RESEARCH ARTICLE

Lucinda K. Hughey Æ Joyce Fung

Postural responses triggered by multidirectionalleg lifts and surface tilts

Received: 23 March 2004 / Accepted: 19 January 2005 / Published online: 7 June 2005� Springer-Verlag 2005

Abstract The aim of the present study was to investi-gate the relationship between proactive and reactivecomponents of postural control. We contrasted thekinematic and electromyographic (EMG) responses tomultidirectional voluntary leg lifts with those elicited byunexpected surface tilts. In particular, we addressed therole of trunk stabilization following either a voluntaryor forced weight shift from double to single limb sup-port. Nine young female subjects stood with a standingposture of 45� toe-out and their arms abducted toshoulder level. On the experimenter’s signal, subjectseither (1) lifted one leg as fast as possible in one of sixdirections (R/L side, R/L diagonal front, R/L diagonalback) to a height of 45� or (2) maintained standing asthe support surface tilted at a rate of 53�/s to a heightof 10� in one of six directions (R/L-up, R/L diagonaltoes-up, R/L diagonal toes-down). For both tasks, ourresults showed that the center of pressure (COP) dis-placement began before or in conjunction with dis-placement of the center of mass (COM), after which theCOP oscillated about the horizontal projection of theCOM. In addition, the muscles were recruited in adistal-to-proximal sequence, either in anticipation ofthe voluntary leg lift or in response to the suddensurface tilt. Thus, the COP was being used dynamicallyto control displacement of the COM. The axial posturalstrategy comprising head, trunk, and pelvis movementswas quantified by means of principal component anal-ysis. More than 95% of the variance in the data couldbe described by the first two eigenvectors, which re-

vealed specific coordination patterns dominated bypelvis rotation in one direction and head/trunk rotationin the opposite direction. Unexpected surface tiltingelicited an automatic response strategy that focused oncontrolling the orientation of the head and trunk withrespect to the vertical gravity vector while trunk verti-cality was compromised for movement generation andthe recovery of postural equilibrium during leg lifting.In conclusion, regardless of the type (voluntary versusinvoluntary) or direction of perturbation, the strategyemployed by the central nervous system to control thebody COM displacement concerns mainly trunk sta-bilization.

Keywords Posture Æ Equilibrium Æ Anticipation ÆFeedback Æ Principal component analysis

Introduction

The purpose of any postural adjustment during uprightstance, whether anticipatory or reactive, is the controlof equilibrium. In the case of anticipatory adjustments,balance is achieved at the expense of postural orienta-tion, whereas automatic responses involve changes insegmental orientation to regain posture (Massion 1992).In either case, displacement of the body’s center ofmass (COM) must be minimized in order to avoid a fall(Winter 1990). However, the means by which the cen-tral nervous system (CNS) selects a control strategyfrom the overabundant number of possibilities is notwell understood. In particular, very little is knownconcerning the relationship between proactive andreactive response strategies. We hypothesized that, forall postural perturbations, the CNS strategy in thecontrol of balance is centered on the task of trunkstabilization.

There is growing evidence that kinematic responsefollowing perturbation is focused on the active controlof trunk orientation. In humans the upper body

L. K. Hughey Æ J. FungSchool of Physical & Occupational Therapy, McGill University,3654 Prom. Sir William Osler, Montreal,Quebec, Canada, H3G 1Y5

L. K. Hughey Æ J. Fung (&)Research Center (site of the Montreal Interdisciplinary ResearchCenter in Rehabilitation, CRIR), Jewish Rehabilitation Hospital,3205 Place Alton Goldbloom, Laval, Quebec, Canada, H7V 1R2E-mail: joyce.fung@mcgill.caTel.: +1-450-6889550Fax: +1-450-6883673

Exp Brain Res (2005) 165: 152–166DOI 10.1007/s00221-005-2295-9

accounts for approximately two-thirds of the total bodymass. Therefore, tasks that require the preservation oferect posture are dependent upon the control of trunkposition and velocity. Also, an internal representationof the trunk axis with respect to an external referenceframe (i.e., gravity) is essential for determining headorientation and stabilization in space (Mergner et al.1993). Because both the vestibular and visual systemsreside in the head, maintenance of head stability, par-ticularly during dynamic activity, is paramount (Pozzoet al. 1995). However, head stabilization in space ap-pears to be task-specific whereas trunk stabilization isindependent of task (Assiante et al. 1997). Assianteet al. (1997) compared angular displacement of thehead, trunk, and leg during single and double foothopping and found that subjects stabilize both the headand trunk at landing but only actively control trunkposition during flight. Thus, they concluded that trunkstabilization in space was a primary controlled variablewhereas head stabilization depended upon the givenpostural task.

Head and trunk stabilization with respect to thevertical gravity axis has been shown to provide an ego-centric reference value for posture and movementcoordination (Mergner et al. 1991; Mouchnino et al.1993). Upper body stabilization aids in focal targetlocation and trajectory planning (Andersen et al. 1993;Hodges et al. 1999). During locomotion, for example,vertical displacement of the head is minimized to pro-mote gaze stabilization (Assiante and Amblard 1993;Pozzo et al. 1990, 1995) while the trunk is stabilized inthe frontal plane (Winter 1990). Baroni et al. (1999) haveshown that during long-term exposure to microgravity,trunk orientation remains stable with little change overtime whereas COM location is adjusted throughoutflight. Postural responses to activities such as voluntaryleg lifting demonstrate the dependence of balance on thedirection of the gravity force vector (Mouchnino et al.1993; Rogers and Pai 1990). Unlike the COM, there aresensory receptors in the trunk and legs that, in con-junction with vestibular input, may contribute to thecalculation of trunk orientation (Allum and Honneger1998; Hlavacka et al. 1996; Mauer et al. 2000). Indeed,consistent measurements of forward trunk inclinationrecorded in microgravity and underwater provide evi-dence that proprioceptive graviceptors aid in evaluatingvertical posture (Clement et al. 1988; Massion et al.1995, 1997).

