Control of wrist position and muscle relaxation by shifting spatial frames of reference for motoneuronal recruitment: possible involvement of corticospinal pathways

J Physiol 588.9 (2010) pp 1551–1570 1551

Control of wrist position and muscle relaxation by shiftingspatial frames of reference for motoneuronal recruitment:possible involvement of corticospinal pathways

Helli Raptis1,2, Liziane Burtet1,2,3, Robert Forget1,2,3 and Anatol G. Feldman1,2

1Department of Physiology, Universite de Montreal and 2Center for Interdisciplinary Research in Rehabilitation (CRIR),Institut de readaptation Gingras-Lindsay de Montreal and Jewish Rehabilitation Hospital, Laval, PQ, Canada3School of Rehabilitation, University of Montreal, Montreal, PQ, Canada

It has previously been established that muscles become active in response to deviations froma threshold (referent) position of the body or its segments, and that intentional motor actionsresult from central shifts in the referent position. We tested the hypothesis that corticospinalpathways are involved in threshold position control during intentional changes in the wristposition in humans. Subjects moved the wrist from an initial extended to a final flexed position(and vice versa). Passive wrist muscle forces were compensated with a torque motor suchthat wrist muscle activity was equalized at the two positions. It appeared that motoneuronalexcitability tested by brief muscle stretches was also similar at these positions. Responses tomechanical perturbations before and after movement showed that the wrist threshold positionwas reset when voluntary changes in the joint angle were made. Although the excitability ofmotoneurons was similar at the two positions, the same transcranial magnetic stimulus (TMS)elicited a wrist extensor jerk in the extension position and a flexor jerk in the flexion position.Extensor motor-evoked potentials (MEPs) elicited by TMS at the wrist extension position weresubstantially bigger compared to those at the flexion position and vice versa for flexor MEPs.MEPs were substantially reduced when subjects fully relaxed wrist muscles and the wrist was heldpassively in each position. Results suggest that the corticospinal pathway, possibly with otherdescending pathways, participates in threshold position control, a process that pre-determinesthe spatial frame of reference in which the neuromuscular periphery is constrained to work.This control strategy would underlie not only intentional changes in the joint position, butalso muscle relaxation. The notion that the motor cortex may control motor actions by shiftingspatial frames of reference opens a new avenue in the analysis and understanding of brainfunction.

(Resubmitted 1 January 2010; accepted after revision 9 March 2010; first published online 15 March 2010)Corresponding author H. Raptis: University of Montreal, Rehabilitation Institute of Montreal (IRM), Research Center,6300 Darlington, 4th floor, Montreal, Qc, H3S 2J4, Canada. Email: [email protected]

Abbreviations DF, degree of freedom; E, extension position; ECR, extensor carpi radialis; ECU, extensor carpiulnaris; F, flexion position; FCR, flexor carpi radialis; FCU, flexor carpi ulnaris; FR, frame of reference; M1, primarymotor cortex; MEP, motor-evoked potential; MT, motor threshold; Ri and Rf , initial and final threshold joint angles;SR, stretch response; TMS, transcranial magnetic stimulation.

Introduction

Studies employing transcranial magnetic stimulation(TMS) in humans revealed that corticospinal facilitationof motoneurons is initiated prior to the onset ofand is modulated during the muscle activation under-lying intentional movements (Hoshiyama et al. 1997;MacKinnon & Rothwell, 2000; Schneider et al. 2004;

Irlbacher et al. 2006). Opinions on the role of primarymotor cortex (M1) in the specification of EMG patternsremain controversial, varying from the idea that M1directly encodes these patterns (Irlbacher et al. 2006;Townsend et al. 2006; Jackson et al. 2007) to that M1 onlyindirectly influences them (Lemon et al. 1995; Grazianoet al. 2002). Single-cell recordings also show that M1activity may correlate with different kinematic and kinetic

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society DOI: 10.1113/jphysiol.2009.186858

1552 H. Raptis and others J Physiol 588.9

variables, such as the direction, velocity and position ofthe hand in peripersonal space as well as with the final armposture (Caminiti et al. 1990; Kalaska et al. 1997; Aflalo& Graziano, 2006; Ganguly et al. 2009). M1 activity mayalso correlate with variables that reflect combined actions(‘synergies’) of body segments involved in the motor task(McKiernan et al. 1998; Park et al. 2001; Holdefer &Miller, 2002; Townsend et al. 2006; Jackson et al. 2007).In addition, in a study of tonic electrical stimulation ofthe pyramidal tract in decerebrated cats, Feldman andOrlovsky (1972) found that descending signals influencea specific neurophysiological variable – the threshold limbposition at which appropriate muscles are silent but readyto be recruited when stretched.

The threshold position can be considered as the originpoint of the spatial frame of reference (FR) for recruitmentof motoneurons (Feldman & Levin, 1995; Feldman, 2009).By producing threshold position resetting, the brain mayshift the spatial FR in which muscles are constrained towork without pre-determining how they should work(Feldman et al. 2007). Within the designated FR, musclesare activated or not depending on the gap between theactual and the threshold limb position as well as onthe rate of change of this gap. This fine organizationof muscle activation in the intact system vanishes aftermuscle deafferentation. Although deafferented patientsregain the possibility of muscle activation, the absenceof threshold position control results in numerous motordeficits, including the inability to stand or walk withoutassistance (e.g. Tunik et al. 2003; see also the website by Jacques Paillard devoted to deafferentation:http://jacquespaillard.apinc.org/deafferented).

The present study addresses the controversy regardingthe functional meaning of corticospinal excitability in thecontrol of joint position with a specific focus onthe question of whether or not the human motorcortex is involved in threshold position resetting formuscles spanning the wrist joint. To derive sometestable predictions, consider the threshold positioncontrol in more detail. Initially revealed in humans(Asatrian & Fel’dman, 1965), the existence of thresholdposition control has been confirmed in decerebrated cats.Vestibulo-, reticulo-, rubro- and cortico-spinal pathwaysinfluencing α-motoneurons directly (mono-synapticallyor pre-synaptically) or indirectly (via spinal interneuronsor γ-motoneurons) can reset the threshold length ofmuscles spanning the ankle joint (Matthews, 1959;Feldman & Orlovsky, 1972; Nichols & Steeves, 1986;Capaday, 1995; Nichols & Ross, 2009). The thresholdmuscle length is velocity dependent (Feldman, 1986). Italso depends on reflex reciprocal inhibition and otherheterogenic reflexes (Feldman & Orlovsky, 1972; Nichols& Ross, 2009). The importance of threshold positioncontrol is emphasized by findings that stroke in adultsand cerebral palsy in children limit the range of threshold

regulation, resulting in sensorimotor deficits such asabnormal muscle co-activation, weakness, spasticity andimpaired inter-joint coordination (Levin et al. 2000;Mihaltchev et al. 2005; Musampa et al. 2007).

The notion that descending systems have the capacity toset and shift the chosen spatial FR has many implications.In particular, it solves the classical posture–movementproblem described in the seminal paper by Von Holstand Mittelstaedt (1950). They emphasized that eachposture of the body or its segments is stabilized suchthat deviations from the posture are met with positionand velocity-dependent resistance generated by variousmuscle, reflex and central posture-stabilizing mechanisms.In contrast, intentional movements away from pre-viously stabilized postures do not evoke resistance fromthese mechanisms. To explain this difference, Von Holstand Mittelstaedt assumed that while sending motorcommands to initiate intentional motion, the nervoussystem simultaneously uses an efference copy (i.e. a copy ofmotor commands to muscle) to suppress motion-evokedafferent feedback. The system thus cancels the afferentinfluences that would otherwise cause resistance tomotion. This assumption conflicts with the findingthat afferent feedback remains functional throughoutintentional movements (Marsden et al. 1972, 1976;Matthews, 1986; Latash & Gottlieb, 1991; Feldman et al.1995; Feldman, 2009). In addition, in isotonic movements,the EMG activity can return to its pre-movement levelwhen the final posture is reached (Ostry & Feldman, 2003;Foisy & Feldman, 2006), implying that the copy of thisactivity (efference copy) also returns to its pre-movementlevel. As a consequence, the resistance resulting fromposition-dependent changes in the afferent signals is notcompensated by efference copy. The resistance would thusdrive the arm back to the initial position, a prediction ofthe efference copy theory that conflicts with experimentalobservations (Ostry & Feldman, 2003; Foisy & Feldman,2006).

In the framework of threshold position control, theposture–movement problem is solved in the following way.Posture-stabilizing mechanisms are effective only whenmuscles are active. This occurs if the limb is deviatedfrom a certain, threshold limb position (initial thresholdjoint angle, Ri in Fig. 1A; Feldman, 2009). Therefore, byshifting the threshold limb position (Ri → Rf (the finalthreshold joint angle)), the system resets (‘re-addresses’)posture-stabilizing mechanisms to a new posture. Theinitial posture appears as a deviation from the futureposture and the same posture-stabilizing mechanismsthat otherwise would resist the movement, now drive thelimb to the new posture. Thus, by shifting the thresholdposition, the nervous system converts movement-resistinginto movement-producing forces, which solves theposture–movement problem. Threshold position controlmay underlie not only isotonic movement but also

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1553

other motor actions (e.g. isometric torque generation ormovements against loads; Fig. 1B and C).

In the present study, we first verified that thresholdposition resetting actually occurs during active changesin the wrist joint angle in humans. Then we tested thepossibility that, in the same motor task, the cortico-spinal pathway, possibly together with other descendingpathways, is involved in threshold position resetting andthus in the specification of a spatial FR in which the neuro-muscular periphery is constrained to work. Note that if M1is involved in threshold position resetting, then its steadystate should change with transition to the new position.Specifically, to produce, say, a wrist flexion, the thresholdwrist angle should be changed in the flexion direction(Fig. 1A). This can be done by increasing corticospinalfacilitation of wrist flexor motoneurons while decreasingcorticospinal facilitation of extensor motoneurons. Thechanged threshold position should be maintained afterthe movement offset, even if the activity and excitabilityof α-motoneurons at the new position return to thosebefore the movement onset. We used a special techniqueto equalize the activity and excitability of α-motoneuronsat pre- and post-movement positions.

