Pulsatile motor output in human finger movements is not dependent on the stretch reflex

Journal of Physiology (1996), 493.3, pp.895-908

Pulsatile motor output in human finger movements is notdependent on the stretch reflex

J. Wessberg and A. B. Vallbo

Department of Physiology, Goteborg University, Medicinaregatan 11, 8-41390 Goteborg,Sweden

1. Stretch perturbations were delivered during slow voluntary finger movements with the aimof exploring the role of the stretch reflex in generating the 8-10 Hz discontinuities thatcharacterize these movements. Afferent activity from muscle spindle primary endings inthe finger extensor muscles was recorded from the radial nerve, along with the EMGactivity of these muscles, and kinematics of the relevant metacarpo-phalangeal joint.

2. Perturbations elicited a distinct response from the muscle spindles appearing at therecording electrode after 13 ms, and weak reflex responses from the muscle with peakvalues at 53 and 63 ms during flexion and extension, respectively.

3. The time relations between kinematics, spindle firing and modulations of EMG activityelicited by the perturbations were compared with those of the self-generated discontinuities.These analyses indicate that stretch reflex mechanisms cannot account for the modulationsofEMG activity that give rise to successive 8-10 Hz discontinuities.

4. A comparison of the reflex responses to perturbations with the EMG modulations duringself-generated movements indicates that the reflex was too weak to account for the pulsatilemotor output during voluntary movements.

5. By inference it was concluded that the 8-10 Hz discontinuities during self-generatedmovements are probably generated by mechanisms within the central nervous system.

It has been demonstrated that voluntary finger movementsin man are not smooth but discontinuous, and usuallycontain a prominent component of 8-10 Hz variations inangular velocity. These discontinuities are brought about bymodulations of muscular activity, often an alternatingincrease of the agonist and antagonist activity (Vallbo &Wessberg, 1993). However, the underlying neuronalmechanisms for this pulsatile motor output remainunknown.

Oscillations in feedback loops from sense organs respondingto finger movements seem to be a potential mechanism.Stretch reflexes have attracted particular interest and it hasbeen suggested that the loop time of the spinal reflex isroughly adequate for sustaining 8-10 Hz finger oscillations,whereas the long-latency stretch reflex component woulddampen rather than promote such oscillations (Stein &Oguzt6reli, 1976).

The spinal stretch reflex has been suggested to account forphysiological tremor as well as enhanced physiologicaltremor during position holding (Lippold, 1970; Joyce &Rack, 1974; Hagbarth & Young, 1979; Young & Hagbarth,1980; Burne, Lippold & Pryor, 1984; Sanes, 1985), althoughthe opposite view also has ample support, i.e. that the

spinal reflex is unlikely to produce 8-12 Hz tremor(Andrews, Burke & Lance, 1973; Elble & Randall, 1978;Prochazka & Trend, 1988; Jacks, Prochazka & Trend,1988). In addition, a number of other mechanisms fortremor have been proposed (Marsden, 1984; Elble & Koller,1990).

The aim of the present study was to test the hypothesisthat the 8-10 Hz discontinuities during slow fingermovements are caused exclusively by reflex oscillationstriggered and sustained by mechanoreceptors in themoving parts, particularly by muscle spindles in theagonist and antagonist muscles. The approach was to studythe strength of the reflex responses to brisk fingerperturbations during voluntary movements and toinvestigate the time relations between kinematics, muscleactivity and firing of muscle spindle primary endings usingtime- and frequency-domain analyses.

It was found that reflex responses to finger movements aretoo delayed and too weak to account for the modulations ofmuscular activity at 8-10 Hz during voluntary fingermovements.

Preliminary reports of some of this work have beenpublished previously (Vallbo & Wessberg, 1995, 1996).

8953978

) by guest on July 15, 2011jp.physoc.orgDownloaded from J Physiol (

http://jp.physoc.org/

896

METHODSThe expeiiments were pierformed on fourteen healthy volunteers,seven females and seven males, aged 19-31 years. Informedconsent was obtained according to the I)eclaration of Helsinki.The study was approved by the Ethical Committee of the AMedicalFaculty of G6teborg University.General procedure and equipmentThe subject was comfortably seated in a reclining chair with theleft armln resting on a su)pporting platformii. The hand was held witha clanm) that permitted fr ee finger movements, but preventedmovenments of the wrist, which was held lightly flexed at a jointangle of 165-175 deg. Mtuch care was taken to position the hand ina mlaniner the subject felt was natural for fine manipulative handmov ements.

One finger, usually the middle, was strapped to a splint, whichpermitted movements only at the metacarpo-phalangeal joint. Thesplint was attached to an actuator device byimeans of a low-masshinged bar. The actuator has been described previously (Al-Falahe& V'allbo, 1988). Transducers of the actuator provided continuousrecording of metacarpo-phalangeal joint angle, velocity andtorque. An integral servX-omotor compensatecl for firiction andiner tia; thus the actuator device presented zeiro load to the miovingfinter.

A nmicrocomputer w,as used to control the actuator, the trackingoscilloscope, and other equipment including a separate samplingcomnputer. In six experimnenits, the computer was programmed togenier ate a perturbation, i.e. a single transient finger flexionmoVement of 3 dceg during the voluntary movement. The imposedmovemient lasted 30 ins, while the actuator was in a positionfeedback mode, and the finger w,as unable to move in any otherway.

An oscilloscope in front of the subject was used for visual tracking.One beam was swept xertically and split into t-wo halves, the tophalf displaying the desirecl target and the bottom half the actualjoint angle. The width of the beam corresponded to a joint angle of0-25 deg. Direct vision of the finger was obscured with a screen.

Electromyography (E1IG) wras recorded with surface electrodesplaced on the dorsal surface of the forearrm, over the area ofminimum thlresholds for transcutaneous electrical stimulation ofthe commnon finger extensor. Electrode iml)edlance was measured,anid only values below 2 5 klQ per electirode were accepted. TheEl IG was root mean square rectified with rise and decay timeconstants of 1-6 and 4 8 ills, respectively, ancl then sampled at800 Hz on-line. The kinemnatic parameters were sampled at400 Hz. Acceleration was derived off-line by digital low-passfilter ing (-3 dB at 52 Hz, zero gain at 123 Hz) and differentiationof' tle velocity signals.

Microneurographic recording and unit classificationIn a subset of three experiments involving perturbations, theactivity of single muscle spinclle afferents wvas recorded from theleft radial nerve using the microneurographic technique (VTallbo,Hagbarth, Torebj6rk & Wallin, 1979). The recording needle was

inserted 5-7 cm proxim-al to the elbow. The search procedlureconsistecl of repeated l)assive flexion of the fingers, and localizedpall)ation over the longr extensor muscles of the fingers. Onlyslowvly adlapting units that respondecl to palpation and passiveflex ion (muscle stretch) and/or voluntary conti'action of the )arentmuiiiscle were rIecor clecl. In the microneurography experiments, thefitger that elicitedl the best response f'rom) the unit was strapped tothte actuator device, as described above. Th'le nerve signal wvas

J. Physiol.493.3

sampled at 12-8 kHz. Each recorded nerve spike was inspected off-line on an expanded time scale, and this validated nerve signalas used for subsequent unit classification and data analysis.

