared activation have shown that the activation peak

FIG. 8. Distributed activation in M1 during finger movements.A: coloredspheres each represent a single neuron recorded in the left hemisphere M1 asa monkey performed individuated movements (flexion or extension) of eachright-hand digit and of the right wrist. Each neuron was consistently related toat least 1 movement, although most neurons were related to multiple differentfinger and/or wrist movements. The sphere representing each neuron is cen-tered at the location of the recorded neuron in the anterior bank of the centralsulcus, with the hemispheric surfaceabove, white matterbelow, lateral to theviewer’s right and medial to theleft. Each sphere is sized according to itsgreatest change in discharge frequency during any of the movements; the whitespheres atleft constitute a scale from 0 to 200 spikes/s, with centers 1 mmapart. Each sphere representing a neuron is colored according to the movementfor which that neuron’s greatest discharge occurred: thumb, red; index finger,orange; middle, yellow; ring, green; little, blue; wrist, violet. Neurons best-related to movements of each digit or the wrist were intermingled throughoutthe same cortical territory.B: centroids of discharge frequency changes cal-culated for each flexion movement and each extension movement are shown inthe same coordinate system as inA, with the scale of white spheres as a visualanchor. Rather than shifting progressively across the field of active cortex forthumb through little finger and wrist movements, these centroids all areclustered together in the center of the field, with only a slight shift formovements of different digits or the wrist, reflecting the extensive overlap ofthe representations of different movements (modified from Schieber andHibbard 1993).

2136 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

during thumb movements is lateral to that during little fingermovements (Kleinschmidt et al. 1997). Similarly, when theactivation peaks during thumb, index finger, wrist, elbow, andshoulder movements were compared using positron emissiontomography (PET), a somatotopic progression from lateral tomedial was demonstrated (Grafton et al. 1993). These obser-vations become consistent with observations of distributedactivation, as well as with observations of convergence, diver-gence, and horizontal interconnections, when the gradual shiftin peak activation or centroid of activation is recognized to bepresent on a base of extensively overlapping representation.

P A R T I A L I N A C T I V A T I O N

A strictly somatotopic organization of M1 also would pre-dict that in some instances lesions should affect certain parts ofthe upper extremity without affecting others. In human pa-tients, where lesions of M1 may be produced by a variety ofdisease processes, the resulting upper extremity weakness af-fects distal strength in the fingers and wrist more profoundlythan proximal strength in the shoulder and elbow (Bucy 1949;Colebatch and Gandevia 1989). Uncommonly, more selectivelesions weaken the fingers and wrist much more profoundlythan the elbow and shoulder (Foerster 1936). Even in thesecases, however, some weakness is evident at these proximaljoints, unless the weakness of the hand itself is minimal. Evenmore uncommonly, human patients have been reported withweakness greater in some fingers than in others. Most often, thethumb is the weakest digit, with some weakness of the index aswell (Lee et al. 1998; Terao et al. 1993). Greatest weakness ofthe thumb (and index) could result from a greater representa-tion of the thumb and index throughout the M1 hand area,rather than selective involvement of a region controlling onlythe thumb. Other cases have been reported, however, in whichthe little and ring fingers were weakest (Foerster 1936; Kim2001; Phan et al. 2000; Schieber 1999). Notably absent arecases in which the index, middle, or ring fingers were theweakest. Human cases thus suggest that, rather than discreteregions of M1 controlling different parts of the upper extrem-ity, control of each part is mediated by an extensive territorythat overlaps with the territories controlling other parts. Nev-ertheless, on top of this widely distributed control of eachfinger, two somatotopic gradients may be present, consistentwith the general order suggested by the homunculus. First, theproximal upper extremity is represented more heavily mediallythan laterally, while the reverse is true for the distal upperextremity. Second, within the distal representation, the thumband index are represented more heavily laterally than medially,while the little and ring fingers are represented more heavilymedially than laterally.

The exact location and extent of lesions in human cases, ofcourse, cannot be controlled experimentally. Relatively fewinvestigations in experimental animals have attempted to cor-relate the location of a lesions within the M1 upper extremityrepresentation with the resulting motor deficits in the upperextremity. In monkeys performing individuated finger move-ments, however, partial inactivation of the M1 hand area pro-duced by intracortical injection of the GABAA agonist, mus-cimol, impaired some finger movements more than others, butwhich finger movements were impaired was unrelated to themediolateral location of the inactivation along the central sul-

cus (Schieber and Poliakov 1998). Similarly, when muscimolwas injected at loci where ICMS evoked thumb and indexfinger movements, movements of the whole hand were im-paired (Brochier et al. 1999). Microinfarction of ICMS definedhand representation in squirrel monkeys resulted not only indecreased use of the hand, but also in tonic flexion at the elbowand adduction of the extremity close to the torso, similar to theinvoluntary tonic posturing of the upper extremity seen inhuman patients after much more extensive lesions of M1 (Frieland Nudo 1998). The deficits produced by controlled lesions inanimal studies, like those resulting from lesions produced bydisease in humans, suggest that control of each finger, and ofeach more proximal joint, is widely distributed in the M1 upperextremity representation.

P L A S T I C I T Y

Observations indicative of what we now call plasticity arealmost as old as stimulation mapping of M1. In their classicstudy of somatotopic organization of M1 in the great apes, forexample, Leyton and Sherrington (1917) took pains to describethe “functional instability of cortical motor points.” Theyfound that after the movement evoked by stimulation of a givenpoint was first identified, intervening stimulation of the samepoint or other nearby points could result in facilitation, rever-sal, or deviation of the movement evoked when stimulation ofthe given point subsequently was repeated. These investigatorsinferred that “. . . the functional instability of cortical motorpoints are indicative of the enormous wealth of mutual asso-ciations existing between the separable motor cortical points,and those associations must be a characteristic part of themachinery by which the synthetic powers of that cortex aremade possible.” When M1 was considered to contain a point-to-point somatotopic map of body parts, movements, or mus-cles, with each corticospinal neuron monosynaptically con-nected to one and only one spinal motoneuron pool, thepossibility of plastic reorganization in M1 seemed remote. Theconvergence, divergence, horizontal interconnections, and dis-tributed activation described above, however, provide a sub-strate that would appear capable of considerable plastic reor-ganization.