Alternatively, if M1 is directly involved in thespecification of EMG patterns, then corticospinalexcitability should return to its pre-movement level ifpost-movement EMG activity of wrist muscles returnsto its pre-movement level. In the present study, we usedtranscranial magnetic stimulation (TMS) to test thesealternative possibilities. In addition, we addressed thequestion of whether or not M1 is involved in thresholdposition resetting in wrist muscles when they are fullyrelaxed (Wachholder & Altenbruger, 1927). In the contextof threshold position control, full muscle relaxation isachieved by shifting the muscle activation thresholdbeyond the upper limit of the biomechanical muscle range(Levin et al. 2000), which reduces reflex responses toperturbations. Results were previously reported in abstractform (Burtet et al. 2007; Raptis et al. 2008, 2009).

Methods

Ethical approval

Nineteen healthy subjects participated in the study aftersigning an informed consent form approved by theinstitutional Ethics Committee (CRIR) in accordance withthe 1964 Declaration of Helsinki.

Subjects

All subjects (8 males and 11 females, age 34.4 ± 11.3 years;range 25–69 years) were right-handed (Edinburgh’s test).They were included in the study if they had no history of

neurological diseases (e.g. epilepsy) or physical deficits ofthe upper extremities. Subjects were excluded from thestudy if they took drugs that could affect the corticalexcitability (e.g. psychoactive drugs).

Apparatus

Subjects sat in a reclining dental chair that supportedthe head, neck and torso in a comfortable position withthe right forearm placed on the table (the elbow anglewas about 100 deg, horizontal shoulder abduction about45 deg; Fig. 2A). The head and neck were additionally

Figure 1. Threshold position controlThe same shift in the threshold position of a body segment from Ri toRf results in a change in the actual position of this segment in isotonic(zero load) condition (point a in A), in muscle torque in isometriccondition (point b in B), or in both the position and muscle torque inintermediate condition (point c in C). Points a, b and c are equilibriumpoints defined as the points of intersection between the leftcontinuous curve representing the final torque–angle characteristic(resulting from muscle activity regulated by proprioceptive feedback)and the dashed line (load torque–angle characteristic) in each panel.

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1554 H. Raptis and others J Physiol 588.9

stabilized with a cervical collar. The hand and forearmwere oriented horizontally in a neutral semi-supinatedposition. The hand with extended fingers was placed in aplastic splint attached to a light horizontal manipulandum.The hand was stabilized inside the splint with foam pads.The manipulandum could be rotated freely about a verticalaxis aligned with the flexion–extension axis of the wristjoint. The subjects were instructed to make wrist flexionor extension without flexing the fingers or making wrist

pronation or supination. They were thus instructed tominimize the involvement of degrees of freedom otherthan wrist flexion–extension. The motion of the forearmplaced on a table was minimized by Velcro straps attachedto the table. A torque motor (Parker iBE342G) connectedto the axis of the manipulandum was used in experimentswith nine subjects to produce brief stretches of wristmuscles to evaluate the excitability of motoneurons at twowrist positions (see below).

Figure 2. Equalizing EMG activity at two wrist anglesA, subjects placed the hand in a vertical splint of a horizontal manipulandum and repeatedly flexed (F) andextended (E) the wrist. B and C, without compensation of the passive components of muscle forces, the tonicEMG activity of wrist flexors was higher when maintaining position F vs. E (F/E EMG ratio > 1). Similarly, extensorswere more active at position E vs. F (E/F EMG ratio > 1, error bars are SDs). D–F, when passive wrist muscleforces were compensated by elastic, spring-like torques, the tonic EMG activity of each of 4 muscles substantiallydiminished and became position independent (EMG ratios ∼1.0). Asterisks indicate P < 0.001 for comparisonof no-compensation with compensation ratios (Kolmogorov–Smirnov test). Representative data from one subject(S7).

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1555

Procedures

Equalizing tonic EMG activity and excitability ofmotoneurons at two wrist positions. Responses to TMS –motor evoked potentials (MEPs) – were recorded from twowrist flexors and two extensors using disc-shaped Ag–AgClbipolar surface electrodes (1 cm diameter, 2–3 cm betweenthe disc centres) placed on the bellies of the flexor carpiradialis (FCR), flexor carpi ulnaris (FCU), extensor carpiradialis (ECR, long head) and extensor carpi ulnaris(ECU). EMG signals were amplified (Grass electro-myograph), filtered (30–500 Hz) and sampled at a rateof 5 kHz.

MEPs could depend not only on the corticospinalsignals transmitted to α-motoneurons either directly orindirectly, via spinal interneurons and γ-motoneurons,but also on the activity and excitability of α-motoneuronsresulting, in particular, from proprioceptive influences onthese motoneurons (Di Lazzaro et al. 1998; Todd et al.2003). We tried to equalize the states of α-motoneurons attwo wrist positions established by subjects (25–30 deg ofwrist extension and 40–45 deg of wrist flexion measuredrelative to the neutral position of 0 deg). If this is done,the difference in MEPs at two wrist positions couldreflect the difference in the corticospinal influences, ratherthan the difference in the excitability of α-motoneuronsof the muscles from which the MEPs were recorded.

The tonic EMG activity of wrist muscles usuallychanges with the transition from one position to another(Fig. 2B), which might be related to the necessity tocounteract the passive resistance of antagonist musclefibres and connective tissues at these positions. Witha deviation of the wrist from the neutral position, forexample, in the flexion direction, passive wrist extensorsare stretched such that tonic activation of flexors isnecessary to hold the wrist at the flexion positionand vice versa for extensors at the extension position.To exclude such position-related changes in the EMGactivity, we mechanically compensated the passive muscletorques. Initially, before we added a torque motor to themanipulandum, we used two elastics (Fig. 2D) to produceposition-dependent (spring-like) compensatory torques(in 7 subjects, S1–S7). One end of each elastic was attachedunderneath the manipulandum to a middle point located2 cm from the axis of the manipulandum. With rotationof the manipulandum away from the neutral position,the moment arm of the force of one elastic increasedwhereas that of the other elastic decreased such that, atthe extension wrist position (E in Fig. 2D), the net elastictorque assisted wrist extension and, at the flexion position(F in Fig. 2D), it assisted wrist flexion. By stretchingor shortening the elastics, it was possible to adjust theassisting forces, individually for each subject so that theycould minimize the tonic EMG activity at each position(compare Fig. 2E with B). In subsequent experiments with

nine participants (S8–S16), the passive muscle torqueswere compensated by similar linear spring-like torquesgenerated by a torque motor connected to the axis of themanipulandum (stiffness coefficient ∼0.003 Nm deg−1,adjusted individually for each subject). Note that thespring-like load, by itself, could not stabilize any position:in the absence of the hand in the splint, the manipulandumrapidly moved towards one or another extreme positiondetermined by mechanical safety stoppers.

Subjects were instructed to minimize, using EMG feed-back on the oscilloscope, co-contraction of wrist musclesat the initial and final positions. Trials in which the tonicEMG activity visually differed at the two wrist positions(1–2 trials per subject) were rejected. In off-line analysis,the residual tonic EMG activity (over 200 ms, 3 s after themovement offset) at each wrist position was measuredas the mean value of EMG envelopes obtained afterrectification and low-pass EMG filtering (0–15 Hz). It wasexpressed as a percentage of the EMG amplitude duringmaximal voluntary contraction, preliminarily measuredindividually for each subject. Subjects placed the hand inthe splint of the manipulandum fastened in the neutralposition. In response to a ‘go’ signal, they produced amaximal wrist flexor or, after about 10 min of rest, extensortorque for about 3 s. The maximal EMG was determinedas the mean of the maximal values of EMG envelopes alsoobtained after EMG rectification and low-pass filtering,and taken from two repetitions of torque production ineach direction. For all subjects, the minimized activity wasin the range of 3–5% from the mean EMG maximum. Foreach subject, the background EMG level was stable anddid not change significantly with position (see Results andFig. 2E and F).

Although tonic EMG levels were equalized at a near-zerolevel at the two wrist positions, the responses to the sameTMS pulse could still depend on the state of excitabilityof α-motoneurons at these positions. In humans, theexcitability of α-motoneurons of wrist muscles is usuallyevaluated by H-reflexes elicited by monopolar electricalstimulation of the median nerve. A major pre-condition ofusing H-reflex for evaluation of motoneuronal excitabilityis a stable M-response (Misiaszek, 2003; Knikou, 2008).By trying this technique, we found that when subjectschange the wrist position and contract wrist muscles,it is difficult to prevent mechanical displacements ofthe H-reflex stimulating electrode, resulting in artificialposition-related changes in both H- and M-responses. Inaddition, to get a reliable H-response, muscles shouldmaintain a contraction force higher than 5% of theMVC (Knikou, 2008), which would conflict with theinstruction to minimize the pre- and post-movementEMG activity in our study. The H-reflex is consideredas an electrical analog of the stretch responses (SRs;Knikou, 2008). Taking into account the limitations ofthe H-reflex technique in our study, we used instead a

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1556 H. Raptis and others J Physiol 588.9

physiologically more natural method in which theexcitability of motoneurons at the two wrist positionswas evaluated by comparing short-latency SRs of wristmuscles to a brief stretch (0.35 Nm, 100 ms) elicited bythe torque motor (Fig. 3) when passive muscle forces atthese positions were compensated by the same motor. Thesubjects were instructed not to intervene voluntarily inresponse to perturbations. The pulse was applied in eitherflexion or extension direction (chosen randomly, 4 blocksof 10 trials alternating between the two positions and twodirections of perturbations) every 5–10 s at each of the twowrist positions. An additional advantage of using stretchesinstead of H-reflexes was the possibility of evaluatingthe motoneuronal excitability of all four wrist musclessimultaneously, which is difficult to achieve with H-reflexstimulation. Some limitations of this technique are relatedto the anatomical difference in the muscle moment armsat the two positions. For the four muscles, the momentarms at the flexion position are about 20% higher than atthe extension wrist position (16, 22, 26 and 22% for FCR,FCU, ECR and ECU, respectively; Gonzalez et al. 1997). Atthe two wrist positions, we used the same torque pulse thatcaused a similar angular displacement (about 25 deg) andpeak velocity (about 300 deg s−1). Due to the differencesin the moment arms, similar angular changes resulted insomewhat different stretches of muscles at these positions.To overcome this problem, the computed magnitudes ofSRs of all muscles at the flexion position were downscaledby the percentage indicated above.