The recorded afferents were classified on the basis of eight criteriausing a Bayesian evaluation procedure as described by Edin &Vallbo (1 990), i.e. each unit was subjected to amp stretches, smallsinusoidal oscillations superimposed on a ramp stretch, a test ofintrafusal myofilament bonds, voluntary contractions ending withbrisk relaxations of the par ent rmuscle, and maximal twitchcontractions evoked by electr ical stimulation. Six units classified as

Ia afferents were includledl in the study. Median probability that

these units were muscle spindle primary afferents according to theBayes procedure was 0 98.

Experimental protocolThe main experimental protocol comprised a visual tracking task,each trial consisting of three phases: a position-holding phase, an

extension or flexion movenment at the metacaipo-phalangeal jointof 20 deg, and a final position-holding phase. In the micro-neurography experiments, movements were concatenated so thatthe subject performed first extension and then flexion in one

uninterrupted sequence. Auditory cues prompted the subjects togo before the individual movement. Thie target moved with a

constant speed corresponding to 10, 16 or 25 deg s-i, starting 0 3 s

after the auditory cue. Subjects were given a short training sessionat the beginning of each experiment. Usually ten to twentymovement sequences were sufficient to make the subjectsadequately familiar Mwith the procedure.

In the perturbation experiments (ot = 6), a transient flexionmovement, stretching the finger extensorimuseles, was imposed ata random point in timiie dLmr'ing a voluntar y finger' movement.Altogether, 342 pertur'l)ation tests were collected, including80 tests with recordings fiom I a units. In addition, 800 voluntarymovements without perturb)ations were collected for the analysisof the EMIG patterns dcuring self-generated mov-ements.

Data analysisThe tinme of occurrence and the size of the stretch reflex were

investigated from the EMG records by averaging a large number of

perturbation test data. Statistical analysis was based on the 95%confidence limits of the EMG during the 100 ms preceding thepeirturbation.

The nerve response to per'turbations was used to computecompound discharge timne histograms. First the number of spikesin 5 nms bins was countedt separately f'or each unit ancl for flexionand extension movements. Seconcd, a Gaussian distributionsuperimposed on a constant, baseline was fitted to the lhistolramnsby miieans of a general least-squares algorithlm. AMean p)eakdischarge latency ancl liselharge variability are giiven as the iiean

and standacrd deviation of the fitted Gaussian distributions.

E1AIG modulations duiring the self-genier ated movements were

investigated using several methods for time- and freqiuency-domain signal analysis. Averages of EMIG in relation to the8-10 Hz discontinuities were produced by manual selection ofcycles and triggering on the local acceleration peak. For runs with

perturbations, an appropriate time window around the disturbance

was excluded fi'omn the averaging procedure.

C:ross-coiielation between EM11G and acceleration was calculatedusing a stanclard procedure (Bendat & Piersol, 1986). The cross-

correlatlion is a statistical(l esipticn of how the two signals; are

linearly correlated over a ranige of (liftlerent lags, i.e. when one

signal is shifted in time wvith rlespect to the other.

J Wessberg and A. B. Valibo



Pulsatile motor output

For frequency-domain analysis, auto- and cross-spectra wereobtained by averaging the Fourier transforms of successivesegments of the ENMG and acceleration time series (Bendat &1'iersol, 1986). For most analyses, a 1 28 s time window was used,giving a frequency resolution of 0-78 Hz. A so-called Hanningtapering function and 50% segmenit overlap w-as used. For thes)ectral analysis of the applied pertuirbations, an untapered 0X64 s

timiie window was used, giving 1-56 Hz resolution. Cross-spectrawere normalized as the coherence spectrum, which is a functiondescribing the linear correlation squared between the two signalsat a given firequency on a scale firom zero to one.

From the cross-spectra, phase spectra wN-ere also calculated, andwere used to estimate accurately mean delay betwreen the tw!oprocesses. Since a fixed delay constitutes a decreasing fraction ofthe total cycle time as frequency increases, the delay should giverise to a tilted line with constant slope in the phase spectrum.Delay can be estimated by finding the slope of this line usingstandard linear regression. Care must be taken to limiit theanalysis of phase data to frequencies that exhibit a coherencesignificantly greater than zero.

Statistical inference based on cross-correlation strictly requir esp)arametric time-series moclelling (Brockwell & Davies, 1991). Thiscannot be made wN-ithout some ultimately subjective assumiiptionsregairding the model paramiieters, ancl wTas not done in the presentstudy. Statistical analysis of the complementary frequency-domiaintools, particularly coherence, is, however, straightforward.

SemanticsThe sign of velocity and acceleration was related to lengtheningand shortening of the parent muscle, i.e. a positive velocitysignifies a flexion movement, which implies a lengthening of themuscle. Similarly a positive acceleration signifies an acceleration ofa movement in the direction of muscle lengthening. Positive-goingand negative-going peaks in the acceleration records of voluntarymovements must be interpreted in relation to the direction ofthese movements. A positive-going peak in the acceleration curveduring an on-going flexion movement implies an increase of thespeed of movement, whereas it implies a clecrease of the speed ofthe on-going extension movemiient. SimilarlY, a negative-goingpeak in the acceleration cui ve may signify either a decrease in thesl)eedl of the flexion movement or an inciease of the speecl of theextension movement.

RESULTSThe main purpose of the present study was to investigatethe role of reflex effects from the moving parts, andparticularly the stretch reflexes, as a potential generator ofthe 8-10 Hz discontinuities that characterize slow fingermovements. The approach was to inject perturbationsduring on-going movements, and to explore the size andthe time relations of the evoked reflex responses. Theseresponses were compared with the EMG modulationsduring unperturbed 8-10 Hz discontinuities. In addition,the covariation between EMG and acceleration duringunperturbed trials Nas investigatedi using statistical signalanalyses in time and frequency domiains.

The basic reasoning was that the reflex response to aperturbation would be large if the stretch reflex, alone orin combination wvith other reflexes, was powerful enough

to produce the discontinuities during self-generatedmovements. Moreover, the time relations betweenkinematics, muscle spindle firing, and EMIG activity wouldmatch in perturbations andl self-generated discontinuities ifthe latter were accounted for by the stretch reflex alone.

Stretch reflexes during movementsIn order to explore the properties of the stretch reflex,perturbations were applied at random during the voluntarymovements. The perturbations wxrere rapid flexionsamounting to 3 deg, which stretched the test muscles,i.e. the finger extensor muscles. Figure 1A shows recordswith perturbations during voluntary flexion moveements,i.e. the parent extensor muscle was lengthening. The threetop records of position, velocity, and EMG originate fromone sample movement, while the following EMG recordsstem from similar perturbed movements. All records havebeen aligned at the onset of perturbations. Compared withthe self-generated discontinuities, the amplitudes of theperturbations were larger by a factor of 2-10, measured asthe incremental change in p)osition during individualdiscontinuities.

Usually, the reflex effects of perturbations were not largeenough to be identified in the individual EMG records. Todefine the reflex output effect, averages were constructedusing the onset of perturbation as; the trigger. Figuie 1Billustrates one average of fifty iiovemnents from the samesubject as in Fig. 1A. It is apparent that the EMG responseto the perturbation is small in relation to the self-generatedIEMG bursts during the individual movements. This was aconsistent finding for reflexes elicited during flexionmovements.