As reviewed in detail elsewhere (Nudo et al. 2001; Sanesand Donoghue 2000), in the past decade M1 has been shown toundergo plastic reorganization in response to a variety ofchanges, including peripheral lesions, central lesions, and mo-tor skill acquisition. Here we focus only on the latter, whichmay be most relevant to the normal organization of somato-topic representation in M1. In a variety of motor skill acqui-sition paradigms, the M1 representation of the trained bodyparts has been shown to enlarge, typically at the expense of therepresentations of less trained body parts. In rats trained toreach and grab small food pellets, for example, the ICMSdefined M1 representation of digit and wrist movements wasexpanded at the expense of elbow and shoulder movementrepresentation (Kleim et al. 1998). In monkeys trained toretrieve small objects, ICMS mapping likewise revealed ex-pansion of the digit representation, whereas in monkeys trainedto perform forearm pronation/supination movements the rep-resentation of forearm movements expanded (Nudo et al.1996). These changes are progressive as training continues,and reverse after training stops (Fig. 9).

2137CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

Additional evidence obtained in human subjects indicatesthat M1 reorganization occurs both within a single session andover the longer term needed to acquire a complex skill. Whennormal human subjects, initially unaware of any sequence,practice a repeating sequence of finger movements instructedby visual cues, the amplitude and extent of finger musclerepresentation assessed bytrans-cranial magnetic stimulationincreases in M1 contralateral to the performing hand as thespeed of performance increases over a single day of training(Pascual-Leone et al. 1994). Several days of such trainingproduce progressive expansion of finger muscle representation,whether the training involves physical practice or only mentalrehearsal (Pascual-Leone et al. 1995). Practicing a fingermovement sequence over several weeks results in greater fMRIactivation of the M1 hand representation during performanceof the practiced sequence than during performance of a com-parable, but unpracticed sequence (Karni et al. 1995). Anexample of very long-term changes related to motor skill isfound in experienced Braille readers, whose M1 representationof the first dorsal interosseous muscle (used to sweep the tip ofthe index finger over Braille letters) is expanded in M1 con-tralateral to their reading finger (Pascual-Leone et al. 1993).

If reorganization of M1 in normal subjects can be driven by

learning and practicing a particular skill, then reorganization islikely to be proceeding continuously as each individual per-forms the motor tasks used frequently in their daily life. Thepatterns of representation in M1 thus are likely to change as anindividual performs more of one motor activity and less ofanother from day to day. Such a continual process of reorga-nization places yet another constraint on somatotopic organi-zation in M1.

W H Y S H O U L D T H E P R I M A R Y M O T O R C O R T E X

H A V E S O D I S T R I B U T E D A N O R G A N I Z A T I O N ?

The ease with which we can comprehend a well-ordered,discrete, somatotopic representation makes the concept of so-matotopy an attractive organizing principle with which tounderstand the function of the primary motor cortex. Somato-topy seems so straightforward that it ought be so. The primarysomatosensory cortex (S1) has a well-ordered somatotopicrepresentation, and the primary visual cortex (V1) has a well-ordered retinotopic representation. The evidence reviewedabove indicating that within-limb somatotopy in M1 is limited,and that a more complex, widely distributed organization existsinstead, therefore is likely to reflect important features of

FIG. 9. M1 Reorganization in relation to motor training.A: after initial ICMS mapping of the distal forelimb representation(Baseline), a squirrel monkey was trained for 11 days on a Kluver board and then re-mapped (Training I), with little change in thetotal area devoted to digit (red), wrist/forearm (green), digit and wrist/forearm (yellow, DIG1 W/FA), digit and more proximal(blue, DIG1 PROX), or wrist/forearm and more proximal (purple, W/FA1 PROX) movements. After additional training on theKluver board for 39 days, re-mapping showed increases in the percent of the total distal forelimb area devoted to movements usedin retrieving food morsels from the small wells on the board: digit and wrist/forearm, finger flexion and wrist extension, and fingerflexion/extension movements (Training II). Then, after 4 mo without practice, re-mapping showed a reversion of these changes backto baseline (Extinction), and finally after another 30 days training, re-mapping showed that similar changes had occurred again(Reaquisition). Only points from which stimulation evoked movements of the digits, wrist, or pronation/supination movements ofthe forearm are shown in these maps; points eliciting only movements of the elbow and/or shoulder are not included, although theytypically surrounded the distal forelimb representations anteriorly (up), laterally (left) and medially (right) as indicated by “prox”in the baseline map.B: bargraphs show the percentage of the total distal forelimb representation in each map from which specificmovements were elicited in each phase of the study (modified from Nudo et al. 1996).

2138 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

functional organization in M1.4 I close with speculations as towhat these features might be.