We compared the amplitudes of SRs of each recordedmuscle to perturbations in each direction at the two wristpositions during the first 30 ms after the SR onset (theinitial part of the first EMG burst). This short windowwas chosen in order to minimize the possible influenceof trans-cortical, triggered and intentional responses onthe evaluations of motoneuronal excitability (Crago et al.1976; Cheney & Fetz, 1984). These responses contributeto later EMG bursts elicited by wrist muscle stretching(Schuurmans et al. 2009).

Threshold wrist positions before and after intentionalmovement. Another method of evaluation ofmotoneuronal excitability in our study was basedon the measurement of threshold wrist positions.A threshold wrist position is the position at whichmotoneurons of wrist muscles are silent but ready to beactivated in response to a small central or reflex input.Physiologically, in the threshold state, the membranepotential of respective motoneurons is slightly belowtheir electrical threshold (Pilon & Feldman, 2006). Thisstate of motoneurons is quite different from their stateduring muscle relaxation when the membrane potentialsof motoneurons are far below their electrical thresholds(see the next section).

The threshold wrist position (λ) prior to active wristmovement onset and after its offset was determined basedon the SR onsets of wrist muscles at the two wrist positionsusing the following formulas. A muscle is activated if

x − λ∗ ≥ 0 (1)

where x is muscle length and λ∗ is its dynamic(velocity-dependent) threshold length (Feldman et al.2007). To a first approximation,

λ∗ = λ − μv (2)

where v is stretch velocity and μ is a time-dimensionalparameter related to the dynamic sensitivity of musclespindle afferents (Feldman, 1986).The threshold musclelength can be found by inserting eqn (2) into eqn (1) andtaking the equality sign:

λ = x + μv, (3)

where x and v are the length and stretch velocity atwhich muscle activation is initiated. While using eqn (3),two delays should be taken into account. The first isthe mechanical delay – the time required for the wristto reach the dynamic threshold – and the second is theminimal reflex delay in signalling the x and v values byproprioceptive afferents to motoneurons. The minimalreflex delay (27.5 ms, see Results) was evaluated in ourstudy based on the latency of EMG responses to stretchesof pre-activated wrist muscles. It has been previouslyidentified that parameter μ is within the range of 40–70 msfor a variety of muscles, such as those spanning the elbow,ankle, knee and hip joints (St-Onge et al. 1997; Gribbleet al. 1998; Pilon & Feldman, 2006; Feldman et al. 2007).To evaluate the maximal deviation of the threshold fromthe actual position (see eqn (3)), we chose the maximalvalue from this range, i.e. 70 ms.

Conventionally, the wrist angle is defined as increasingwith lengthening of wrist flexor muscles; the neutral wristposition resembles 0 deg, extension angles are positive andflexion angles are negative. Note that eqn (3) remains thesame if instead of the length-dimensional variables, λ, xand v, respective angular variables are considered. Thresholdjoint angles (R) were determined for the initial and finalwrist positions (Ri and Rf ), separately for flexors andextensors.

Comparison of the threshold state with the state ofmuscle relaxation. Since the EMG activity of wristmuscles was equalized at a near-zero level at the twowrist positions, one could suggest that subjects simplyrelaxed muscles at each position after motion to it. Thestate of full muscle relaxation is characterized not only bythe absence of EMG activity but also by the absence ofstretch reflex responses to substantial and rapid changesin the joint angle (Fig. 4A). In terms of the λ model,

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1557

the state of full muscle relaxation is achieved when thethreshold muscle length exceeds the maximal musclelength in the biomechanical range (Feldman, 1986). Inthis case, the muscle cannot be activated within the entirebiomechanical range unless it is stretched at a very highspeed as in the case of the knee reflex.

Using brief perturbations, the excitability ofmotoneurons was evaluated in eight subjects (S9–S16,10 trials per position and direction of stretch) not onlyduring active maintenance of the wrist positions but alsowhen subjects relaxed wrist muscles and the requiredpositions were established passively, by the experimenter.It appeared that subjects had difficulties in relaxingtheir wrist muscles and a short training was necessaryto help them in this. To ensure profound relaxation, theexperimenter first manually changed the position of themanipulandum in the whole biomechanical range ofwrist angle while the subject looked at the EMG activity ofthe four muscles displayed on an oscilloscope and tried,as requested, to minimize the background EMG andresponses to passive wrist displacements elicited by theexperimenter (Fig. 4A). This training lasted approximately5 min until EMG responses were excluded. After that, theexperimenter moved the manipulandum and establishedthe subject’s wrist flexion or extension position. Theexperimenter stopped holding the manipulandum justbefore the perturbation pulse was delivered. Thus, wetested that no SR could be evoked in the wrist muscles inresponse to mechanical pulses similar to those employedin testing SR during active positioning (Fig. 4B).

In three subjects, we determined whether or not musclestiffness (the slope of static torque–angle characteristic)associated with minimal EMG activity during activewrist positioning was different from that in the stateof full muscle relaxation. This was done by comparingwrist displacements from the flexion and extension wristpositions elicited by a small constant torque (±0.12 Nm)generated by the torque motor in the two musclestates. Only one-directional perturbations were appliedat each initial position (flexor torque at position E andextensor torque at position F) since perturbations in theother direction could bring the joint to the adjacentbiomechanical limit of the wrist joint when muscleswere relaxed. Subjects were instructed not to intervenevoluntarily to perturbations.

TMS

TMS was produced by single pulses applied to adouble-cone coil (70 mm outer diameter, 45 deg betweenthe axes of each half of the coil; Magstim 200, UK) suchthat the magnetic fields created by the currents in the twohalves were summated, resulting in a maximal stimulusat the intersection point (Rothwell et al. 1991; Hallett,

2007). TMS was delivered to the left M1. The TMS coilwas placed on the surface of the scalp in such a way thatthe point of intersection between the two circles of thecoil was approximately 2 cm anterior and 6 cm lateral tothe vertex (Cz), according to the 10-20 system for EEGelectrode placement (Jasper, 1958; Bonnard et al. 2003).From this position, the coil was moved (less than 0.5 cm)in the anterior–posterior and medial–lateral directionsto a position where the threshold for eliciting MEPs inwrist flexors or extensors was minimal (Wassermann et al.1992). MEPs were recorded electromyographically fromtwo wrist flexors and two extensors (see above).

The optimal spot for TMS was defined as elicitinga MEP of more than 50 μV in the ECR (17 subjects)or FCR (2 subjects) at a minimal stimulation intensity(motor threshold, MT) in at least 5 out of 10 sequentialtrials when the wrist was in the neutral position specifiedby the subject while maintaining minimal EMG activity.The TMS intensity was chosen individually in each sub-ject (range 1.2–1.4 MT for the group of subjects) to geta minimal supra-threshold stimulus that caused MEPsthat not only clearly exceeded the background EMG levelbut also had a stable amplitude (about 200 μV) in fivesequential trials in a flexor and an extensor wrist muscles.Since the sensitivity to TMS was different in differentsubjects, the most important characteristic of responses(relative stability) could not be ensured by a standard TMSintensity across all subjects.

Once the TMS intensity was determined, it wasunchanged in each experiment. The optimal point wasmarked with a felt pen on the scalp. Four marks intotal on the scalp and around the perimeter of the coilserved as a visual reference to maintain the coil positionthroughout the experiment. To let the subjects rest, thecoil was removed and re-placed from time to time butonly between different experimental sessions.

Subjects were asked to establish a 25 deg wrist extensionposition. At this initial position, a single TMS pulse wasdelivered. After about 2–3 s, in response to a go signal(a beep), subjects moved the wrist, in a self-paced way,to a 45 deg flexion position (the total change in the jointangle was about 70 deg). In order to establish the requiredpositions, subjects looked at a computer screen where theirwrist angle was displayed on-line. Two to three secondsafter the end of flexion movement, a second TMS pulsewas delivered. This pulse was thus produced in a final staticposition, at the time when not only the transitional EMGbursts responsible for the wrist movement, but also theterminal agonist–antagonist co-activation, usually visibleafter the movement offset, receded to a minimum.

The relatively long intervals for TMS prior to the onsetand after the offset of movement were chosen since wewanted to evaluate the difference in the corticospinalfacilitation at two steady-state positions, while avoidingtransitional changes in the MEPs resulting not only from

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1558 H. Raptis and others J Physiol 588.9

corticospinal co-facilitation of agonist and antagonistmuscles but also from temporary changes in the localexcitability of α-motoneurons following the movement.Such long inter-stimulus intervals were also chosen toavoid significant intra-cortical interactions due to the pre-vious TMS pulse (Rothwell et al. 1991). As verified byoff-line analysis (see below in the Methods and Results),the chosen time for TMS (2–3 s after the movementoffset) was sufficient for EMG signals and excitabilityof wrist motoneurons to recede to their pre-movementlevels.

After 15 s, the trial was finished and subjects returnedthe wrist to the extension position to prepare for thenext trial. In one block of 10–12 trials, the experimentaltesting started at the extension position (E → F sequence).In the second block (also 10–12 trials) testing started atthe flexion position (F → E sequence). The order of theseblocks as well as the order of other tests was randomizedacross subjects.

Changes in the corticospinal excitability couldspecifically be related to the established static wristposition or also reflect the history of reaching this position.For example, the same wrist position could be reachedafter wrist flexion or extension. To determine whether ornot the corticospinal excitability is characteristic of thewrist position or/and history of its reaching, we comparedMEPs at the same position (20 deg of flexion or 20 degof extension) established after motion from a more flexedor extended position (±20 deg, 10 trials per direction ofreaching and per position reached). This test was done insix subjects (S7, S9, S16–S19).

Changes in TMS responses following full musclerelaxation. After subjects fully relaxed wrist muscles (noSRs to rapid passive perturbations), the experimenterrotated the manipulandum at a moderate speed to bringthe wrist to the flexion or extension position. A TMSpulse was delivered at each position (5–10 s betweensequential pulses; 10 trials with a TMS pulse for eachposition). Directions of passive displacements towardseach wrist position were randomized. We compared theMEPs recorded at the two wrist positions establishedactively by subjects with those when subjects fully relaxedwrist muscles. In experiments involving stretch responses,active and passive positioning were conducted separatelyin the same day. Experiments in which history-dependenteffects were tested were conducted on a different day. Theorder of different experiments was randomized across sub-jects.

Data recording and analysis

Wrist position and velocity were measured with an opticalencoder coupled to the shaft of the manipulandum.