Figure 2 shows a comparable set of data for extensionmovements, i.e. wThen the extensor muscle was activelyshortening. Because the perturbation goes in the oppositedirection to the voluntary imiovement, it constitutes a morepowerful disturbance than witlh flexion movements. Theaverage reflex effect in the EMIG, shown in Fig. 2B, is largerin magnitude than during flexion (Fig. 1B). This miilhtpartly be an effect of the larger meclhaniical (lisruption ofthe movements, and partly an effect of stronger centralfacilitation of the motor neurone pool as the extensormuscle is driving the movement.

In order to illustrate the details of the reflex effects shownin Figs 1 and 2, the recordings are also presented onexpanded timne scales in Figs 3A and 4B. During flexionnlovem,elt (Fig. 3A), a significant increase of the EMG wNasseen 34 ms after peak acceleration (mean; range 23-45 ms),while the peak occurried at 53 nis (range 42-64 ins) (fivesubjects, averages of fiftv perturbations or more firomn eachsubject). The initial EAIG burst was followed by a decrease,and a succession of waves that seeiimed to be relate(d to thelater phases of the comiiplex kinemiiatic structure of theaverage perturbation (see ailso Fig. I B). The firequency of'thiese waves was in the 14-20 Hz ranige. Hence it is obvious

J. Physiol.493.3 897



J Wessberg and A. B. Vallbo

that they do not represent a resetting of the 8-10 Hzoscillations, but are of a different nature. It is possible thatthey represent underdamped passive oscillations of themechanical system, i.e. the finger and the actuator, in thewake of the strong perturbation. The kinematics of theperturbations and the relationship between accelerationand EMG were also analysed using a statistical signalanalysis approach (see below).

Figure 3A also shows muscle spindle data based onrecordings from six muscle spindle primary afferentsduring eighty-eight stretch perturbations. A distinctafferent activation with muscle stretch was seen in allafferents. Five of them responded consistently with morethan one impulse at high firing rates. It may be seen thatthe increase in I a firing started very shortly after the onsetof the perturbation. On the basis of Gaussian distributions

AJoint ]angle

(1 0 deg)

Velocity 1(100 deg s-I) o

EMG 1(100 1UV)

fitted to the individual discharge histograms it was deducedthat the peak discharge occurred 13 ms after peakacceleration. This latency was the same for both directionsof movements. Standard deviations were 5-7 and 6-4 ms forflexion and extension, respectively. This information isuseful because it allows the accurate calculation of adistinct part of the total stretch reflex loop time, i.e. theaverage latency from the afferent volley passing therecording site to the EMG response simply by subtracting13 ms from the latency between peak acceleration and theEMG response as given above.

It seemed intuitively appealing to compare directly theaveraged EMG response to perturbations with the averagedEMG of the self-generated 8-10 Hz discontinuities, toexplore whether amplitudes and timings fit the hypothesisthat the latter are produced by movement-induced reflexes.

-1L -.LS LIL, LLAII LMBLsJLAwL4*LL LL L&L

BJointangle

(1 0 deg) IVelocity 1

(100 deg s') J

EMG 1(100 UV) I

M100o ms

Figure 1. Perturbations during flexion movementsA, 3 deg stretch of the finger extensor muscle applied during voluntary flexion movements. Records fromtop: joint angle, joint velocity and finger extensor EMG for a single sample stretch, and the EMG recordsfor another four samples. In these movements, the subject was tracking a visual target that moved at25 deg s-' (cf. Methods). B, average of fifty stretches as in A.

I L, t iLl L,Lid A -II.

.......................................................

-Al

898 J Physiol. 493.3

A




In order to define the characteristics of the EMG associatedwith the self-generated discontinuities, the average EMGwas extracted from a set of 8-10 Hz cycles selected forprominent local accelerations, using acceleration peaks astriggers. As shown in Fig. 3C and in a previous report(Wessberg & Vallbo, 1995), the peak I a firing occurs veryclose to peak acceleration of the 8-10 Hz discontinuities, atleast during muscle lengthening. Therefore it seemed thatthis trigger point would be optimal to reveal any reflexeffects due to angular speed variations at 8-10 Hz. Sampleaverages from flexion movements are shown in Fig. 3B. Toenable a direct comparison between the effect of theinjected perturbation and the naturally occurring pattern,the scales are identical in Fig. 3A and B. It is obvious thatthe EMG modulation associated with the relatively in-conspicuous self-generated discontinuity is larger than themodulation elicited by the much larger perturbation.

A

Considering the time relations within the individualdiscontinuity during voluntary flexion movement, it isobvious that peak positive acceleration occurs while theflexor muscle is momentarily propelling the finger (Vallbo &Wessberg, 1993). During this phase, when the extensormuscle is passively stretched, its spindles would launch thestimulus for a potential stretch reflex that would be seen inthe EMG, appropriately timed in relation to the peakacceleration. However, it has previously been found that theextensor muscle will produce a braking pulse very shortlyafter the peak (Vallbo & Wessberg, 1993). The exact timerelations are shown in the sample average of Fig. 3B, whereit may be seen that the extensor muscle EMG started toincrease roughly simultaneously with the peak acceleration,and peaked 12 ms later. Hence the timing of the extensorEMG during the self-generated discontinuities is hardlyconsistent with a reflex nature.

Jointangle

(1 0 deg)

Velocity 1(100 deg s1)]

EMG 1(200 juV) j i..I.iL LL it..iLL ~ A- Alm1Mi aJ

Li-L,E , iL.

kAiIA- 1-1- A .-.L,L i ~JkaL l.'iI A.Ll "AL

B

A

Figure 2. Perturbations during extension movementsA, 3 deg stretch of the finger extensor muscle applied during voluntary extension movements. B, averageof fifty stretches as in A. Tests and records as in Fig. 1. Data from the same subject as in Fig. 1.

Jointangle

(1 0 deg) IVelocity

(100 deg s_1) J

EMG200 tV

M100 ms

J Physiol. 493.3 899



900 J Wessberg a

Obviously, the time relations between the kinematic eventsthat might possibly elicit a reflex response and increase ofthe EAIG activity is vastly different in the twAo situations(Fig. 3A and B), because in the self-generated movementsthe increase of EMGi follows much earlier after theacceleration than in the reflex response to the perturbations.Similar records as in Fig. 3 were obtained in all subjectsandlNwere corroborated bY cross-correlation analysis, asdescribed below.

In order to allow a direct comparison betwreen perturbationsand self-generated movements with regard to musclespincdle firing, material from a previous study is displayedin Fig. 3C (Wessberg & Vallbo, 1995), i.e. the activity oftwenty-two single Ia afferents in relation to 8-10 Hzmovement discontinuities.

The average of kinematics and EMG response toperturbations from Fig. 2B for extensioni moveirments are

Ind

A

Jointangle(5 deg)

A. B. Vallbo J Physiol.493.3

shown on an expanded time scale in Fig. 4A. The delayfrom peak acceleration to reflex onset was 36 ms (n = 5,range 26-49 ms), and to peak EMG it was 63 ms (range53-75 ms). The reflex in the finger extensor muscle wasthus slightly more delayed during muscle shortening thanduring lengthening movements.

Figure 4B shows the average of a number of 8-10 Hzdiscontinuities during e,xten8sion movemenlts. It may be seenthat the average did not rieveal any increase whatsoever inthe EAIG that was time-locked to the potential stimulus fora reflex, i.e. peak acceleration when the muscle spindleafferents wNould fire maximally (Wessberg & Vallbo, 1995).