One such feature may be the dimensionality of the informa-tion processed in M1. Well-ordered representations exist wherea two-dimensional receptor sheet (the skin surface or retina,respectively) can be mapped isomorphically onto the two-dimensional cortex. Movements and the muscles that generatethem are three-dimensional, however, and cannot be mappedsimply into a two-dimensional cortex. The number of dimen-sions represented in M1 is arguably much more than three, ifeach muscle, each degree of freedom at each joint, and eachkinematic or dynamic parameter of movement constitutes apossible dimension. Even in S1, the most discrete and well-ordered somatotopic representation of the different fingers isfound in area 3b, where cutaneous inputs predominate(Iwamura et al. 1983a; Pons et al. 1987). In areas 1 and 2,where cutaneous inputs are combined with inputs from deepmechanoreceptors in joints and muscles, increasing numbers ofreceptive fields span multiple digits, and somatotopic organi-zation becomes more complex, particularly in awake animals(Iwamura et al. 1980, 1983b, 1993; Pons et al. 1985). Incontrast, V1 represents additional dimensions by nesting themwithin the two-dimensional retinotopic representation. Oculardominance columns, orientation columns, and color blobs canbe considered additional dimensions of visual stimuli, therepresentations of which are nested within the two-dimensionalrepresentation of each retinotopic location. Little evidence ofsuch a fine-grained nesting has been found in M1, however,which presumably reflects some additional difference in corti-cal processing for control of movement versus perception ofsensory stimuli.

A second feature of functional organization may have to dowith what needs to be processed simultaneously by M1. Thewell-ordered representations in S1 (area 3b in particular) andV1 are thought to be computationally advantageous becausetwo adjacent receptors are much more likely to receive similarinput simultaneously than two distant receptors. If a mechano-receptor on the thumb is responding to an indenting stimulus,for example, another mechanoreceptor on the thumb is muchmore likely to be responding simultaneously than a mech-anoreceptor on the little finger. Some economy of neuralprocessing presumably is achieved by representing thumbmechanoreceptors close to one another, with little fingermechanoreceptors represented at a distance. In the muchless likely event that the thumb and little finger are indentedsimultaneously, however, the requisite neural processing ismore costly than if the thumb and little finger mechanore-ceptor representations were intermingled with one another.

Control of movement, particularly the control provided byM1, is fundamentally different. Innumerable combinations ofmuscle contractions and movements with relatively similarlikelihood must be represented. In this way M1 provides thecapacity to generate a huge repertoire of movements, as well asthe potential to generate previously unperformed movements.To achieve these abilities, the organizational substrate of M1must be able to access virtually any different combination ofmuscle contractions and body part movements with equalfacility. A well-ordered, discrete, somatotopic representationwould limit its ability to do so. Such a well-ordered somato-topic representation in M1 often has been likened to a pianokeyboard, on which other cortical areas play out movements, asillustrated in Fig. 10A, where colors have been added to thewhite keys to identify individual notes. Although many differ-ent tunes can be played on such a keyboard, a 21st centurycomposer might be disappointed that certain combinations ofnotes simply cannot be played. For example, a single pianistcannot play the five notes indicated by black dots in Fig. 10A.If, however, a modern two-dimensional keyboard is created in

4 Indeed, the resolution of somatotopic organization in area 3b exceeds thatwhich would be expected based on the divergence of thalamocortical afferentscarrying somatosensory input from a given finger, and the overlap of thalamo-cortical afferents carrying input from different fingers (Garraghty et al. 1989;Rausell et al. 1998). The precise somatotopy in area 3b therefore indicates thatactive mechanisms normally increase the somatotopic resolution in S1, incontrast to the mechanisms organizing M1.

FIG. 10. Cortical pianos.A: the standardpiano keyboard constitutes a well-orderedmap of musical notes. Here the 42 whitekeys have been colored to distinguish each ofthem. Although the strict order and singlerepresentation make it easy to locate eachnote, these same features make it impossiblefor a pianist to play certain combinations ofnotes, such as the 5 notes marked by blackdots beneath.B: a nonstandard keyboard canbe created by re-representing each note atmultiple locations and in a wide variety oforders. While this arrangement appears morecomplex, with each note represented over awide territory that overlaps with the territo-ries representing many other notes, this dis-tributed organization provides ready accessto many more combinations of notes than thestandard keyboard. The 5 adjacent notesmarked by black dots inB, for example, arethe same notes indicated inA.

2139CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

which each note (white keys only for simplicity) is re-repre-sented at multiple locations (Fig. 10B), then at some locationthe desired combination of five notes can be accessed as easilyas any other combination. Distributed organization thus canprovide much more flexible access to a wide variety of com-binations.

In theory, all possible combinations could be equally repre-sented. In practice, equivalent representation of all possiblecombinations might come at the cost of an excessively largecortical area. More area is required to re-represent any elementmultiple times (compare the size of Fig. 10,A vs. B), and theright connections must be established and maintained through-out a relatively large area, with relatively long conductiondelays. A compromise therefore might exist in which morefrequently used combinations are represented at more locationsthan less frequently used combinations. Hence when activatedby electrical stimulation, the extent of cortical territory repre-senting these frequently used combinations, and the corre-sponding body parts, would appear magnified relative to otherless frequently used combinations and body parts. As theindividual learns new motor skills (or quits practicing old ones)such a distributed system (with the underlying structural sub-strates of convergence, divergence, and horizontal interconnec-tion) could be reorganized to represent new combinations (atthe expense of now less used combinations) more readily thana well-ordered somatotopic system.

Distributed organization also provides greater resistance tothe disruptive effects of lesions. A tiny lesion the size of asingle piano key, for example, could eliminate any ability toproduce a given note (such as the pale yellow E) in thewell-ordered keyboard of Fig. 10A. The same tiny lesion wouldgo virtually unnoticed in the distributed keyboard of Fig. 10B.A considerably larger lesion covering many keys, would beneeded to noticeably compromise use of any given element onthe distributed keyboard. When such a lesion occurs, however,use of many notes all along the musical scale would be com-promised, consistent with the observations reviewed above thatinactivation in the M1 hand representation sufficient to producedetectable deficits affects movements of multiple fingers, notjust one.