EMG activity, MEPs and wrist kinematics were recordedon-line, stored on a PC and analysed with LabViewand Matlab softwares specifically adapted to thisproject.

To compare the tonic EMG levels prior to TMS atthe initial and final positions, we used de-correlationtechniques (Dong, 1996; Battaglia et al. 2007; Clancyet al. 2008). It takes into account that low-amplituderegular signals usually coming from external electricalsources such as power supplies can induce someundesirable correlation within and between EMG samples.De-correlation reduces autocorrelation within a signal, orcross-correlation between different samples of the signalwhile preserving their basic aspects. Although small, theexternal regular signals could produce noticeable peaksin the power spectrum of EMG. After eliminating thesepeaks, computer software produced an inverse Fouriertransformation of the EMG power spectrum to obtainEMG samples that were less affected by these signals.The software then found the time interval after whichthe auto-correlation function for each of the two residualEMG samples fell below 10%. This interval was used asthe new sampling interval for selecting 20 de-correlatedEMG values from each of the two EMG samples. Theduration of the original pre- and post-movement EMGsamples preceding TMS to which this procedure wasapplied exceeded 60 ms. Based on the de-correlated values,the EMG samples were compared (Kolmogorov–Smirnovtest, the level of significance P < 0.05). Data were analysedtrial by trial, individually for each subject.

For each of the four muscles in each trial we measuredthe MEP peak-to-peak amplitude, area and latency (thetime between the TMS artefact and the first sign that theEMG deviates from the background activity) at the initialand final positions. In order to evaluate the differencein the MEPs at the two wrist positions, we comparedMEP amplitudes at these positions, individually for eachmuscle and subject. For group averaging and groupstatistics, MEPs were normalized by dividing each muscleaverage response by the value of the maximal averageMEP amplitude, individually for each subject, and thenaveraged across subjects.

Statistical analysis

The influence of position on the MEPs was assessedby calculating the mean and standard deviation (SD,represented by error bars in histograms) of the amplitudes,areas and latencies individually for each muscle. Itappeared that MEP variances at the extension and flexionpositions were different and that the values did not followa normal distribution (Levene’s test, Shapiro–Wilk test,P < 0.05). Therefore, we used non-parametric tests forstatistical analyses. The Kolmogorov–Smirnov test was

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1559

used to compare measures in the same subject (e.g. flexionvs. extension, active vs. relaxed state) while Wilcoxonmatched-pairs test was used to study the significance ofgroup results for the same conditions. After verificationof the variance similarity and distribution normality ofthe EMG envelope amplitudes elicited after mechanicalperturbations, dependent samples t tests were used toevaluate the difference between muscle reflex reactions inthe group of subjects tested in the two wrist positions.For individual results, amplitudes of EMG (envelope)responses at the two wrist positions were compared foreach perturbation direction using Kolmogorov–Smirnovtest. The significance level of P < 0.05 was chosen in alltests.

Results

Threshold position resetting associated with activetransition from one wrist position to another

When passive muscle torques at the two selected wristpositions were compensated, the EMG activity of allrecorded wrist muscles at the initial and final positionswas minimized. After transitional EMG bursts during themovement, the activity of each of the four wrist musclesin all subjects was reduced to its pre-movement level(Kolmogorov–Smirnov test for two samples, P > 0.05;Fig. 2E and F), with the exception of one or, very rarely,two muscles in some trials (trials in which EMG levelswere not equalized were excluded from the analysis ofMEPs). In most cases for the group of subjects (95.1%),the EMG activity at the initial and final positions did notdiffer and was often only slightly above the backgroundnoise in EMG recordings (<5% from the EMG level duringmaximal voluntary contraction).

To test whether or not the reflex excitability ofmotoneurons was different at the two actively establishedwrist positions, we recorded EMG stretch responses (SRs)to torque pulses generated by the motor, in eight sub-jects. The pulses caused similar wrist deviations andvelocities at these positions (Fig. 3A and B). Torque pulsesin the extension direction initially evoked SRs in wristflexors (FCU, FCR) and those in flexion direction evokedSRs in wrist extensors (ECU, ECR). All four musclesshowed distinctive SRs (Fig. 3A and B) in both wristpositions. After these initial SRs, muscles usually generatedseveral EMG bursts alternating between wrist flexors andextensors.

We compared latencies of EMG SRs at the two wristpositions measured from the beginning of perturbation.These latencies did not differ between positions for eachof four muscles (P > 0.05) and were in the range of51 ± 9 ms. The comparatively long latencies of SRs couldresult from the use of moderate stretches in terms of themagnitude and speed: SR latency tested by faster stretches

of pre-activated muscles in our study was about 25–33 ms(mean 27.5 ms).

The amplitudes of short-latency SRs were evaluated bythe maximal values of rectified EMG envelopes during thefirst 30 ms after EMG onsets. By focusing on the earliestcomponent of SRs, we excluded long-latency responsesthat could be mediated by trans-cortical loops or resultedfrom intentional or triggered reactions to perturbations.To minimize the effect of the difference in the momentarms at the two wrist positions, the SRs at the flexionposition were downscaled as described in the Methods.

Overall, for the group and all muscles (4 muscles in8 subjects tested, 32 muscles in total), the amplitudes ofshort-latency SRs normalized to the maximum of EMGresponses within the 30 ms windows (individually for eachsubject, and then averaged across subjects; Fig. 3D and E)did not differ at the two wrist positions (P = 0.64 forFCR, P = 0.26 for ECU, P = 0.68 for ECR and P = 0.06for ECU, dependent samples t tests). Thus, in most cases,the excitability states of wrist motoneurons at the twopositions were indistinguishable in terms of latency andmagnitude of SRs, with some exceptions (included in thestatistical analyses): in 1 of 4 muscles in three subjects, SRswere smaller at the position at which the length of thismuscle was larger (e.g. in ECU when the wrist was flexed);in two other subjects, responses at the two positions tendedto differ in 2 out of 4 muscles. In total, the individualstatistical tests (Kolmogorov–Smirnov) showed that SRsdid not differ at the two positions in 78% of cases (25/32muscles) and that, in most cases, not only the EMG levelsbut also the excitability of α-motoneurons of wrist musclesat the two actively established wrist positions were similar.

Even in those cases when muscles were silent most ofthe time at either of the two positions, they irregularlyshowed small intermittent EMG activity. This implies thatmotoneurons of these muscles were near their recruitmentthresholds and were activated from time to time. Tofurther verify this suggestion, we used SRs and eqn (3)to determine the recruitment threshold wrist positionsbefore the onset and after the offset of intentionalmovements. We took into account that, because ofreflex delay in the transmission of afferent signals toα-motoneurons and muscles, the values of position andvelocity that were responsible for SRs actually occurredearlier than the measured latency of SRs (51 ± 9 ms).The transmission delay was measured as the latency ofEMG responses to stretching of pre-activated muscles atthe pre- and post-movement positions, in four subjects(see Methods). For all muscles and subjects, the trans-mission delay was in a narrow range of 25–33 ms (mean27.5 ms). We then measured the angular displacementand velocity 27.5 ms before SR onsets at the pre- andpost-movement positions to evaluate the respective initialand final threshold positions (Ri and Rf ; see Methods).The difference between the threshold positions and the

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1560 H. Raptis and others J Physiol 588.9

respective actual wrist positions for each of eight subjectsdid not exceed 2–7 deg, which is substantially less than thedifference between the pre- and post-movement positions(about 70 deg). We thus confirmed that motoneuronsof wrist muscles at the initial and final positions werenear their recruitment thresholds and that the thresholdwrist position (i.e. the position at which muscles werealmost silent but ready to be activated when stretched)was reset when the wrist angle was changed, by a valuethat was close to the actual angular wrist displacement(about 70 deg).

Muscle relaxation

The threshold states in actively specified positions weresubstantially different from the state when wrist muscleswere relaxed and the initial and final positions wereestablished passively, by the experimenter. It appeared that

most subjects had difficulties in relaxing wrist musclesto fully exclude SRs to perturbations. This was achievedafter some training with the use of EMG feedback andrandom passive joint rotations in the whole biomechanicalrange. Subjects were instructed to exclude EMG responsesto perturbations. After training, no significant EMGresponses were observed during either passive wristrotations by the experimenter or pulse perturbationselicited by the torque motor (Fig. 4). In this state,only 3–5 times stronger and more rapid perturbationscould evoke SRs comparable to those during active wristpositioning.

The difference between the threshold and relaxationmuscle states is also reflected in muscle stiffness (slopeof torque–angle characteristic), evaluated by suddenapplication of small constant torques (±0.12 Nm) atthe two wrist positions and by measuring the wristdisplacement elicited by these torques (see Methods).During active positioning, muscle stiffness was 6–32 times

Figure 3. Testing motoneuronalexcitability at the actively establishedpositions E and F based on early EMGSRs to perturbationsA, perturbations in the extension directionelicited a short-latency SR in wrist flexors(FCR, FCU) and a later response inextensors (ECR, ECU). B, perturbations inthe flexion direction elicited a short-latencySR in stretched extensors and a laterresponse in flexors. C, pre-perturbationEMG levels in position E and F (normalizedto the maximal SR amplitude, individuallyfor each muscle and subject) were similar(P > 0.05). D and E, earliest (30 msduration) EMG SRs were also similar at thetwo positions.

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1561

higher than during relaxation at the same positions in thethree subjects tested (P < 0.001; Fig. 5).

Corticospinal excitability at different activelyspecified wrist positions

We compared the MEPs in each muscle only in those caseswhen the EMG activity of wrist muscles was equalizedat the two positions. In each trial, a TMS pulse wasproduced before and after a voluntary wrist movementfrom extension to flexion (E → F movement) and viceversa (F → E movement).

The same TMS pulse regularly elicited a brief wristflexor jerk at the flexion position and an extensor jerkat the extension position (Fig. 6). This behaviour wasobserved in 10 out of 16 subjects. In four subjects, onlyan extensor jerk was elicited at the extension positionand in two subjects no mechanical response to TMSwas observed at either position (even if MEPs were pre-sent). The reciprocal changes in motor responses to thesame TMS pulse imply that the states of M1 at the twowrist positions were substantially different, even thoughthe states of wrist motoneurons were similar. Note thatthis finding shows that changes in the M1 were stronglyrelated to position but, like in the cases of correlations

Figure 4. Muscle relaxationA, absence of EMG responses of fully relaxed wrist muscles to passivemovements in the whole biomechanical joint range. B, in contrast toactive wrist positioning (Fig. 3), no SR occurred in response to forcepulses in relaxed muscles.

between variables, it does not imply causality (functionaldependency) underlying this relationship: it could beobserved if, for example, the M1 caused changes in wristposition or, vice versa, changes in wrist position couldinfluence the M1 excitability (e.g. via trans-cortical loops;see Discussion).