Smoothing effect of cycle variability. It might appearsurprising, at first sight, that the modulations of the EMG thatapparently generated the discontinuities shown in the top recordof Fig. 4B did not appear in the average, particularly since theextensor muscle is the prime mover in these movements. However,the lack of EMIG modulation can be understood if the

B

IVelocity 1

(100 deg s-1) j

Acceleration 1(5000 deg s-2) j

EMG ](50 ,uV)

V A X...........................................I.................I...................

Ia dischargehistogram

(1 spike bin-') IC

loo ms loo ms

Figure 3. Averages of perturbations and self-generated discontinuities during flexion movementA, average of perturbation tests applied during finger flexion on expanded scales (cf. Fig. 1). Records fromtop: joint angle, joint velocity, joint acceleration, extensor muscle EMG and compound I a dischargehistogramns. 'rhe 95% confidence limits of the EMG baseline, measured 100 ms prior to the stretch, areindicated by continuous lines. B, averages triggered on local positive (i.e. muscle lengthening) accelerationpeaks of fifty typical 8-10 Hz movement (liscontinuities selected from the samne inovements as in A.Records in B represent the samne variables as in A and have the same scaling. C, activity pattern oftwenty'-two single I a miiuscle spindle afferents in relationi to 8-10 Hz inovemnent cliscontinuities repro(lucedfiom a previous study (WVessberg & Vallbo, 1995). An albsolute scale as for the I a histogram in A is, notav-ailable for these data. The time bars have been aligned with the positive peaks in the accelerationrecords.

_ _ _ _ _ _ _0




characteristics of the 8-10 Hz discontinuities and the smoothingeffect of the averaging proceduire are taken into account. Becausethe 8-10 Hz discontinuities exhibit considerable cycle-to-cyclev-ariability with ilegard to period length, the increases in EMOG willoccuIr at different points in timne with respect to the peak of theupward-going, i.e. braking, acceleration. However, by triggeringon the negative-going acceleration peak, the average EMG burstpropelling the finger could be demonstrated, as shown in Fig. 4C.The increase in EMG peaks some 20 ins before the negative-goingacceleration peak. Note that the scale of the acceleration record inFig. 4A is five times smaller than in B and C, while the scales forthe EMG records are the same.

Phase dependence of reflex response. It may be asked if the sizeof the reflex was dependent on the phase of the self-generateddiscontinuity at which the perturbation was injected. As the fingerextensor muscle is modulated at 8-10 Hz during both extensionand flexion, the compliance of the muscle as vell as the centralexcitatory drive might vary with the phase. In order to explorethe significance of phase-dependent factors, the EAIG response tothe perturbation (integrated over 25-75 ins aftei peak stietchvelocity) was plotted against acceleration of the self-generateddiscontinuity at onset of the perturbation (not illustrated). Thereflex size generally exhibited a considerable variability, but no

A

dependence of reflex size on phase of the discontinuities could bediscerned in any subject. Linear regressions were altogether non-significant.

Covariation between EMG and accelerationAnother approach to the exploration of the significance ofthe reflex mechanisms for generating the 8-10 Hzdiscontinuities was to employ standard statistical signalanalysis tools. This analysis complements the abovecomparison between averages of injected perturbations andaverages of naturally occurring discontinuities. Animportant difference between the two approaches is thatonly selected sections of the continuous data strings areused for averaging, whereas the full lengths of all relevantrecordings were used in the signal analysis. Thereforeinteresting features may appear in the latter that are notrevealed by the former.

Time-domain analyses. The temporal relationshipbetween EMG and acceleration during the voluntarymovements was explored using the EMGI to accelerationcross-correlation functions, in order to reveal reflex effects

B

Jointangle(5 deg)

Velocity ](1 00 deg s'1)

Acceleration 1(5000 deg s-2)j

EMG(50 ,uV) ]

I a dischargehistogram

(1 spike bin-') I

1 000 deg s-2

-LLE=-i ^I I I m

1 00 ms

50,uV ] ~...............................................................................

1000 deg s-2 j_ _>

50,uV ]

100 ms

Figure 4. Averages of perturbations and self-generated discontinuities during extensionmovementsA, average of perturbation tests applied during finger extension on expandecl scales (cf. Fig. 2). Recorlds asin Fig. 3. B, averages triggered on local positive (i.e. muscle lengthening) acceleration peaks of fifty typical8-10 Hz movement discontinuities from the same movements as in A. Records fiomn top: acceleration anclEAIG. C, averages triggered on fifty local negative (i.e. muscle shortening) acceleratiorn peaks dulLinlg tflesame 8-10 Hz movement discontinuities as in B; records as in B. Note that accelerationi signials aremagnified by a factor of 5 in B and C compared with A.

J Phystol. 493.3 901

fll-




and define the electrokinematic delay, i.e. the delay fromthe excitation of the muscle, as seen by the surfaceelectrodes, to the ensuing change in speed of jointmovement. Examples from one subject are shown inFig. 5A, while Fig. 5B shows the cross-correlation functionsfrom all twelve subjects superimposed.

The most prominent feature of Fig. 5A and B is a largenegative-going peak at relatively uniform lags (mean19 ms, range 12-32 ms, standard deviation 6-0 ms,n = 12) with similar amplitude and time course in bothflexion and extension movements. The positive lagindicates that a change in EMG preceded a change inacceleration, while the fact that the peak is negative impliesthat increases of the EMG of the finger extensor musclescovaried with accelerations in the direction of muscleshortening, i.e. the two variables went in oppositedirections at this lag. It seems apparent that the lag fromzero to the negative peak represents the electrokinematicdelay. Similar delays were found in the averaged records ofFigs 3B and 4B. Factors accounting for the variationbetween subjects were not explored, although it seemsreasonable to assume that the location of the EMGelectrodes and the inertia of the finger were significant, inaddition to the random error inherent in estimating timedelays with cross-correlation.

For a central issue of the present study, i.e. the role of thespinal stretch reflex in generating the 8-10 Hzdiscontinuities during voluntary movements, it wasparticularly pertinent to explore whether there was a

A0-2 -

0-

-0 2-

B

positive relation between EMG and acceleration at negativelags matching the loop time of the spinal stretch reflex. Ifthe spinal stretch reflex was working effectively, a stretchof the muscle would be followed by an increased EMG after30-50 ms. This would give rise to a positive peak in thecross-correlation functions at a negative lag correspondingto the reflex delay. Admittedly, our data demonstratedpositive peaks at negative lags, but these peaks were smalland the lag varied widely between subjects (-15 to120 ms). Hence it seems justified to conclude that the lagwas inconsistent with a reflex response in many subjects(Fig. 5).Moreover, the existence of a positive peak at a negative lagof about 50 ms is a very ambiguous support for a reflexeffect because, regardless of which mechanisms account forthe 8-10 Hz modulations, a pattern of peaks and troughswith a period time of 100-125 ms is bound to appear in thecorrelogram. This may be illustrated by the followingreasoning. A corollary of the self-generated periodicity isthat the main increase in EMG, which is one of the twofactors accounting for the negative-going peak in the cross-correlogram, will be followed by a relative decrease inmuscle activity which, in turn, will cause an acceleration inthe direction of muscle lengthening. In the correlogram thiswould appear as a positive-going peak at positive lags. Thiswas also seen in all cross-correlation records (mean 62 ms,standard deviation 8-2 ms).