In a distributed system, networked by convergence, diver-gence, and horizontal interconnections, somatotopic organiza-tion is theoretically unnecessary. Even if one subset of loca-tions representing a particular element (such as the pale yellowE, or thumb flexion) is active during one movement, andanother subset is active during another movement, no a priorirequirement forces the two different subsets to be spatiallysegregated. Movement control from the primary motor cortexis not distributed to the point of homogeneity, however. Asnoted above, the face, arm and leg representations are distinctfrom one another. Within the upper extremity representation,gradual somatotopic gradients also can be identified on top ofan underlying distributed representation. At some point, thecosts of distributed representation outweigh the benefits, whichmay have to do with a third feature of functional organizationin M1, the biomechanical interactions of the body parts beingcontrolled.

The degree of somatotopic segregation in M1 generallyparallels the biomechanical independence of different bodyparts. Thumb movements are biomechanically independent oflip movements, and the representations of the thumb and lips

therefore can be quite segregated in M1. Movements of thethumb and the wrist are not so independent. Extrinsic musclesacting on the thumb (flexor pollicis longus, extensor pollicislongus, and abductor pollicis longus) act across the wrist aswell, and because the proximal segment of the thumb is con-nected to the wrist directly, motion of the thumb will exertinteraction torques at the wrist. Precise control of thumb move-ment therefore will always require some control of the wrist,even when the wrist is being stabilized so as not to move whenthe thumb does. Because movement of the thumb alwaysrequires some degree of simultaneous control of the wrist, then,representation of the thumb and the wrist is intermingled to aconsiderable degree in M1. Even more intermingled in M1 arerepresentations of thumb and fingers. Movements of the dif-ferent digits are not entirely independent (Hager-Ross andSchieber 2000; Schieber 1991). In functional uses of the hand,even when performing sophisticated tasks such as typing orplaying the piano, the thumb and fingers are in motion simul-taneously (Engel et al. 1997; Fish and Soechting 1992; Santelloand Soechting 1998).

The need to control a wide variety of movements in biome-chanically coupled, simultaneously moving body parts mayhave constrained evolution of a well-ordered, spatially segre-gated, discrete somatotopic map in M1. Indeed, when Hugh-lings Jackson initially proposed localization of control ofmovements in the brain, he recognized that, “. . . since themovements of the thumb and fingers could scarcely be devel-oped for any useful purpose without fixation of the wrist . . . ,we should a priori be sure that the center discharged, although itmight represent movements in which the thumb had the leadingpart, must represent also certain other movements of the forearm,upper arm, etc., which serve subordinately” (Jackson 1958, p. 69).

The author thanks J. Gardinier for preparing illustrations, M. Hayles foreditorial comments, and the anonymous reviewers for helpful critiques.

This work was supported by National Institutes of Health Grants R01-NS-27686 and P41-RR-09283.

REFERENCES

ANDERSEN P, HAGAN PJ, PHILLIPS CG, AND POWELL TP. Mapping by micro-stimulation of overlapping projections from area 4 to motor units of thebaboon’s hand.Proc R Soc Lond B Biol Sci188: 31–36, 1975.

ARONIADOU VA AND KELLER A. The patterns and synaptic properties ofhorizontal intracortical connections in the rat motor cortex.J Neurophysiol70: 1553–1569, 1993.

ASANUMA H, ARNOLD A, AND ZARECKI P. Further study on the excitation ofpyramidal tract cells by intracortical microstimulation.Exp Brain Res26:443–461, 1976.

ASANUMA H AND ROSEN I. Topographical organization of cortical efferentzones projecting to distal forelimb muscles in the monkey.Exp Brain Res14: 243–256, 1972.

ASANUMA H AND ROSEN I. Spread of mono- and polysynaptic connectionswithin cat’s motor cortex.Exp Brain Res16: 507–520, 1973.

BAKER SN, KILNER JM, PINCHES EM, AND LEMON RN. The role of synchronyand oscillations in the motor output.Exp Brain Res128: 109–117, 1999.

BAKER SN, OLIVIER E, AND LEMON RN. Coherent oscillations in monkey motorcortex and hand muscle EMG show task-dependent modulation.J Physiol(Lond) 501: 225–241, 1997.

BAKER SN, OLIVIER E, AND LEMON RN. An investigation of the intrinsiccircuitry of the motor cortex of the monkey using intra-cortical microstimu-lation. Exp Brain Res123: 397–411, 1998.

BEEVOR CE. (1) The Croonian Lectures on Muscular Movements and(2)Remarks on Paralysis of the Movements of the Trunk in Hemiplegia.NewYork: The MacMillan Company, 1903, p. 1–79.

BEEVOR CE AND HORSLEY V. A minute analysis (experimental) of the variousmovements produced by stimulating in the monkey different regions of the

2140 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

cortical centre for the upper limb, as defined by Professor Ferrier.PhilosTrans R Soc Lond B Biol Sci178: 153–167, 1887.

BEISTENER RCW, LANZENBERGER R, EDWARD V, CUNNINGTON R, ERDLER M,GARTUS A, STREIBL B, MOSER E, AND DEECKE L. Finger somatotopy inhuman motor cortex.NeuroImage13: 1016–1026, 2001.

BROCHIERT, BOUDREAU MJ, PARE M, AND SMITH AM. The effects of muscimolinactivation of small regions of motor and somatosensory cortex on inde-pendent finger movements and force control in the precision grip.Exp BrainRes128: 31–40, 1999.

BUCY PC. Effects of extirpation in man. In:The Precentral Motor Cortex,edited by Bucy PC. Urbana, IL: Univ. of Illinois Press, 1949, p. 353–394.

BUYS EJ, LEMON RN, MANTEL GW, AND MUIR RB. Selective facilitation ofdifferent hand muscles by single corticospinal neurones in the consciousmonkey.J Physiol (Lond)381: 529–549, 1986.

CAPADAY C, DEVANNE H, BERTRAND L, AND LAVOIE BA. Intracortical connec-tions between motor cortical zones controlling antagonistic muscles in thecat: a combined anatomical and physiological study.Exp Brain Res120:223–232, 1998.