Position-related changes in the corticospinal excitabilitywere also evaluated by measuring the MEP peak-to-peakamplitude and area of the rectified MEPs in wristextensors (ECR, ECU) and flexors (FCR, FCU) at thetwo wrist positions. The two methods yielded consistentresults. A typical recording of wrist position, MEPs andEMG activity in a representative subject are shown inFig. 6A for E → F → E movement. Two seconds beforethe movement onset (left segment), the EMG activityof wrist muscles was close to zero (background noiselevel). After transient EMG bursts, the wrist reached theflexion position, at which the EMG activity graduallyreturned to its pre-movement, near-zero level. Althoughthe EMG activity of all muscles was minimal at eitherof the two positions, the MEP amplitude substantiallychanged with the transition from one position to another(for S15 in Fig. 6A, P < 0.01 for FCR, ECR, ECU andP < 0.1 for FCU, Kolmogorov–Smirnov test): MEPs ofwrist flexors at the flexion position exceeded those at theextension position, and vice versa for extensor MEPs. Thehistogram for the group of subjects (Fig. 6C) shows similarposition-related differences: MEPs for all muscles acrossthe repeated trials were different at F and E positions (forFCR, P = 0.0008; FCU, P = 0.003; ECR, P = 0.0003; ECU,P = 0.0003, Wilcoxon matched-pairs test).

Reciprocal changes in flexor and extensor MEPs(extensor MEPs at E position exceeded those at F positionand vice versa for flexor MEPs) were also observed in theremaining subjects although not always in all muscles.Specifically, in 6 out of 16 subjects, the position-relatedchanges were significant for MEPs of all four muscles

Figure 5. Muscle stiffness in threshold and relaxation statesAlthough EMG activity was minimal at both wrist positions (F and E),wrist stiffness (the slope of torque–angle characteristic) during activewrist position control was substantially higher than during full musclerelaxation at the same positions but established passively.

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1562 H. Raptis and others J Physiol 588.9

(Kolmogorov–Smirnov test). In nine subjects, this was thecase in 3 out of 4 muscles, while reciprocal changes in MEPsof the fourth muscle were at the border of significancein 3 of these 9 subjects. In one subject, reciprocal MEPchanges were significant in only two (FCR, ECR) muscles(MEP changes in FCU were on the border of significance).Thus, except for few cases, the TMS responses wereposition-related, both for flexors and extensors. This resultwas obtained regardless of whether the passive muscletorques were compensated by elastics (7 subjects) or bythe torque motor (9 subjects; Kolmogorov–Smirnov test,P > 0.05).

To determine whether or not corticospinal excitabilityat each of the two positions depended on the directionof movement towards these positions, we first compared

MEPs at F and E positions when F → E or E → Fmovements were made. The movement direction effectwas not significant for all four muscles (Wilcoxonmatched-pairs test, FCR: P = 0.12, FCU: P = 0.12, ECR:P = 0.60, ECU: P = 0.92). In an additional experiment,subjects (n = 6) reached 20 deg of wrist extension froma more extended position (40 deg) or from the neutralposition (0 deg). The corticospinal excitability (MEPs) at20 deg of extension was independent of how this positionwas reached (P > 0.1 for all muscles, Fig. 7). A similarresult was obtained for reaching 20 deg of wrist flexionafter opposite-direction movements. These results implythat the corticospinal influences on wrist motoneurons ata given position were the same regardless of whether itwas reached by previous active wrist extension or flexion.

Figure 6. Typical mechanical and EMG responses to TMS at two static actively established positions,F and E, when the EMG activity and excitability of motoneurons of wrist muscles were equalized (as inFig. 3)A, the left panel shows the wrist angle and MEPs for 4 wrist muscles at position E before movement. The nextpanel shows EMG activity during wrist movement from position E to position F at which a second TMS pulse wasdelivered. The last 2 panels show movement to position E and kinematic and EMG responses to TMS pulse atthis position. The same TMS pulse elicited a flexor jerk at position F and an extensor jerk at position E (verticalarrows). Although the excitability of motoneurons was similar at the two positions, flexor MEPs at position F weresubstantially bigger than at position E and vice versa for extensor MEPs (reciprocal pattern). B, pre-TMS EMGlevels in position E and F (normalized as in Fig. 3C) were similar (P > 0.05). C, group mean MEP amplitudes for 16subjects.

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1563

In other words, TMS responses reflected the corticospinalinfluences at each position, rather than the history of itsreaching.

MEP latency was measured by identifying the firstdeflection of the EMG trace from the background levelafter the TMS artefact. In all cases of active positioning,the latency was in the range 14–24 ms (mean 18.8 ± 0.3 msfor all muscles and positions). The wrist position andmovement direction had no effect on the latency of MEPs(P > 0.1, Wilcoxon matched-pairs test for the group).

Corticospinal excitability before and after musclerelaxation

Results described in the previous sections showed thatmotoneurons of wrist muscles at the two activelyestablished wrist positions were near their recruitmentthresholds and were ready to be activated in responseto external perturbations, which was not the case formotoneurons after muscle relaxation (Fig. 4). The stateof M1 also changed with muscle relaxation. For example,the same TMS pulse that produced an wrist extensorjerk at the extension position but a flexor jerk at theflexion position established actively, became inefficientin eliciting a jerk after muscle relaxation at the same

Figure 7. Corticospinal excitability reflects the state of M1 thatis specific to each wrist position, regardless of how it wasreachedA, mean normalized MEP amplitudes at extension position(E = 20 deg) that was reached either by wrist flexion from positionE +20 deg or by extension from position E –20 deg. B, similar test fora flexion position (F = −20 deg), after leaving a more extended ormore flexed position. Error bars indicate SDs. Reaching direction hadno effect on MEPs (P > 0.1; data for 6 subjects).

positions established passively (Fig. 8A), in all subjects.Muscle relaxation also resulted in a decrease in MEPamplitude and area at both positions in 93.1% of allcases, by a factor of 1.5–5.0 depending on the muscleand subject (P < 0.05 for group, Wilcoxon matched-pairstest, Fig. 8C). The reciprocal pattern of changes in MEPs,characteristic of active positioning, became less systematicafter muscle relaxation (only in 3 muscles in 2 out of the 9subjects tested; in 1 or 2 muscles in 2 other subjects). In 5out of 9 subjects, no position-related changes in the MEPamplitude were observed after relaxation (P > 0.05 for allmuscles in these subjects, Kolmogorov–Smirnov test; forthe group, MEP amplitudes appeared position-related in2 muscles out of 4; P = 0.03 for FCU and ECR, P > 0.05for FCR and ECU, Wilcoxon matched-pairs test; Fig. 8C).

As in active wrist positioning, MEP latency after musclerelaxation was similar in the two wrist positions (P > 0.2)but mean latency significantly increased, by about 3 ms,from 18.8 ± 0.3 to 21.6 ± 0.5 ms (P < 0.05, Wilcoxonmatched-pairs test).

Discussion

Threshold position control by descending systems

It is known that MEPs might depend not only on cortico-spinal influences but also on motoneuronal excitability(Komori et al. 1992). To minimize the role of the latterfactor in the evaluation of corticospinal excitability attwo wrist positions, we compensated passive muscletorques and thus equalized EMG levels at these positions.It appeared that motoneuronal excitability was alsoequalized as shown by the similarity in the short (30 ms)initial components of muscle SRs at these positions. Thesecomponents probably reflected not only mono-synapticSRs of motoneurons but also pre- and poly-synapticSRs of motoneurons mediated by spinal interneurons.Corticospinal influences on motoneurons could also bemediated by spinal interneurons of reflex loops (Porter,1985; Lemon, 2008). However, this type of cortico-spinal influence could not play a major role in theevaluation of excitability of motoneurons based on theshort initial components of SRs, otherwise these SRcomponents would reflect the position-related changesfound in MEPs. Instead, these components were similarfor the two wrist positions. In the additional test ofmotoneuronal excitability, we found that, at both wristpositions, motoneurons of wrist muscles were neartheir activation thresholds. This test of motoneuronalexcitability was based on detecting the earliest sign, ratherthan the magnitude of stretch-evoked EMG activity. Inother words, this test relied on mono-synaptic SRs ofmotoneurons. The similarity of MEP latencies at the twopositions also supports the assertion that motoneuronalexcitability did not differ at these positions. Our data

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1564 H. Raptis and others J Physiol 588.9

thus confirmed the assumption that motoneurons ofwrist muscles at the initial and final positions were neartheir recruitment thresholds and that the threshold wristposition (i.e. the position at which muscles were almostsilent but ready to be activated in response to perturbation)was reset when the wrist angle was actively changed.Our analysis thus complements previous findings ofthreshold position resetting in intentional single- andmulti-joint movements in humans (St-Onge & Feldman,2004; Archambault et al. 2005; Foisy & Feldman, 2006).Such resetting is fundamental in controlling movements

without any concern for the posture–movement problem(see Introduction).

Our results showed that changes in corticospinalexcitability accompanying threshold position resettingmight be unrelated to and independent of EMG activity.They confirm the hypothesis that the corticospinalinfluences are involved in threshold position resetting,possibly in combinations with influences of otherdescending systems. One should also have in mind thatthe threshold wrist position may or may not coincidewith the actual wrist position (see Fig. 1 and Results).

Figure 8. Responses to TMS (intensity as in Fig. 6) at positions E and F established passively, by theexperimenter, after muscle relaxationA, compared to active specifications of wrist positions (Fig. 6A), mechanical responses to TMS (jerks) were absentand MEPs substantially diminished. B, pre-TMS EMG levels in position E and F (normalized as in Fig. 3C) weresimilar (P > 0.05). C, group mean normalized MEP amplitudes. After relaxation, position-related changes in MEPamplitudes were observed only for 2 of 4 muscles. D, histogram for the group showing a decrease in MEPamplitudes for all muscles after relaxation compared to those in active positioning, at both wrist positions.