Furthermore, due to the cyclic nature of the discontinuities,a relative muscle relaxation will also occur prior to an EMG

ExtensionFlexion0-2 -

r 0-

V-100 0

Lag (ms)

0-2 -

-0 2 -

100 -100 0Lag (ms)

100

-100 0Lag (ms)

-100 0Lag (ms)

Figure 5. EMG to acceleration cross-correlationsA, EMG to acceleration cross-correlation function during unperturbed finger flexion movements (Flexion)and extension movements (Extension) for one subject. B, superimposed cross-correlations for flexion andextension movements in twelve different subjects. Positive lags represent EMG preceding the acceleration,while at negative lags acceleration is leading EMG.

100

0

902 J Physiol.493.3

-022-

100




burst. This would give a positive-going peak at negativelags. Hence the peaks at about -50 ms in Fig. 5 areexpected simply from the periodic EMG modulations andany effect of a reflex would inevitably merge with a peak atcorresponding lags due to the periodicity of the processes.Hence the two positive peaks at negative and positive lagsin the correlograms of Fig. 5 might well be caused by thesame mechanism.

As indicated above (see Fig. 5), the peaks at negative lagswhose timing could possibly be consistent with a reflexresponse were rather flat and variable in time comparedwith both the main negative-going peak and the positive-going peak at positive lags. Considering that the latter wasvery likely to be due to trivial and variable modulations ofEMG activity inherent in the 8-10 Hz pattern, it seemsobvious that the reflex hypothesis would predict a muchlarger and more distinct correlation peak at negative lags,because the EMG at this lag would produce the drivingforce for the 8-10 Hz discontinuities.

Frequency-domain analysis of acceleration and EMG.The above findings were corroborated by analysis in thefrequency domain. Figure 6A shows a sample coherencespectrum for EMG and acceleration. There was generally abroad-band significant coherence for both extension andflexion movements. However, the peak was regularly in therange 8-10 Hz and was very prominent. This indicatedthat the coupling between EMG and acceleration, revealedby cross-correlation analysis (Fig. 5), was particularlystrong at this frequency range.

The frequency-domain analysis also provides an additionalreliable way of assessing the delay between the two signals.Figure 6B shows the phase spectrum of the same samplerecords as in Fig. 6A, demonstrating the phase shiftbetween EMG and acceleration as a function of frequency.As described in Methods, the roughly linear slope in thephase spectrum indicated a constant delay over a frequency

A

a)c

Ca)

0

0-5-

0-4-

0-3-

0-2

0-1

0

range. The result of linear regression over part of the phasespectrum is shown by the straight line. In this sample case,the slope signifies a delay of 16 ms between EMG andacceleration.

Signal analysis of the perturbation experimentsThe stretch perturbation tests described above were alsoanalysed using signal analysis tools in the time andfrequency domains in order to compare perturbationresponse with self-generated discontinuities. Figure 7Ashows the results for the analysis of acceleration and EMGover the 640 ms immediately preceding the onset of thestretch perturbations for all extension movement trials inthe same subject as in Figs 1-4. The power spectrum of theacceleration demonstrates a 8-10 Hz peak. The EMGpower spectrum is broader, but a peak is discernible around8-10 Hz. The cross-correlation and phase spectrum aresimilar to those illustrated in the previous sections, withthe time delay from EMG to acceleration of 23 ms ascalculated from the phase spectrum.

Figure 7B shows the analogous analyses for the 640 msimmediately following the onset of the perturbations forthe same set of movements. Compared with Fig. 7A, the8-10 Hz peak in the acceleration power spectrum has beenreplaced by a broad peak in the 10-20 Hz range, with itsmaximum at 15-16 Hz. This corresponds to the frequenciesof the complex kinematic structure of the perturbations, asdescribed above (see also Figs 1B, 2B, 3A and 4A). In thepower spectrum for the EMG, there is also a peak in thesame range (13-19 Hz), but, in addition, another peak inthe 8-10 Hz range. These two peaks in the EMG powerspectrum would indicate a modulation of the EMG by theapplied perturbations, and, moreover, a persisting 8-10 Hzmodulation of the EMG during the perturbation period.The cross-correlation reveals a wave-like structure of about15 Hz, whose maximum is clearly to the right of lag zero,i.e. demonstrating a delay in time from acceleration to

B180 -

0)

0~

W-o-

-180I.I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

0 10 20 30 40 50

Frequency (Hz)0 10 20 30 40

Frequency (Hz)

50

Figure 6. Frequency-domain analysis ofEMG versus acceleration in unperturbed movementsA, coherence spectrum between EMG and acceleration for one sample subject. The horizontal line indicatesthe approximate 99% confidence limit. B, phase spectrum for the same data base as in A. Linearregression over part of the spectrum yields an estimated delay between EMG and acceleration of 16 ms,indicated by a thin line superimposed on the phase data. The positive slope of the line indicates that theEMG is leading the acceleration.

J. Physiol.493.3 903

A. -r



J Wessberg and A. B. Vallbo J Physiol.493.3

a)

C)CO

(5LU

0 10 20 30Frequency (Hz)

15

C0(oa1)L-o

0 10 20 30Frequency (Hz)

-200 0 200Lag (ms)

a)C/),aCL

-10 10 20 30Frequency (Hz)

6 0 _0.4 180C

ca

Ca~~~~~~~oc

~ 0 00

co~~2L 10-0-2-0 )a)----- -0-4 I-180

0 10 20 30 0 10 20 30 -200 0 200 0 10 20 30Frequency (Hz) Frequency (Hz) Lag (ms) Frequency (Hz)

Figure 7. Statistical signal analysis of perturbationsA, time- and frequency-domain analysis of a 640 ms time window immediately preceding onset ofperturbations extracted fIom the same fifty movements as analysed in Figs 2 an(d 4. From left to right:power spectrum of the acceleration, power spectrum of the EMIG, cross-correlation between accelerationand EMG and phase spectrum between acceleration and E..MG. The power spectr a show peaks at 8-1 0 Hz.Cross-correlation and phase spectrum indicate that the EAIG is leading the acceleration by 23 ins. Linearlregression over parts of the phase spectra in A and B are indicated by straight lines, as in Fig. 6.B, analysis of a 640 ms time window immediately following the onset of perturbations for the same

movement data base as in A. The acceleration and EMIG powei spectra show peaks in the 10-20 Hz range,

with a maximum at 15 Hz. The cross-correlation and phase spectrum indicate that the acceleration isleading the EMG by 60 ins.

EMG, although the amount is difficult to assess accuratelyfrom the cross-correlation alone. The delay is also reflectedby the shape of the phase spectrum, which shows a clearnegative slope in the 10-20 Hz range, which indicates thatthe EMG is lagging the acceleration rather than vice versa

as in Fig. 7A, where the slope is positive. The average timedelay from acceleration to EAIG calculated from the phasespectrum in Fig. 7B is 60 ms. The coherence (not illustrated)was significantly above zero for a broad range of frequenciesfrom 1 to 20 Hz, with a non-significant notch at 8-9 Hz,wrhich again provides support for the notion that there is anon-going 8-10 Hz activity in the EAIG, which is onlyweakly reflected in the acceleration during the appliedperturbations.