CHANG H-T, RUCH TC, AND WARD AA JR. Topographical representation ofmuscles in motor cortex of monkeys.J Neurophysiol11: 39–56, 1947.

CHENEY PD AND FETZ EE. Comparable patterns of muscle facilitation evokedby individual corticomotoneuronal (CM) cells and by single intracorticalmicrostimuli in primates: evidence for functional groups of CM cells.J Neurophysiol53: 786–804, 1985.

CHEYNE D, KRISTEVA R, AND DEECKE L. Homuncular organization of humanmotor cortex as indicated by neuromagnetic recordings.Neurosci Lett122:17–20, 1991.

COLEBATCH JG AND GANDEVIA SC. The distribution of muscular weakness inupper motor neuron lesions affecting the arm.Brain 112: 749–763, 1989.

DEFELIPE J, CONLEY M, AND JONES EG. Long-range focal collateralization ofaxons arising from corticocortical cells in monkey sensory-motor cortex.J Neurosci6: 3749–3766, 1986.

DEFELIPE J AND JONES EG. Vertical organization of gamma-aminobutyricacid-accumulating intrinsic neuronal systems in monkey cerebral cortex.J Neurosci5: 3246–3260, 1985.

DONOGHUE JP, LEIBOVIC S, AND SANES JN. Organization of the forelimb areain squirrel monkey motor cortex: representation of digit, wrist, and elbowmuscles.Exp Brain Res89: 1–19, 1992.

DONOGHUE JP, SANES JN, HATSOPOULOSNG, AND GAAL G. Neural dischargeand local field potential oscillations in primate motor cortex during volun-tary movements.J Neurophysiol79: 159–173, 1998.

ENGEL KC, FLANDERS M, AND SOECHTING JF. Anticipatory and sequentialmotor control in piano playing.Exp Brain Res113: 189–199, 1997.

FERRIER D. Experimental researches in cerebral physiology and pathology.West Riding Lunatic Asylum Medical Report3: 30–96, 1873.

FETZ EE AND CHENEY PD. Muscle fields of primate corticomotoneuronal cells.J Physiol Paris74: 239–245, 1978.

FETZ EE AND CHENEY PD. Postspike facilitation of forelimb muscle activity byprimate corticomotoneuronal cells.J Neurophysiol44: 751–772, 1980.

FISH J AND SOECHTING JF. Synergistic finger movements in a skilled motortask.Exp Brain Res91: 327–334, 1992.

FOERSTERO. Motorishce Felder und Bahnen. In:Handbuch der Neurologie,edited by Bumke O and Foerster O. Berlin: Springer, 1936, p. 1–357.

FRIEL KM AND NUDO RJ. Recovery of motor function after focal cortical injuryin primates: compensatory movement patterns used during rehabilitativetraining.Somatosens Mot Res15: 173–189, 1998.

FULTON JFAND KELLER AD. The Sign of Babinski: A Study of the Evolution ofCortical Dominance in Primates.Springfield, IL: Charles C. Thomas, 1932.

GARRAGHTY PE, PONS TP, SUR M, AND KAAS JH. The arbors of axonsterminating in middle cortical layers of somatosensory area 3b in owlmonkeys.Somatosens Mot Res6: 401–411, 1989.

GATTER KC AND POWELL TPS. The intrinsic connections of the cortex of area4 of the monkey.Brain 101: 513–541, 1978.

GOULD HJ, CUSICK CG, PONS TP, AND KAAS JH. The relationship of corpuscallosum connections to electrical stimulation maps of motor, supplemen-tary motor, and the frontal eye fields in owl monkeys.J Comp Neurol247:297–325, 1986.

GRAFTON ST, WOODS RP, AND MAZZIOTTA JC. Within-arm somatotopy inhuman motor areas determined by positron emission tomography imaging ofcerebral blood flow.Exp Brain Res95: 172–176, 1993.

GRAMMONT F AND RIEHLE A. Precise spike synchronization in monkey motorcortex involved in preparation for movement.Exp Brain Res128: 118–122,1999.

HAGER-ROSSCK AND SCHIEBER MH. Quantifying the independence of humanfinger movements: comparisons of digits, hands and movement frequencies.J Neurosci20: 8542–8550, 2000.

HARI R AND SALENIUS S. Rhythmical corticomotor communication.Neurore-port 10: R1–R10, 1999.

HERN JEC, LANDGREN S, PHILLIPS CG, AND PORTER R. Selective excitation ofcorticofugal neurones by surface-anodal stimulation of the baboon’s motorcortex.J Physiol (Lond)161: 73–90, 1962.

HINES M. Movements elicited from precentral gyrus of adult chimpanzees bystimulation with sine wave currents.J Neurophysiol3: 442–466, 1940.

HLUSTIK P, SOLODKIN A, GULLAPALLI RP, NOLL DC, AND SMALL SL. Soma-totopy in human primary motor cortex and somatosensory hand represen-tations revisited.Cereb Cortex11: 312–321, 2001.

HUANG CS, SIRISKO MA, HIRABA H, MURRAY GM, AND SESSLE BJ. Organi-zation of the primate face motor cortex as revealed by intracortical micro-stimulation and electrophysiological identification of afferent inputs andcorticobulbar projections.J Neurophysiol59: 796–818, 1988.

HUMPHREY DR. Representation of movements and muscles within the primateprecentral motor cortex: historical and current perspectives.Federation Proc45: 2687–2699, 1986.

HUMPHREY DR AND REED DJ. Separate cortical systems for control of jointmovement and joint stiffness: reciprocal activation and coactivation ofantagonist muscles.Adv Neurol39: 347–372, 1983.