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1565

These positions should be treated as different variables.The threshold position can be considered as a parameterthat defines the origin (referent) point in the spatialFR in which wrist muscles start their recruitment orde-recruitment (see Introduction). In other words, byinfluencing the origin point, descending systems wouldspecify a spatial FR in which the neuromuscular peripheryis constrained to work whereas EMG activity emergesdepending on the difference between the actual and thethreshold position.

In a small number of cases (in 1 of 4 muscles in9 out of 16 subjects), MEPs did not differ at the twowrist positions. This observation does not necessarilyconflict with the suggestion that the corticospinal systemis involved in threshold position control. It might reflectthe complexity of neural interactions between wristmuscles (the ‘enslaving’ phenomenon; Zatsiorsky et al.2000; Schieber & Santello, 2004). Despite the instructionto use only one, flexion–extension degree of freedom(DF), subjects could unintentionally involve other wristDFs: although the hand was fixed in the splint of themanipulandum, subjects could exert small pressure on thesplint walls in the pronation or supination direction. Inthese comparatively rare cases, corticospinal excitabilityand, as consequence, the resulting pattern of thresholdposition resetting could depend on how different DFs werecombined (Ginanneschi et al. 2005, 2006).

Corticospinal pathways, possibly in combination withother descending pathways accomplish independent(‘open-loop’) control of motoneurons

While discussing our findings, one should have in mindthat TMS applied over the primary motor cortex is knownto influence motoneurons not only via the corticospinalbut also via other corticofugal pathways. The most efficientTMS spot in the M1 for cortico-reticulo-spinal effects isdifferent from the M1 spot we stimulated and requiresmuch higher intensity of TMS compared to that usedin our study (Gerloff et al. 1998; Ziemann et al. 1999).The latency of these effects also exceeds that of responsesobserved in our study. Therefore, it seems unlikely that, inour study, cortico-reticulo-spinal pathways played a majorrole in the position-related changes in MEPs. However,corticospinal neurons may send collaterals to neurons inrubro-, reticulo- and vestibulo-spinal descending systemsinfluencing motoneurons (Keizer & Kuypers, 1984, 1989).Therefore, the changes in MEPs associated with wristrepositioning in our study might reflect in part the changesin the excitability of other descending systems.

Our findings revealed concomitant changes in theMEP and wrist position (Figs 6 and 7). One can saythat these changes were correlated. As in other casesof observations of correlation between variables, theposition-related MEP changes do not imply causality

between the variables. Figure 1 shows that, in differentexternal conditions, the same control signal (a changein the threshold joint angle, R) generated independentlyof position and force can correlate with one or bothof these variables depending on external conditions. Wesuggest that corticospinal influences evaluated by MEPswere functionally independent of the actual wrist positionbut they became correlated with this position becauseof the isotonic condition of our experiments. Considerfirst alternative hypotheses. Corticospinal influences couldfunctionally depend on wrist position if proprioceptivemuscle length-dependent signals resulting from activewrist movement influenced motoneurons via spinaland/or trans-cortical reflex loops. According to thishypothesis, it is the neuromuscular periphery, ratherthan corticospinal and other descending influences, thatwould be responsible for correlation of MEPs with wristposition. Spinal reflexes per se, such as the stretch reflexand reciprocal Ia inhibition could not be responsiblefor the observable correlation, for two reasons. (1) Theshortening of agonist muscles during wrist motion wouldresult in a decrease in the reflex autogenic facilitationof agonist motoneurons following shortening of musclespindles as well as from Ia inhibition of these motoneuronsfollowing lengthening of antagonist muscles. In contrast,our results showed that MEPs of shortening agonistmuscles increased with the transition to the new position.(2) Such reflexes alone would change the excitability ofmotoneurons in a position-dependent way whereas ourtests showed that excitability of motoneurons was the sameat both positions.

It is also unlikely that MEPs became correlatedwith position because of trans-cortical reflexes. (1) Ifpresent (Wiesendanger, 1986; Macefield et al. 1996),such reflexes usually represent long-latent supplementsto the spinal stretch reflex (Evarts & Fromm, 1981;Matthews, 1991), although the pattern of responses canbe different in the case of strong perturbations (not usedin the present study), resulting in involuntary discrete(‘triggered reactions’) or voluntary responses (Crago et al.1976). Thus, like the spinal stretch-reflex (see above),stretch-reflex-like trans-cortical influences on the M1could not be responsible for positional MEP relation.(2) To be consistent with observed MEP changes, thetrans-cortical reflex facilitation of motoneurons shouldincrease with muscle shortening, i.e. be in oppositionto the stretch reflex. Such trans-cortical reflex wouldcorrespond, in engineering terms, to positive feedbackthat could destabilize wrist position control. In particular,once initiated, wrist flexion could be augmented by suchfeedback, eventually flexing the joint to its biomechanicallimit.

Consider the hypothesis that motoneurons of wristmuscles were controlled by corticospinal influencesindependently of wrist position, i.e. independently of the

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1566 H. Raptis and others J Physiol 588.9

changes in proprioceptive feedback resulting from thetransition from one wrist position to another. In iso-tonic conditions, this facilitation could elicit a changein wrist position (Fig. 1A), resulting in correlationbetween these events. The finding that changes incorticospinal facilitation of agonist motoneurons startprior to the onset of muscle activation in rapid wristmovements (MacKinnon & Rothwell, 2000; Schneideret al. 2004; Irlbacher et al. 2006) is also a sign ofindependent initiation of corticospinal facilitation. Wecall this strategy ‘open-loop’ control. It does not implyan absolute independence on the periphery. Based on pre-vious proprioceptive and other sensory information, thesystem may decide how to change corticospinal influencesaccording to the task demand. However, once a decision ismade, these influences are accomplished independently ofperipheral feedback, unless this feedback or other sensorysignals call on the necessity of movement correction.

Our study implies that previous observations ofcorrelations of cortical activity with electromyographicand mechanical variables (e.g. Holdefer & Miller, 2002;Kurtzer et al. 2005; Townsend et al. 2006; Jacksonet al. 2007; Griffin et al. 2009) may not be sufficientto conclude that M1 is involved in programming ofthese variables (see also Shah et al. 2004). An additionallimitation of studies that found correlation of corticalactivity with arm posture (e.g. Fortier et al. 1993;Kalaska et al. 1997) is that they leave unanswered thequestion what posture (actual or threshold) the cortexdeals with. Future studies are also needed to test thepossibility that, even during the dynamic phase offast movement from one position to another, cortico-spinal facilitation remains predominantly independentof, although influences and therefore correlates with,the emerging EMG patterns and kinematic variables(cf. MacKinnon & Rothwell, 2000; Irlbacher et al. 2006;Kalaska, 2009). Our study demonstrates the disparitybetween corticospinal facilitation and tonic EMG patterns.Preliminary data showed that this might also be the casefor the dynamic phase of rapid point-to-point movements(Raptis et al. 2009).

The motor cortex and muscle relaxation

Our findings suggest that muscles can appear inactivewhen subjects either relax wrist muscles or maintaina wrist posture by holding host motoneurons neartheir activation thresholds. These states of the neuro-muscular system are substantially different: the formerbut not the latter case is associated with minimal, if notabsent, reflex reactions to perturbation and low stiffness(see also Marsden et al. 1976). The absence of SRs atboth wrist positions when subjects were relaxing impliesthat the excitability of α-motoneurons was suppressedby descending systems during muscle relaxation. This

process was probably accompanied by a decrease indescending facilitation of γ-motoneurons, thus reducingthe positional sensitivity of muscle spindle afferents(Lennestrand & Thoden, 1968). After achieving musclerelaxation, subjects could tolerate passive changes in thewrist angle by keeping descending influences unchanged.In this case, low sensitivity of muscle spindle afferent couldbe responsible for the low positional dependency of MEPsobserved in our study.

Note that, unlike active positioning accomplishedby reciprocal changes in the descending influences onagonist and antagonist motoneurons, relaxation involvesparallel de-facilitation or inhibition of wrist agonist andantagonist motoneurons by descending systems. Themotor cortex has been shown to have both types ofprojection on motoneurons (Lemon, 2008). However,a decrease in motoneuronal excitability associated withmuscle relaxation could lead to a decrease in MEPs withouta concomitant decrease in corticospinal facilitation.Therefore, although the participation of the cortico-spinal influences in muscle relaxation seems likely, ourfindings do not rule out the possibility that relaxation isaccomplished by other descending systems.

In a previous study, changes in MEPs were analysedduring passive muscle shortening and lengthening (Lewiset al. 2001), yielding ambiguous results – MEPs couldincrease, decrease or remain unchanged with musclelengthening. The ambiguity might result from severalfactors. Subjects could produce different amounts ofmuscle relaxation resembling different sub-thresholdstates of motoneurons. They could also either beindifferent to or assist passive hand motion by maintainingthe same or modulating corticospinal influences at asub-threshold level. While not conflicting with these data,our findings do show that muscle relaxation, if ensured byappropriate tests, is associated with a substantial reductionof the amplitude and position-related modulation ofMEPs, compared with the cases when wrist angle is activelyspecified by subjects.

Neurophysiological interpretations

Threshold position control implies that, due toproprioceptive feedback, electrochemical synaptic signalsfrom the brain are transformed (‘decoded’) into aposition-dimensional variable, R. In this way, actionsbecome related to body space. The membrane ofα-motoneurons may be the site of this transformation(Pilon et al. 2007; Feldman, 2009). Based on thisassumption, one can explain basic findings in the presentstudy (Fig. 9).

When, say, a flexor muscle is quasi-staticallystretched, position-dependent motoneuronal facilitation,predominantly from muscle spindle afferents, increases(Fig. 9A, low diagonal line) until the electrical threshold

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1567

for motoneuronal recruitment is reached. This electricalthreshold is reached at a certain muscle length or jointangle Re. In our experiments, this threshold angle could beclose to the wrist actual extension position, Qe. In contrast,if descending systems elicit a position-independent increasein the membrane potential (vertical arrow), then thesame muscle stretch results in muscle recruitment at asmaller threshold joint angle, Rf , that could be close tothe actual flexion position (Qf ). Note that the spatialthreshold can be reset even if the electrical thresholdof motoneurons (V +) remains constant. Changes inelectrical motoneuronal threshold (Krawitz et al. 2001;Fedirchuk & Dai, 2004) may be an additional sourceof shifting the spatial threshold (see Pilon & Feldman,2006). Thus, the amount of threshold position resettingis a physiological measure of independent descendinginfluences on motoneurons. Therefore, by changing theseinfluences (either directly or indirectly, via interneuronsor γ-motoneurons), descending systems, including thecorticospinal system, may participate in threshold positionresetting (Re → Rf ).