The findings demonstrated in Fig. 7 were identical for allsubjects and movement directions. In summary, the cross-

correlations and phase spectra for the time periodsimmediately preceding and following the onsets ofperturbations reveal that the pattern of EMG leading theacceleration by about 20 ms found during the naturalmiovemnents is replaced by a pattern of acceleration leadingthe EMG by about 55-60 ms during the perturbations.These findings are consistent with those obtained with theaveraging technique described previously.

DISCUSSIONThe main aim of the present study was to test thehypothesis that the 8-10 Hz discontinuities that seem to bepresent in all subjects during slow finger movements are

caused exclusively by reverberations in external feedbackloops, particularly stretch reflex loops.

Considering the effects of muscle spindle primary afferentsat the spinal level, it seemns that several mechlanisms wouldtend to promote discontinuities. As soon as the movementis started, the stretch of muscle spindles in the lengtheningmuscle would provide reflex excitation of its mnotoneuronesand reciprocal inhibition of the shortening muscle that isdriving the movement. Both effects would tend to deceleratethe voluntary movement. Simultaneously, the spindles ofthe shortening musele would be unloaded and their firingwould diminish if not compensated by fusimotor activity.The resulting decrease of the I a input from the shorteningmuscle would support the deceleration of the movementthrough two separate reflex effects, i.e. reduced excitationof the motoneurone pool of this muscle and reciprocaldisinhibition of its antagonist. Hence the four mechanismswAtould altogether tend to slowT down the movemnent.Subsequently, wlhen the angular velocity declines, these

904

A

a)a 10

InC

co 52*-

a)a)0< 0

B




reflex effects would diminish and the synaptic drive fromhigher centres would exert a relatively stronger effect again,and produce a new phase of acceleration in the desireddirection, i.e. another similar cycle would be restarted.

The hypothesis that the 8-10 Hz discontinuities areproduced by stretch reflexes requires that the timing aswell as the strength of the reflex effects are sufficient tofully account for the observed discontinuities. The findingsof the present study indicate that the hypothesis can berefuted on both accounts, as will be elaborated below.

The timing of stretch reflexes did not fit the hypothesisThe EMG response to the perturbations reached its peak50-60 ms after peak acceleration. This was demonstratedin direct measurements of the response latency as wTell aswith signal analyses in the time and frequency domains.These figures can be compared directly with the EMGmodulations during the self-produced discontinuities,where EAIG peaked 12 ms after peak acceleration. Hencethere was an obvious mismatch between the reflex response

and the self-generated discontinuities with regard to theEAIG pattern in relation to the kinematics (cf. Figs 3 and4A with Figs 3 and 4B). It seenms that this mismatch intiming would exclude the spinal as well as the supraspinalstretch reflex as the main mechanism for the self-generateddiscontinuities. This conclusion was confirmed with signalanalyses of time windows before and after the perturbations(Fig. 7), which clearly indicated that two different processeswere working, one before the perturbation and an additionalone after the perturbation.

It is interesting to consider the timing of the reflexresponse to perturbations in relation to published data onthe H reflex, which is generally accepted as the equivalentof the spinal stretch reflex. The latency of the H reflex asmeasured from the EMIG in forearm muscles is about 16 ms,with a peak effect a few milliseconds later (Deschuytere,Rosselle & De Keyser, 1976). For a comparison with ourdata it is fortunate that the electrical stimulus used to elicitthe H reflex was applied approximately at the same levelalong the arm where we recorded the afferent volley toperturbations. Hence it is highly relevant to compare theH reflex latencies with the latencies from the moment whenthe afferent impulses appear at our recording electrode tothe reflex response in the EMG, i.e. 40 and 50 ms forflexion and extension movements, respectively. Thesefigures were obtained by subtracting, from the total delaybetween perturbation and EMIG response (53 and 63 ms,riespectively), the time of the peripheral events up to theafferent impulses passing the recording electrode, i.e. 13 ms.Hence these estimates yielded a large mismatch betweenH reflex latency (16 ms) and the response to perturbation(40-50 ms), indicating that the main reflex response due tostretch perturbations of the present study was not spinalbut more likely to be transcortical in origin (Matthews,

Stretch reflexes were weakThe perturbations, which consisted of 3 deg fast movements,constituted considerably stronger stimuli for the musclespindles than the stretch phase of the self-generateddiscontinuities. The diffierence was explicitly reflected inthe response of the single I a afferents as shown in Fig. 3Aandc C, because the imposed perturbations elicited regulallytwo or more impulses with low temporal scatter, whereas

the self-generated discontinuities were merely associatedwith a broad modulation of firing probability, as describedin a previous study (WVessberg & Vallbo, 1995). Acorresponding difference was not found in EMG activity,rather the opposite because the reflex effects of theperturbations were actually smaller than the EMGmodulations of the discontinuities, at least during flexionmovements, in spite of the afferent activity being strongerwith perturbations. Although the reflex response toperturbation was stronger during extension than duringflexion, there was no E1IG modulation appropriately timeldto be consistent with a reflex effect during self-generatedmovements (Fig. 4B). It seems that the mismatch in the sizerelation between afferent response and EMG activity isfurther strong support for the conclusion that thediscontinuities were not produced by stretch reflexmechanisms, either spinal or supraspinal.

The findings that the reflexes were generally quite weak,and particularly that the spinal component was

considerably weaker than the long-latency component, are

congruent with previous observations of reflexes in thedistal limb segments in humans, notably in the musclescontrolling the hand (Melvill-Jones & Wtatt, 1971; VTallbo,1974; Houk & Rymer, 1981).

Time- and frequency-domain analyses confirm weakstretch reflexes during self-generated movementsAs a compleiment to the perturl)ation analysis, the Cr'oss-correlation betwTeen EMG and acceleration wvas examined.The lhypothesis that the stretch reflexes produced the8-10 Hz oscillations predicts a rather strong and consistentpositive peak at short negative latencies in the cIross-

correlogram, because muscle length changes would be themain factor to elicit the EMG modulations of the 8-10 Hzdiscontinuities. Although the analysis revealed a peak atnegative latencies, it was small and highly variable in timeas well as in size. Actually it was smaller and more variablethan the positive peak at positive latency which is very

likely to be due to the periodic 8-10 Hz modulations of theEMG. Hence it seems justified to conclude that the peak atnegative lags was also due to such periodic modulations.The complemnentary analysis in the fiequency domain fullyagreed with this conclusion.

Possible stretch reflex effects on the movementkinematics?While it appears obvious that stretchl eflexes cannot

Farmer & Ingram, 1990; Palmer & Ashby, 1992).

J. P/lysiol.493.3 905

produce the 8-10 Hz discontinuities, it may be of intei-est




to consider if the reflexes contribute at all to the kinematics.As outlined above, a basic effect would be that reflexexcitation of the lengthening muscle would contribute todeceleration of the movement. The EMG response toperturbations of the lengthening muscle peaked 53 msafter peak acceleration during flexion movements. Since theelectrokinematic delay was 19 ms, as found in the cross-correlation analysis, any decelerative effect due to thestretch reflex would peak about 53 + 19 = 72 ms afterpeak acceleration. This is clearly too late for the lengtheningmuscles to contribute much to the deceleration because thedecelerations are over within 40-60 ms after peakacceleration, at least when subjects perform very slowmovements (cf. Vallbo & Wessberg, 1993, their Fig. 6).Hence a reflex response at 72 ms after the peak accelerationwould not contribute to the deceleration.