HUNTLEY GW AND JONESEG. Relationship of intrinsic connections to forelimbmovement representations in monkey motor cortex: a correlative anatomicand physiological study.J Neurophysiol66: 390–413, 1991.

INDOVINA I AND SANES J. Journey to the somatotopic representations center forfinger movements in human primary motor cortex and supplementary motorareas.NeuroImage13: 1027–1034, 2001.

IWAMURA Y, TANAKA M, AND HIKOSAKA O. Overlapping representation offingers in the somatosensory cortex (area 2) of the conscious monkey.BrainRes197: 516–520, 1980.

IWAMURA Y, TANAKA M, SAKAMOTO M, AND HIKOSAKA O. Functional subdi-visions representing different finger regions in area 3 of the first somato-sensory cortex of the conscious monkey.Exp Brain Res51: 315–326, 1983a.

IWAMURA Y, TANAKA M, SAKAMOTO M, AND HIKOSAKA O. Converging pat-terns of finger representaiton and complex response properties of neurons inarea 1 of the first somatosensory cortex of the conscious monkey.Exp BrainRes51: 327–337, 1983b.

IWAMURA Y, TANAKA M, SAKAMOTO M, AND HIKOSAKA O. Rostrocaudalgradients in the neuronal receptive field complexity in the finger region ofthe alert monkey’s postcentral gyrus.Exp Brain Res92: 360–368, 1993.

JACKSON JH. On the anatomical and physiological localisation of movementsin the brain. In:Selected Writings of John Hughlings Jackson. On Epilepsyand Epileptiform Convulsions,edited by Taylor J, Holmes G, and WalsheFMR. New York: Basic Books, 1958, vol. 1, p. 37–76.

JACOBS KM AND DONOGHUEJP. Reshaping the cortical motor map by unmask-ing latent intracortical connections.Science251: 944–947, 1991.

JANKOWSKA E, PADEL Y, AND TANAKA R. Projections of pyramidal tract cellsto alpha-motoneurones innervating hind-limb muscles in the monkey.J Physiol (Lond)249: 637–667, 1975a.

JANKOWSKA E, PADEL Y, AND TANAKA R. The mode of activation of pyramidaltract cells by intracortical stimuli.J Physiol (Lond)249: 617–636, 1975b.

KAKEI S, FUTAMI T, AND SHINODA Y. Projection pattern of single corticocor-tical fibers from the parietal cortex to the motor cortex.Neuroreport7:2369–2372, 1996.

KARNI A, MEYER G, JEZZARD P, ADAMS MM, TURNER R, AND UNGERLEIDER

LG. Functional MRI evidence for adult motor cortex plasticity during motorskill learning.Nature377: 155–158, 1995.

KELLER A. Intrinsic connections between representation zones in the cat motorcortex.Neuroreport4: 515–518, 1993.

KELLER A AND ASANUMA H. Synaptic relationships involving local axoncollaterals of pyramidal neurons in the cat motor cortex.J Comp Neurol336:229–242, 1993.

KIM JS. Predominant involvement of a particular group of fingers due to smallcortical infarction.Neurology56: 1677–1682, 2001.

KITAMURA J, SHIBASAKI H, AND KONDO T. A cortical slow potential is largerbefore an isolated movement of a single finger than simultaneous movementof two fingers.Electroencephalogr Clin Neurophysiol86: 252–258, 1993.

KLEIM JA, BARBAY S,AND NUDO RJ. Functional reorganization of the rat motorcortex following motor skill learning.J Neurophysiol80: 3321–3325, 1998.

KLEINSCHMIDT A, NITSCHKE MF, AND FRAHM J. Somatotopy in the humanmotor cortex hand area. A high-resolution functional MRI study.EurJ Neurosci9: 2178–2186, 1997.

2141CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

KRINGST, NAUJOKAT C, AND VON KEYSERLINGK DG. Representation of corticalmotor function as revealed by stereotactic transcranial magnetic stimulation.Electroencephalogr Clin Neurophysiol109: 85–93, 1998.

KUJIRAI T, CARAMIA MD, ROTHWELL JC, DAY BL, THOMPSONPD, FERBERTA,WROE S, ASSELMAN P, AND MARSDEN CD. Corticocortical inhibition inhuman motor cortex.J Physiol (Lond)471: 501–519, 1993.

KWAN HC, MACKAY WA, MURPHY JT,AND WONG YC. Spatial organization ofprecentral cortex in awake primates. II. Motor outputs.J Neurophysiol41:1120–1131, 1978.

KWAN HC, MURPHY JT, AND WONG YC. Interaction between neurons inprecentral cortical zones controlling different joints.Brain Res400: 259–269, 1987.

LANDAU WM, BISHOPGH, AND CLARE MH. Site of excitation in stimulation ofthe motor cortex.J Neurophysiol28: 1206–1222, 1965.

LANDGREN S, PHILLIPS CG, AND PORTER R. Cortical fields of origin of themonosynaptic pyramidal pathways to some alpha motoneurones of thebaboon’s hand and forearm.J Physiol (Lond)161: 112–125, 1962.

LANDRY P, LABELLE A, AND DESCHENESM. Intracortical distribution of axonalcollaterals of pyramidal tract cells in the cat motor cortex.Brain Res191:327–336, 1980.

LEE PH, HAN SW, AND HEO JH. Isolated weakness of the fingers in corticalinfarction.Neurology50: 823–824, 1998.

LEMON RN. Variety of functional organization within the monkey motorcortex.J Physiol (Lond)311: 521–540, 1981.

LEMON RN, MANTEL GW, AND MUIR RB. Corticospinal facilitation of handmuscles during voluntary movement in the conscious monkey.J Physiol(Lond) 381: 497–527, 1986.

LEYTON ASF AND SHERRINGTON CS. Observations on the excitable cortex ofthe chimpanzee, orang-utan, and gorilla.Q J Exp Physiol11: 137–222,1917.