Figure 9A also illustrates how, physiologically, cortico-spinal influences could elicit resetting of the thresholdposition of a body segment, while maintaining thesame motoneuronal activity and excitability at the pre-and post-movement positions. The basic idea is that,because of muscle shortening, the autogenic proprio-ceptive facilitation of flexor muscles would decrease withthe transition from an extension to a flexion position.

Figure 9. Physiological origin of threshold position control andexplanation of basic findingsA, integration of position-dependent proprioceptive (lower diagonalline) and position-independent central inputs (vertical arrow) to aflexor α-motoneuron. The central input shifts the threshold positionfor motoneuronal recruitment (Re → Rf). B, full muscle relaxation inthe whole biomechanical joint range [Q−, Q+] is achieved byminimizing the corticospinal facilitatory influences as well as those ofother descending systems such that the threshold angle appearsoutside that range (R+ > Q+ for flexors, R− < Q− for extensors).Following the decrease in corticospinal facilitation and excitability ofmotoneurons, TMS responses in relaxed muscles are diminishedcompared to responses to TMS pulses applied during active wristpositioning.

However, the independent corticospinal facilitation(partly mediated by γ-motoneurons) restores the stateof flexor α-motoneurons to its pre-movement level,but at the new threshold position, Rf . Figure 9A alsoshows that descending facilitation of flexors at a flexionposition should be greater than at an extension position,even though, at both positions, motoneurons are neartheir recruitment thresholds and therefore have thesame excitability. A similar diagram can be plottedfor extensor motoneurons by taking into account thatthe length of extensors decreases with the joint angle;corticospinal facilitation of extensors should be greaterat the extension position, which explains the reciprocalchanges in corticospinal facilitation with the change inwrist position. Figure 9B shows that by shifting muscleactivation threshold outside the biomechanical range (toR+ for flexors and to R− for extensors, the system canexclude muscle activation in the whole biomechanicalrange of the joint (from Q− to Q+), thus accomplishingfull muscle relaxation. Our data imply that the M1 mayparticipate in threshold position resetting underlyingnot only intentional changes of posture but also musclerelaxation.

Figure 9A illustrates that threshold position resettingmay result from sub-threshold changes in the state ofmotoneurons (Ghafouri & Feldman, 2001; Feldman,2009). In other words, such resetting is a feedforwardprocess in the sense that it starts prior to the EMGonset and can be accomplished before the end of theresulting motor action. Therefore, feedforward and pre-dictive properties may result from natural physiologicalprocesses in the absence of any computations based oninternal models of the system dynamics (Dubois, 2001;Turvey & Fonseca, 2009; Stepp & Turvey, 2010).

Conclusion

Our findings imply that the corticospinal system, possiblytogether with other descending systems, participatesin a fundamental control process – threshold positionresetting, thus pre-determining the spatial frame ofreference in which the neuromuscular periphery isconstrained to work.

References

Aflalo TN & Graziano MS (2006). Partial tuning of motorcortex neurons to final posture in a free-moving paradigm.Proc Natl Acad Sci U S A 103, 2909–2914.

Archambault PS, Mihaltchev P, Levin MF & Feldman AG(2005). Basic elements of arm postural control analyzed byunloading. Exp Brain Res 164, 225–241.

Asatrian DG & Fel’dman AG (1965). [Functional tuning of thenervous system with control of movement or maintenance ofa steady posture – I. Mechanographic analysis of the work ofthe joint during the performance of a postural task].Biophysics 10, 925–935 (Engl transl of Biofizika 10, 837–846).

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1568 H. Raptis and others J Physiol 588.9

Battaglia D, Brunel N & Hansel D (2007). Temporaldecorrelation of collective oscillations in neural networkswith local inhibition and long-range excitation. Phys Rev Lett99, 238106.

Bonnard M, Camus M, de Graaf J & Pailhous J (2003). Directevidence for a binding between cognitive and motorfunctions in humans: a TMS study. J Cogn Neurosci 15,1207–1216.

Burtet L, Raptis H, Tunik E, Latash ML, Forget R & FeldmanAG (2007). Threshold control of wrist movements revealedby transcranial magnetic stimulation of the motor cortex(abstract). Congress of Journee scientifique du CRIR(Montreal).

Caminiti R, Johnson PB & Urbano A (1990). Making armmovements within different parts of space: dynamic aspectsin the primate motor cortex. J Neurosci 10, 2039–2058.

Capaday C (1995). The effects of baclofen on the stretch reflexparameters of the cat. Exp Brain Res 104, 287–296.

Cheney PD & Fetz EE (1984). Corticomotoneuronal cellscontribute to long-latency stretch reflexes in the rhesusmonkey. J Physiol 349, 249–272.

Clancy EA, Bertolina MV, Merletti R & Farina D (2008). Time-and frequency-domain monitoring of the myoelectric signalduring a long-duration, cyclic, force-varying, fatiguinghand-grip task. J Electromyogr Kinesiol 18, 789–797.

Crago PE, Houk JC & Hasan Z (1976). Regulatory actions ofhuman stretch reflex. J Neurophysiol 39, 925–935.

Di Lazzaro V, Restuccia D, Oliviero A, Profice P, Ferrara L,Insola A, Mazzone P, Tonali P & Rothwell JC (1998). Effectsof voluntary contraction on descending volleys evoked bytranscranial stimulation in conscious humans. J Physiol 508,625–633.

Dong D (1996). Associative decorrelation dynamics in visualcortex. In Lateral Interactions in the Cortex: Structure andFunction, ed. Sirosh J, Miikkulainen R & Choe Y. UTCSNeural Networks Research Group, Austin, TX.

Dubois DM (2001). Computing anticipatory systems. AIP ConfProceedings 573(XI), 706.

Evarts EV & Fromm C (1981). Transcortical reflexes and servocontrol of movement. Can J Physiol Pharmacol 59, 757–775.

Fedirchuk B & Dai Y (2004). Monoamines increase theexcitability of spinal neurones in the neonatal rat byhyperpolarizing the threshold for action potentialproduction. J Physiol 557, 355–361.

Feldman AG (1986). Once more on the equilibrium-pointhypothesis (lambda model) for motor control. J Mot Behav18, 17–54.

Feldman AG (2009). New insights into action-perceptioncoupling. Exp Brain Res 194, 39–58.

Feldman AG, Adamovich SV & Levin MF (1995). Therelationship between control, kinematic andelectromyographic variables in fast single-joint movementsin humans. Exp Brain Res 103, 440–450.

Feldman AG, Goussev V, Sangole A & Levin MF (2007).Threshold position control and the principle of minimalinteraction in motor actions. Prog Brain Res 165, 267–281.

Feldman AG & Levin MF (1995). The origin and use ofpositional frames of reference in motor control. Behav BrainSci 18, 723–806.

Feldman AG & Orlovsky GN (1972). The influence of differentdescending systems on the tonic stretch reflex in the cat. ExpNeurol 37, 481–494.

Foisy M & Feldman AG (2006). Threshold control of armposture and movement adaptation to load. Exp Brain Res175, 726–744.

Fortier PA, Smith AM & Kalaska JF (1993). Comparison ofcerebellar and motor cortex activity during reaching:directional tuning and response variability. J Neurophysiol69, 1136–1149.

Ganguly K, Secundo L, Ranade G, Orsborn A, Chang EF,Dimitrov DF, Wallis JD, Barbaro NM, Knight RT &Carmena JM (2009). Cortical representation of ipsilateralarm movements in monkey and man. J Neurosci 29,12948–12956.

Gerloff C, Cohen LG, Floeter MK, Chen R, Corwell B & HallettM (1998). Inhibitory influence of the ipsilateral motor cortexon responses to stimulation of the human cortex andpyramidal tract. J Physiol 510, 249–259.

Ghafouri M & Feldman AG (2001). The timing of controlsignals underlying fast point-to-point arm movements. ExpBrain Res 137, 411–423.

Ginanneschi F, Del Santo F, Dominici F, Gelli F, Mazzocchio R& Rossi A (2005). Changes in corticomotor excitability ofhand muscles in relation to static shoulder positions. ExpBrain Res 161, 374–382.

Ginanneschi F, Dominici F, Biasella A, Gelli F & Rossi A(2006). Changes in corticomotor excitability of forearmmuscles in relation to static shoulder positions. Brain Res1073–1074, 332–338.

Gonzalez RV, Buchanan TS & Delp SL (1997). How musclearchitecture and moment arms affect wrist flexion-extensionmoments. J Biomech 30, 705–712.

Graziano MS, Taylor CS, Moore T & Cooke DF (2002). Thecortical control of movement revisited. Neuron 36, 349–362.

Gribble PL, Ostry DJ, Sanguineti V & Laboissiere R (1998). Arecomplex control signals required for human arm movement?J Neurophysiol 79, 1409–1424.

Griffin DM, Hudson HM, Belhaj-Saif A & Cheney PD (2009).Stability of output effects from motor cortex to forelimbmuscles in primates. J Neurosci 29, 1915–1927.

Hallett M (2007). Transcranial magnetic stimulation: a primer.Neuron 55, 187–199.

Holdefer RN & Miller LE (2002). Primary motor corticalneurons encode functional muscle synergies. Exp Brain Res146, 233–243.

Hoshiyama M, Kakigi R, Koyama S, Takeshima Y, Watanabe S& Shimojo M (1997). Temporal changes of pyramidal tractactivities after decision of movement: a study usingtranscranial magnetic stimulation of the motor cortex inhumans. Electroencephalogr Clin Neurophysiol 105, 255–261.

Irlbacher K, Voss M, Meyer BU & Rothwell JC (2006).Influence of ipsilateral transcranial magnetic stimulation onthe triphasic EMG pattern accompanying fast ballisticmovements in humans. J Physiol 574, 917–928.

Jackson A, Mavoori J & Fetz EE (2007). Correlations betweenthe same motor cortex cells and arm muscles during atrained task, free behaviour, and natural sleep in themacaque monkey. J Neurophysiol 97, 360–374.

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

J Physiol 588.9 Corticospinal control of wrist position 1569

Jasper HH (1958). Report of the Committee on Methods ofClinical Examination in Electroencephalography.Electroenceph Clin Neurophysiol 10, 370–371.