In faster movements, on the other hand, the reflex mightenter more appropriately to exert a braking action becausethe accelerative as well as the decelerative phases last longerand together they cover the whole cycle. Hence a reflexresponse at 72 ms coincides with part of the decelerativephase and might provide some assistance to the brakingaction. To what extent an effect of this nature is significantor not remains to be assessed, however.

Other reflex mechanismsOn the basis of the present findings it seems justified toconclude that neither the spinal nor the supraspinal stretchreflex can account for the periodic modulations of the EMG.It may be asked if sensory feedback from other receptorsmay contribute significantly to an oscillation at 8-10 Hzduring these movements. Candidates would be skin afferentsas well as groups II and Ib muscle afferents. Generally, itseems that several arguments put forward for the stretchreflex also rule out reflex effects from these systems, aselaborated below.

The latency of reflexes from group II muscle spindleafferents would be at least as long as the latency of the I a-elicited spinal stretch reflex, which was shown to be toodelayed to produce the 8-10 Hz discontinuities. Reflexeffects from skin afferents in the fingers have latencies inthe 40-60 ms range as measured by the EMG response,implying that the timing is not appropriate for sustainingan 8-10 Hz oscillation (Jenner & Stephens, 1982; Chen &Ashby, 1993).

A final alternative would be Golgi tendon organ afferents.A reflex effect that would decelerate the voluntarymovement seems reasonable at first sight through autogenicinhibition of the driving muscle and reciprocal facilitationof the antagonist. In a previous study it was found thatsome of the single Ib afferents were activated roughly inthe opposite phase compared with the muscle spindleafferents, i.e. close to peak activation of the parent muscle.However, a larger proportion of them were not significantlymodulated by the discontinuities (Wessberg & Vallbo,

1995). Moreover, the firing pattern of the Ib unitssuggested that the discharge was correlated with theactivity of one or a few motor units rather than with theoverall kinematic or EMG pattern during the movements.Later studies, as well as complementary analyses in thefrequency domain, agreed with this interpretation(N. Kakuda & J. Wessberg, unpublished observations).These considerations make it unlikely that Ib afferentswould be a significant factor in modulating the motoroutput.

Finally, the finding that the correlation between kinematicsand subsequent changes in EMG is weak and variable, asshown by the small and variable peak at negative lags inFig. 5, is a general and strong argument against reflexmechanisms, not only against an individual reflex butagainst combined effects as well. Hence it seems reasonableto infer that the pulsatile motor output is largely producedby neural mechanisms within the central nervous system.

Central mechanismsSince reflex mechanisms seem to play a minor role in thegeneration of the 8-10 Hz discontinuities, it is of interestto consider which central mechanisms might be involved.Recurrent inhibition has been suggested as a factorpromoting oscillations in tremor (Elble & Randall, 1976).However, it has been shown that this system would tendto desynchronize motoneurone activity rather thansupport an oscillation (Stein & Oguztoreli, 1984; Windhorst& Kokkoroyiannis, 1992). Moreover, a number ofinvestigations suggest that the Renshaw system is lackingin the motoneurones innervating the finger muscles (Person& Kozhina, 1978; H6rner, Illert & Kiummel, 1991; Katz,Mazzocchio, Penicaud & Rossi, 1993).

Another central factor to consider is the intrinsic frequency-regulating properties of the motoneurones that might giverise to modulations of the motor output. Newly recruitedneurones typically fire with frequencies in the 6-10 Hzrange (Freund, Biidingen & Dietz, 1975). The force producedby these motor units could modulate the total muscle forceat their firing rates, since the last recruited units are thestrongest. This mechanism has been suggested forphysiological tremor, and it has been emphasized that asynchronization or a correlation between the firing of theindividual motoneurones is not required to achieve amodulation of the net motor output (Taylor, 1962; Allum,Dietz & Freund, 1978).

However, it seems that the pronounced modulations of thegross EMG that were often seen during slow fingermovements could only be produced by a substantialsynchronization. One mechanism which would provide acertain degree of synchronization, but probably notsufficient to explain our findings, may result from acommon synaptic input even when the motoneurones arenot driven or modulated at the input frequency; their firingrates are determined by the intrinsic properties of the

906 J Physiol.493.3




motoneurones. An alternative hypothesis would be thatthere exists an 8-10 Hz modulation of the synaptic inputto the motoneurone pool that would modulate themotoneurone firing at this rate, regardless of their meanfiring rate. It seems that the latter hypothesis is easier toreconcile with two prominent features of the 8-10 Hzdiscontinuities. One is the consistent 8-10 Hz frequency ofthe discontinuities regardless of overall speed of movement,and the other is the close coupling between activities ofagonist and antagonist muscles, which was often seen. Infact, a recent study indicates that motor units aremodulated at 8-10 Hz, even when they are firing at othermean rates (Kakuda & Wessberg, 1995), suggesting thatthe synaptic input to the finger motoneurones is modulatedat 8-10 Hz during voluntary movements.

If it can be ruled out that the 8-10 Hz discontinuitiesresult from intrinsic frequency-regulating motoneuroneproperties, a pulsatile modulation of the commandreaching the motoneurone pool seems to be the likelyremaining alternative. A pulsatile input to motoneurones at8-10 Hz might be generated either at the spinalinterneuronal level or at supraspinal levels, the latteralternative implying that the descending motor commandfor slow finger movements contains a pulsatile componentrather than being smoothly continuous, and entails asophisticated pattern of alternating pulses to the agonistand antagonist muscles, as speculated in the report thatoriginally described the 8-10 Hz discontinuities (Vallbo &Wessberg, 1993). While the present study exclusivelyconcerns such periodic phenomena during finger movements,it should be noted that a purely central 8-12 Hzcomponent of physiological tremor during position holdinghas previously been suggested (Elble & Randall, 1978;Elble & Koller, 1990). The possible relationship betweenthese two phenomena remains to be assessed.

AL-FALAHE, N. A. & VALLBO, A. B. (1988). Role of the humanfusimotor system in a motor adaptation task. Journal of Physiology401, 77-95.

Au.l um, J. H. J., DIETZ, V. & FREUND, H. J. (1978). Neuronalmechlanisms underlying physiological tremor. Journ al ofNeurophysiology 41, 557-571.

ANDREWS, C. J., BURKE, D. & LANCE, J. W. (1973). The comnparison oftremors in normal, Parkinsonian and athetotic man. Journal of theNeurological Sciences 19, 53-61.

BENDAT, J. S. & PIERSOL, A. G. (1986). Ranidomr Data. John XViley &Sons, New York.

BROCKWELL, P. J. & DAVIES, R. A. (1991). Time Series: Theory andAMethods. Springer-Verlag, New York.

BURNE, J. A., LIPPOLD, 0. C. J. & PRYOR, Al. (1984). PropIioceptorsand normal tremor. Journal of Physiology 384, 559-572.

CHEN, R. & ASHBY, P. (1993). Reflex responses in upper limb musclesto cutaneous stimnuli. Caniadiani Journial of Neurological Scienlces20, 271-278.

DESCHUYTERE, J., ROSSELLE, N. & DE KEYSER, C. (1976).Monosynaptic reflexes in the superficial forearm flexors in man andtheir clinical significance. Journal of Neurology, Neurosurgery andPsychiatry 39, 555-565.