MATSUMURA M, CHEN DF, SAWAGUCHI T, KUBOTA K, AND FETZ EE. Synapticinteractions between primate precentral cortex neurons revealed by spike-triggered averaging of intracellular membrane potentials in vivo.J Neurosci16: 7757–7767, 1996.

MAYNARD EM, HATSOPOULOS NG, OJAKANGAS CL, ACUNA BD, SANES JN,NORMANN RA, AND DONOGHUE JP. Neuronal interactions improve corticalpopulation coding of movement direction.J Neurosci19: 8083–8093, 1999.

MCKIERNAN BJ, MARCARIO JK, KARRER JH, AND CHENEY PD. Corticomo-toneuronal postspike effects in shoulder, elbow, wrist, digit, and intrinsichand muscles during a reach and prehension task.J Neurophysiol80:1961–1980, 1998.

MURPHY JP AND GELLHORN E. Multiplicity of representation versus punctatelocalization in the motor cortex.Arch Neurol Psychiatry54: 256–273, 1945.

MURPHY JT, KWAN HC, MACKAY WA, AND WONG YC. Spatial organization ofprecentral cortex in awake primates. III. Input-output coupling.J Neuro-physiol41: 1132–1139, 1978.

MURTHY VN AND FETZ EE. Coherent 25- to 35-Hz oscillations in the senso-rimotor cortex of awake behaving monkeys.Proc Natl Acad Sci USA89:5670–5674, 1992.

MURTHY VN AND FETZ EE. Oscillatory activity in sensorimotor cortex ofawake monkeys: synchronization of local field potentials and relation tobehavior.J Neurophysiol76: 3949–3967, 1996a.

MURTHY VN AND FETZ EE. Synchronization of neurons during local fieldpotential oscillations in sensorimotor cortex of awake monkeys.J Neuro-physiol76: 3968–3982, 1996b.

NOWAK LG AND BULLIER J. Axons, but not cell bodies, are activated byelectrical stimulation in cortical gray matter. I. Evidence from chronaxiemeasurements.Exp Brain Res118: 477–488, 1998a.

NOWAK LG AND BULLIER J. Axons, but not cell bodies, are activated byelectrical stimulation in cortical gray matter. II. Evidence from selectiveinactivation of cell bodies and axon initial segments.Exp Brain Res118:489–500, 1998b.

NUDO RJ, MILLIKEN GW, JENKINS WM, AND MERZENICH MM. Use-dependentalterations of movement representations in primary motor cortex of adultsquirrel monkeys.J Neurosci16: 785–807, 1996.

NUDO RJ, PLAUTZ EJ, AND FROST SB. Role of adaptive plasticity in recoveryof function after damage to motor cortex.Muscle Nerve24: 1000–1019,2001.

PARK MC, BELHAJ-SAI F A, GORDON M, AND CHENEY PD. Consistent featuresin the forelimb representation of primary motor cortex in rhesus macaques.J Neurosci21: 2784–2792, 2001.

PASCUAL-LEONE A, CAMMAROTA A, WASSERMANN EM, BRASIL-NETO JP, CO-

HEN LG, AND HALLETT M. Modulation of motor cortical outputs to thereading hand of braille readers.Ann Neurol34: 33–37, 1993.

PASCUAL-LEONE A, DANG N, COHEN LG, BRASIL-NETO JP, CAMMAROTA A,AND HALLETT M. Modulation of muscle responses evoked by transcranialmagnetic stimulation during the acquisition of new fine motor skills.J Neu-rophysiol74: 1037–1045, 1995.

PASCUAL-LEONEA, GRAFMAN J,AND HALLETT M. Modulation of cortical motoroutput maps during development of implicit and explicit knowledge.Science263: 1287–1289, 1994.

PATTON HD AND AMASSIAN VE. Single- and multiple-unit analysis of corticalstage of pyramidal tract activation.J Neurophysiol17: 345–363, 1954.

PENFIELD W AND BOLDREY E. Somatic motor and sensory representation in thecerebral cortex of man as studied by electrical stimulation.Brain 37:389–443, 1937.

PENFIELD W AND RASMUSSEN T. The Cerebral Cortex of Man.New York:MacMillan, 1950.

PHAN TG, EVANS BA, AND HUSTON J. Pseudoulnar palsy from a small infarctof the precentral knob.Neurology54: 2185, 2000.

PHILLIPS CG. Cortical motor threshold and the thresholds and distribution ofexcited Betz cells in the cat.Q J Exp Physiol41: 70–83, 1956.

PHILLIPS CG. Laying the ghost of ‘muscles versus movements.’Can J NeurolSci 2: 209–218, 1975.

PHILLIPS CG AND LANDAU WM. Upper and lower motor neuron: the little oldsynecdoche that works.Neurology40: 884–886, 1990.

PONS TP, GARRAGHTY PE, CUSICK CG, AND KAAS JH. The somatotopicorganization of area 2 in macaque monkeys.J Comp Neurol241: 445–466,1985.

PONS TP, WALL JT, GARRAGHTY PE, CUSICK CG, AND KAAS JH. Consistentfeatures of the representation of the hand in area 3b of macaque monkeys.Somatosens Res4: 309–331, 1987.

RATHELOT J-A AND STRICK PL. Anatomical evidence that area 3A has mono-synaptic connections with hand motoneurons.Soc Neurosci Abstr26: 1483,2000.

RAUSELL E, BICKFORD L, MANGER PR, WOODSTM, AND JONESEG. Extensivedivergence and convergence in the thalamocortical projection to monkeysomatosensory cortex.J Neurosci18: 4216–4232, 1998.

REMY P, ZILBOVICIUS M, LEROY-WILLIG A, SYROTA A, AND SAMSON Y.Movement- and task-related activations of motor cortical areas: a positronemission tomographic study.Ann Neurol36: 19–26, 1994.