Kalaska JF (2009). From intention to action: motor cortex andthe control of reaching movements. Adv Exp Med Biol 629,139–178.

Kalaska JF, Scott SH, Cisek P & Sergio LE (1997). Corticalcontrol of reaching movements. Curr Opin Neurobiol 7,849–859.

Keizer K & Kuypers HGJM (1984). Distribution ofcorticospinal neurons with collaterals to lower brain stemreticular formation in cat. Exp Brain Res 54, 107–120.

Keizer K & Kuypers HGJM (1989). Distribution ofcorticospinal neurons with collaterals to the lower brainstem reticular formation in monkey (Macaca fascicularis).Exp Brain Res 74, 311–318.

Knikou M (2008). The H-reflex as a probe: pathways andpitfalls. J Neurosci Methods 171, 1–12.

Komori T, Watson BV & Brown WF (1992). Influence ofperipheral afferents on cortical and spinal motoneuronexcitability. Muscle Nerve 15, 48–51.

Krawitz S, Fedirchuk B, Dai Y, Jordan LM & McCrea DA(2001). State-dependent hyperpolarization of voltagethreshold enhances motoneurone excitability during fictivelocomotion in the cat. J Physiol 532, 271–281.

Kurtzer I, Herter TM & Scott SH (2005). Random change incortical load representation suggests distinct control ofposture and movement. Nat Neurosci 8, 498–504.

Latash ML & Gottlieb GL (1991). An equilibrium-point modelfor fast, single-joint movement: II. Similarity of single-jointisometric and isotonic descending commands. J Mot Behav23, 179–191.

Lemon RN (2008). Descending pathways in motor control.Annu Rev Neurosci 31, 195–218.

Lemon RN, Johansson RS & Westling G (1995). Corticospinalcontrol during reach, grasp, and precision lift in man. JNeurosci 15, 6145–6156.

Lennerstrand G & Thoden U (1968). Position and velocitysensitivity of muscle spindles in the cat. 3. Static fusimotorsingle-fibre activation of primary and secondary endings.Acta Physiol Scand 74, 30–49.

Levin MF, Selles RW, Verheul MH & Meijer OG (2000).Deficits in the coordination of agonist and antagonistmuscles in stroke patients: implications for normal motorcontrol. Brain Res 853, 352–369.

Lewis GN, Byblow WD & Carson RG (2001). Phasicmodulation of corticomotor excitability during passivemovement of the upper limb: effects of movement frequencyand muscle specificity. Brain Res 900, 282–294.

Macefield VG, Rothwell JC & Day BL (1996). The contributionof transcortical pathways to long-latency stretch and tactilereflexes in human hand muscles. Exp Brain Res 108,147–154.

McKiernan BJ, Marcario JK, Karrer JH & Cheney PD (1998).Corticomotoneuronal postspike effects in shoulder, elbow,wrist, digit, and intrinsic hand muscles during a reach andprehension task. J Neurophysiol 80, 1961–1980.

MacKinnon CD & Rothwell JC (2000). Time-varying changesin corticospinal excitability accompanying the triphasicEMG pattern in humans. J Physiol 528, 633–645.

Marsden CD, Merton PA & Morton HB (1972). Servo action inhuman voluntary movement. Nature 238, 140–143.

Marsden CD, Merton PA & Morton HB (1976). Servo action inthe human thumb. J Physiol 257, 1–44.

Matthews PB (1959). A study of certain factors influencing thestretch reflex of the decerebrate cat. J Physiol 147,547–564.

Matthews PB (1986). Observations on the automaticcompensation of reflex gain on varying the pre-existing levelof motor discharge in man. J Physiol 374, 73–90.

Matthews PB (1991). The human stretch reflex and the motorcortex. Trends Neurosci 14, 87–91.

Mihaltchev P, Archambault PS, Feldman AG & Levin MF(2005). Control of double-joint arm posture in adults withunilateral brain damage. Exp Brain Res 163, 468–486.

Misiaszek JE (2003). The H-reflex as a tool in neurophysiology:its limitations and uses in understanding nervous systemfunction. Muscle Nerve 28, 144–160.

Musampa NK, Mathieu PA & Levin MF (2007). Relationshipbetween stretch reflex thresholds and voluntary arm muscleactivation in patients with spasticity. Exp Brain Res 181,579–593.

Nichols R & Ross KT (2009). The implications of force feedbackfor the lambda model. Adv Exp Med Biol 629, 663–679.

Nichols TR & Steeves JD (1986). Resetting of resultant stiffnessin ankle flexor and extensor muscles in the decerebrate cat.Exp Brain Res 62, 401–410.

Ostry DJ & Feldman AG (2003). A critical evaluation of theforce control hypothesis in motor control. Exp Brain Res 153,275–288.

Park MC, Belhaj-Saif A, Gordon M & Cheney PD (2001).Consistent features in the forelimb representation of primarymotor cortex in rhesus macaques. J Neurosci 21, 2784–2792.

Pilon JF, De Serres SJ & Feldman AG (2007). Thresholdposition control of arm movement with anticipatoryincrease in grip force. Exp Brain Res 181, 49–67.

Pilon JF & Feldman AG (2006). Threshold control of motoractions prevents destabilizing effects of proprioceptivedelays. Exp Brain Res 174, 229–239.

Porter R (1985). The corticomotoneuronal component of thepyramidal tract: corticomotoneuronal connections andfunctions in primates. Brain Res 357, 1–26.

Raptis HA, Burtet L, Forget R & Feldman AG (2008).Corticospinal excitability before and after active movementsand unloading reflexes: implications for theposture-movement problem. Society for Neuroscience,Washington, DC 860.10/JJ1.

Raptis H, Rothwell J, Forget R & Feldman AG (2009).Corticospinal excitability and EMG profiles during activemovements and muscle unloading: motor control by shiftingspatial frames of reference. Congress of Progress in MotorControl VII (Marseille), abstract C2–20.

Rothwell JC, Thompson PD, Day BL, Boyd S & Marsden CD(1991). Stimulation of the human motor cortex through thescalp. Exp Physiol 76, 159–200.

St-Onge N, Adamovich SV & Feldman AG (1997). Controlprocesses underlying elbow flexion movements may beindependent of kinematic and electromyographic patterns:experimental study and modelling. Neuroscience 79,295–316.

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society

1570 H. Raptis and others J Physiol 588.9

St-Onge N & Feldman AG (2004). Referent configuration ofthe body: a global factor in the control of multiple skeletalmuscles. Exp Brain Res 155, 291–300.

Schieber MH & Santello M (2004). Hand function: peripheraland central constraints on performance. J Appl Physiol 96,2293–2300.

Schneider C, Lavoie BA, Barbeau H & Capaday C (2004).Timing of cortical excitability changes during the reactiontime of movements superimposed on tonic motor activity.J Appl Physiol 97, 2220–2227.

Schuurmans J, de Vlugt E, Schouten AC, Meskers CG, de GrootJH & Van Der Helm FC (2009). The monosynaptic Ia afferentpathway can largely explain the stretch duration effect of thelong latency M2 response. Exp Brain Res 193, 491–500.

Shah A, Fagg AH & Barto AG (2004). Cortical involvement inthe recruitment of wrist muscles. J Neurophysiol 91,2445–2456.

Stepp N & Turvey MT (2010). On strong anticipation. CognSyst Res 11, 148–164.

Todd G, Petersen NT, Taylor JL & Gandevia SC (2003). Theeffect of a contralateral contraction on maximal voluntaryactivation and central fatigue in elbow flexor muscles. ExpBrain Res 150, 308–313.

Townsend BR, Paninski L & Lemon RN (2006). Linearencoding of muscle activity in primary motor cortex andcerebellum. J Neurophysiol 96, 2578–2592.

Tunik E, Poizner H, Levin MF, Adamovich SV, Messier J,Lamarre Y & Feldman AG (2003). Arm-trunk coordinationin the absence of proprioception. Exp Brain Res 153,343–355.

Turvey MT & Fonseca S (2009). Nature of motor control:perspectives and issues. Adv Exp Med Biol 629, 93–123.

Von Holst E & Mittelstaedt H (1950). Daz reafferezprincip.Wechselwirkungen zwischen Zentralnerven-system undPeripherie, Naturwiss 37, 467–476. English translation (1973:The reafference principle). In The behavioural physiology ofanimals and man. The collected papers of Erich von Holst.Martin R (translator), ed. University of Miami Press, pp.139–173. Coral Gables, Florida.

Wachholder K & Altenbruger H (1927). Do our limbs have onlyone rest length? A contribution to the measurement of elasticforces in passive and active movements. English translationand comments by Sternad D in Motor Control, 2002, 6,299–318. ed. Pfluger’s Archive fur die gesamte Physiologie.

Wassermann EM, McShane LM, Hallett M & Cohen LG (1992).Noninvasive mapping of muscle representations in humanmotor cortex. Electroencephalogr Clin Neurophysiol 85, 1–8.

Wiesendanger M (1986). Experimental evidence for theexistence of a proprioceptive transcortical loop. Prog BrainRes 64, 67–74.

Zatsiorsky VM, Li ZM & Latash ML (2000). Enslaving effects inmulti-finger force production. Exp Brain Res 131, 187–195.

Ziemann U, Ishii K, Borgheresi A, Yaseen Z, Battaglia F, HallettM, Cincotta M & Wassermann EM (1999). Dissociation ofthe pathways mediating ipsilateral and contralateralmotor-evoked potentials in human hand and arm muscles.J Physiol 518, 895–906.

Author contributions

All the experiments were performed in the Motor ControlLaboratory at the Institute of Rehabilitation of Montreal(University of Montreal). All authors contributed to thefollowing parts: (1) Conception and design of the experiments;(2) Collection, analysis and interpretation of data; (3) Draftingthe article or revising it critically for important intellectualcontent. All authors approved the final submitted version.

Acknowledgements

We thank Trevor Drew and Mindy Levin for valuable commentson a draft of this paper and Michel Goyette and ValeriGoussev for help in programming and data analysis in thisstudy. Supported by Canadian Institutes of Health Research,Collaborative health research projects program (CHRP), NaturalSciences and Engineering Research Council, Fonds de recherchesur la nature et les technologies and Fonds de la recherche ensante du Quebec (Canada).

C© 2010 The Authors. Journal compilation C© 2010 The Physiological Society