EDIN, B. B. & VALLBO, A. B. (1990). Classification of human musclestretch receptor afferents: A Bayesian approach. Journal ofNeurophysiology 63, 1314-1322.

ELBLE, R. J. & KOLLER, NV. C. (1990). Tremor. The Johns HopkinsUniversity Press, Baltimore.

ELBLE, R. J. & RANDALL, J. E. (1976). Mlotor-unit activity responsiblefor 8- to 12-Hz component of human physiological finger tremor.Journal of Nleurophysiology 39, 370-383.

ELBLE, R. J. & RANDALL, J. E. (1978). Alechanistic components ofnormal hand trem-or. Electroencephalography and Clin icalNeurophysiology 44, 72-82.

FREUND, H.-J., BODINGEN, H. J. & DLETZ, V. (1975). Activity ofsingle motor units from human forearnm muscles during voluntaryisometric contractions. Journal of Neurophysiology 38, 933-946.

HAGBARTH, K.-E. & YOUNG, R. R. (1979). Participation of the stretchreflex in human physiological tremor. Brain 102, 509-526.

HORNER, Al., ILLERT, 1\I. & KOMMEL, H. (1991). Absence of recurrentaxon collaterals in motoneurones to the extrinsic digit extensormuscles of the cat forelimb. Neuroscience Letters 122, 183-186.

HoUK, J. C. & RYMER, W. Z. (1981). Neural control of muscle lengthand tension. In Handbook of Physiology, section 1, The N\ervousSystem, vol. II, Mlotor Control, part 1, edl. BROOKS, V. B.,pp. 257-324. \Villiam & WN'ilkins, Baltimore.

JACKS, A., PROCHAZKA, A. & TREND, P. 8. J. (1988). Instability inhuinan forearm movemients studied with feed-back-controlledelectrical stimulation of muscles. Journial of Physiology 402,443-461.

JENNER, J. R. & STEPHENS, J. A. (1982). Cutaneous reflex res)Consesand their central nervous pathways stu(lies in man. Journal ofPhysiology 333, 405-419.

JOYCE, G. C. & RACK, P. M. H. (1974). The effects of load and force ontremor at the normal human elbow joint. Journal of Physiology240, 375-396.

KAKUDA, N. & WESSBERG, J. (1995). AModulation of the impulse trainsof single motor units in 8-10 Hz discontinuities during slow fingermovement in man. In Abstracts of the Fourth International BrainResearch Orgatnizationi lVorld Contgress of Neuroscience, p. 339.Rapid Communications, Oxford.

KATZ, R., AIAZZOCCHIO, R., PENICAUD, A. & Rossi, A. (1993).Distribution of recuirent inhibition in the hluman upper limb. ActaPhysiologica Scandinavica 149, 183-198.

LiPPOLD, 0. C. J. (1970). Oscillation in the stretch reflex arc and theorigin of the rhythmnical, 8-12 C/S coml)onent of plhysiologicaltremor. Journial of Physiology 206, 359-382.

MARSDEN, C. D. (1984). Origins of normual and pathological tremor. InMIovemenit Disorders: T'remtor, ed. FINDLAEY, L. J. & (CAP1LI)bEO, R.,pp. 37-84. AMacmillan, London.

AIATTHEWS, P. B. C., FARMER, S. F. & ING,RAM, D. A. (1990). On thelocalization of the stretch reflex of intrinsic hand muscles in apatient with mirror movements. Journial of Physiology 428,561-577.

AIELVILL-JONES, G. & WATT, D. G. D. (1971). AMuscular control oflanding from unexpected falls in man. Journial of Physiology 219,729-737.

PALMER, E. & AsHBY, P. (1992). Evidence that a long latency stretchreflex in humans is transcortical. Journal of Physiology 449,

J. Phy8iol.493.3 907

429-440.



908 J Wessberg and A. B. Vallbo J Physiol.493.3

PERSON, R. S. & KoZH1NA, G. V. (1978). Study of orthodromic andantidiomic effects of nerve stimulation on single motoneurones ofhumiian hand muscles. Electromnyography and Clini1cal N21europhysiology18, 437-456.

PIROC(HAZKA, A. & TREND, P. S. J. (1988). Instability in humnanforearm movements studied with feed-back-controlled musclevibration. Jouernial of Physiology 402, 421-442.

SANES, J. N. (1985). Absence of enhanced physiological tremor inpatients without mluscle or cutaneous afferents. Journal ofNeurology, Neurosurgery and(l Psychiatry 48, 645-649.

STEIN, R. B. & OGUZTORELI, Al. N. (1976). Tremor ancl otheroscillations in neuromnuscular systems. Biological cybernetics 22,147-157.

STEIN, R. B. & OGUZTORELI, Al. N. (1984). Alodification of muscleresponses by spinal circuitry. Neuroscience 11, 231-240.

TAYLOR, A. (1962). The significance of grouping of motor unit activity.Journlal of Physiology 162, 259-269.

VALLBO, A. B. (1974). The significance of intrcamuscular receptors inloacl compensation during voluntary contractions in man. In ControlofJ Posture and Locomotiont, ed. STEIN, R. B., PEARSON, K. C.,SMI1TH, R. S. & REDFORI), J. B., pp. 211-226. Plenum Press, NexwYor0k.

VAiLIHo, A. B., HAG4BARTH, K.-E., TOREBJ61RK, H. E. & XVALLIN, B. G.(1979). Somatosensory, proprioceptive, and sympathetic activity inlhumnan pei-iplheral nerves. Physiological Reviews 59, 919-957.

VALLBO, A. B. & WXESSBERG, J. (1993). Organization of motor outputin slow finger movements in man. Journial of Physiology 469,673-691.

VALLBO, A. B. & WESSBERCG, J. (1995). Are slow finger movementsprocluced by a pulsatile motor command? In Alpha and GammaMlIotor Systems, ed. TAYLOR, A., GLADDEN, At. H. & DURBABA, R.,pp. 601-608. Plenum Piess, New York.

VALLBO, A. B. & NVTESBERCG, J. (1996). Proprioceptive mechanisms andthe control of finger miovements. In Hand and Brainl:Neurophysiology anld Psychology of Hand Moveinemnt, ed. WVING, A.,HAGGARD, P. & FLANAGAN, R., pp. 363-379. Academic Press,San Diego, CA, USA.

WVESSB3ERG, J. & VALLBO, A. B. (1995). Coding of pulsatile motoroutput by human muscle afferents during slow finger moxvemnents.Journal of Physiology 485, 271-282.

WINDHORST, U. & KOKKOROYIANNIS, T. (1992). Dynamic b)ehaviour ofalpha-motoneurons subjected to recurrent inhibition and reflexfeedback via muscle spindles. Neuroscience 47, 897-907.

YOUNG, R. R. & HAGBARTH, K.-E. (1980). Physiological tremiiorenhanced bry manoeuvres affecting the segmental stretch reflex.Joutrntal of Neurology, Neurosurgery anid Psychiatry 43, 248-256.

AcknowledgementsThis study was supported by the Swedish Medical ResearchCouncil (grant 14X-3548) and the G6teborg Medical Society. Wewould like to thank Sven-Ojvind Swxahn for valuable technicalsulport.Author's email addressJ. \Vessberg: w\[email protected]

Received 8 Jatnuary 1996; accepted 2 February 1996.



Pulsatile motor output in human finger movements is not dependent on the stretch reflex

Documents