RIEHLE A, GRUN S, DIESMANN M, AND AERTSENA. Spike synchronization andrate modulation differentially involved in motor cortical function [see com-ments].Science278: 1950–1953, 1997.

ROSENI AND ASANUMA H. Peripheral afferent inputs to the forelimb area of themonkey motor cortex: input-output relations.Exp Brain Res14: 257–273,1972.

SALENIUS S, PORTIN K, KAJOLA M, SALMELIN R, AND HARI R. Cortical controlof human motoneuron firing during isometric contraction.J Neurophysiol77: 3401–3405, 1997.

SANES JN AND DONOGHUE JP. Plasticity and primary motor cortex.Annu RevNeurosci23: 393–415, 2000.

SANES JN, DONOGHUE JP, THANGARAJ V, EDELMAN RR, AND WARACH S.Shared neural substrates controlling hand movements in human motorcortex.Science268: 1775–1777, 1995.

SANTELLO M AND SOECHTING JF. Gradual molding of the hand to objectcontours.J Neurophysiol79: 1307–1320, 1998.

SATO KC AND TANJI J. Digit-muscle responses evoked from multiple intracor-tical foci in monkey precentral motor cortex.J Neurophysiol62: 959–970,1989.

SCHIEBER MH. How might the motor cortex individuate movements?TrendsNeurosci13: 440–445, 1990.

SCHIEBER MH. Individuated finger movements of rhesus monkeys: a means ofquantifying the independence of the digits.J Neurophysiol65: 1381–1391,1991.

SCHIEBER MH. Muscular production of individuated finger movements: theroles of extrinsic finger muscles.J Neurosci15: 284–297, 1995.

SCHIEBER MH. Somatotopic gradients in the distributed organization of thehuman primary motor cortex hand area: evidence from small infarcts.ExpBrain Res128: 139–148, 1999.

SCHIEBER MH AND HIBBARD LS. How somatotopic is the motor cortex handarea?Science261: 489–492, 1993.

SCHIEBER MH AND POLIAKOV AV. Partial inactivation of the primary motorcortex hand area: effects on individuated finger movements.J Neurosci18:9038–9054, 1998.

2142 M. H. SCHIEBER

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from

SCHOTT GD. Penfield’s homunculus: a note on cerebral cartography.J NeurolNeurosurg Psychiatry56: 329–333, 1993.

SESSLE BJ AND WIESENDANGER M. Structural and functional definition of themotor cortex in the monkey (Macaca fascicularis). J Physiol (Lond)323:245–265, 1982.

SHINODA Y, KAKEI S, FUTAMI T, AND WANNIER T. Thalamocortical organiza-tion in the cerebello-thalamo-cortical system.Cereb Cortex3: 421–429,1993.

SHINODA Y, YOKOTA J, AND FUTAMI T. Divergent projection of individualcorticospinal axons to motoneurons of multiple muscles in the monkey.Neurosci Lett23: 7–12, 1981.

SMITH WS AND FETZ EE. Effects of synchrony between primate corticomo-toneuronal cells on post-spike facilitation of muscles and motor units.Neurosci Lett96: 76–81, 1989.

STONEY SD JR, THOMPSONWD, AND ASANUMA H. Excitation of pyramidal tractcells by intracortical microstimulation: effective extent of stimulating cur-rent.J Neurophysiol31: 659–669, 1968.

STRICK PL AND PRESTON JB. Multiple representation in the primate motorcortex.Brain Res154: 366–370, 1978.

STRICK PL AND PRESTON JB. Two representations of the hand in area 4 of aprimate. I. Motor output organization.J Neurophysiol48: 139–149, 1982.

TERAO Y, HAYASHI H, KANDA T, AND TANABE H. Discrete cortical infarctionwith prominent impairment of thumb flexion.Stroke24: 2118–2120, 1993.

TOKUNO H AND TANJI J. Input organization of distal and proximal forelimbareas in the monkey primary motor cortex: a retrograde double labelingstudy.J Comp Neurol333: 199–209, 1993.

WALSHE FMR. Critical Studies in Neurology.London: E. and S. LivingstoneLtd., 1948.

WASSERMANN EM, MCSHANE LM, HALLETT M, AND COHEN LG. Noninvasivemapping of muscle representations in human motor cortex.Electroencepha-logr Clin Neurophysiol85: 1–8, 1992.

WATERS RS, SAMULACK DD, DYKES RW, AND MCKINLEY PA. Topographicorganization of baboon primary motor cortex: face, hand, forelimb, andshoulder representation.Somatosens Mot Res7: 485–514, 1990.

WEISS DS AND KELLER A. Specific patterns of intrinsic connections betweenrepresentation zones in the rat motor cortex.Cereb Cortex4: 205–214, 1994.

WONG YC, KWAN HC, MACKAY WA, AND MURPHY JT. Spatial organization ofprecentral cortex in awake primates. I. Somatosensory inputs.J Neuro-physiol41: 1107–1119, 1978.

WOOLSEYCN, ERICKSONTC, AND GILSON WE. Localization in somatic sensoryand motor areas of human cerebral cortex as determined by direct recordingof evoked potentials and electrical stimulation.J Neurosurg51: 476–506,1979.

WOOLSEY CN, SETTLAGE PH, MEYER DR, SENCERW, HAMUY TP, AND TRAVIS

AM. Patterns of localization in precentral and “supplementary” motor areasand their relation to the concept of a premotor area.Res Pub Assoc Res NervMent Dis30: 238–264, 1952.

2143CONSTRAINTS ON SOMATOTOPY IN M1

J Neurophysiol• VOL 86 • NOVEMBER 2001• www.jn.org

on March 26, 2009

jn.physiology.orgD

ownloaded from