Active control of the trunk may be achievedthrough the coordinated displacement of one or morebody segments. Mouchnino et al. (1992) found thatduring a lateral leg lift, trained dancers were able toretain trunk verticality by means of pelvis rotation.The feedforward rotation was achieved through cou-pled movement of the ankle and hip joints. Controlsubjects, on the other hand, were unable to maintainvertical trunk orientation despite specific instructionsto do so. Following externally triggered balance per-turbations, central response strategies have been shown

to involve coordination at the neck and hip joints sothat trunk verticality can be regained as quickly aspossible regardless of head position (Allum et al.1997).

Thus, assuming that active control of trunk orien-tation is fundamental to balance maintenance, animportant question is this: Do similarities exist in theway in which trunk orientation is controlled given acommon postural goal? Also, what is the relationshipbetween response strategies generated by self-initiatedmovement and those elicited by unexpected perturba-tion? To answer these questions, we examined posturalresponses following either a voluntary or unexpectedweight shift from single to double limb support. Apartfrom trunk orientation, of particular interest was thedisplacement of the center of pressure (COP), as theCOP can influence regulation of COM positioning.During stance, if vertical trunk orientation is con-trolled, balance can be maintained through smalladjustments in COM positioning as achieved throughthe foot-support surface interface (Roll et al. 2002;Kavounoudias et al. 2001, 1998). Therefore, in order toevaluate reactive and proactive CNS control of theCOM, we compared and contrasted muscle responsesand the subsequent three-dimensional changes in jointorientations and COP under the feet to expected,internally generated versus unexpected, externally trig-gered weight shifts that occur during multidirectionalvoluntary fast leg lifts and sudden support surface tilts.We suggest that regardless of the type (voluntary versusinvoluntary) or direction of perturbation, trunk stabil-ization will be the primary variable controlled by theCNS.

Methods

Subjects

A convenience sample of nine active young women(mean age 26±7.1) with no history of neurological ormusculoskeletal problems was recruited for this study.Their heights and weights ranged from 154.7 cm to169.8 cm and 41.94 kg to 60.61 kg, respectively. Allsubjects participated in some level of physical activity,from karate to basketball. All subjects gave their in-formed consent to the experimental protocol previouslyapproved by the institutional ethics committee.

Procedure

Two experimental protocols were performed: voluntaryleg lifts and unexpected surface tilts. The initial stanceposition was the same for both experiments: barefoot,weight evenly distributed across both feet, with heelstouching and oriented in a 45� toe-out position. Eachfoot was placed on an individual force plate (AMTIOR6-7). Arms were abducted to shoulder height (see

153

Fig. 1a). Subjects were instructed to maintain this initialposture.

Experiment 1

Following an audio signal, subjects were required to lifteither their right or left leg in one of three directions (45�diagonal front, �45� diagonal back, or 0� pure side,Fig. 1b) to a height corresponding to an inter-leg sepa-ration angle of 45�. Subjects were instructed to performthe movement as fast as possible with no knee flexionand to maintain the final leg position for a minimum of3 s. The choice of lifting leg and lift direction wererandomized within test blocks (i.e., one lift in each of thesix directions). Subjects completed five test blocks for atotal of 30 trials.

Experiment 2

Upon the experimenter’s trigger, the support surface wastilted in the pitch and roll planes. The support surfacewas mounted over a six-degree-of-freedom motion base

servo controlled by six electrohydraulic actuators (Fungand Johnstone 1998). Ten degrees of ramp-and-holdsurface tilt in one of six random axial directions (0� rightor left side-up, 45� diagonal toes-up, and 45� diagonaltoes-down) was presented at a peak velocity of 53�/s(Fig. 1b). The tilt magnitude and speed were sufficientenough to cause a shift from the initial two-footedstance (with an even weight distribution) to a position inwhich more weight was supported by a single limb (seevertical forces shown in Fig. 1b). However, the pertur-bation did not elicit a stepping response for balancerecovery from these subjects. Two catch trials with nosurface tilt were included to reduce the subject’s antici-pation of upcoming perturbation direction. Tilt direc-tion was randomized within test blocks (i.e., one tilt ineach of the eight directions). Subjects completed five testblocks for a total of 40 trials.

All experiments were conducted in a single day toensure that comparisons within subjects could be made.Subjects were given a practice session prior to recording.In addition, subjects were offered rest periods at anytime during testing, and the task was alternated betweenconditions upon completion of one test block in an effort

(b) Perturbation directions and coordinates

45°45°

45°

0° 0°L side lift L side-up

(L foot unload)

R diag front liftR toe unload

R diag back lift R heel unload

L diag back lift

L heel unload

L diag front liftL toe unload

Voluntary leg lift

L loadR load

Unexpected surface tilt

L loadR load

RL

600

300

0-250 0 250 500

Time (ms)

600

300

0-250 0 250 500

Time (ms)

600

300

0-250 0 250 500

Time (ms)

600

300

0-250 0 250 500

Time (ms)

600

300

0-250 0 250 500

Time (ms)

600

300

0-250 0 250 500

F z (

N)

F z (

N)

F z (

N)

F z (

N)

F z (

N)

F z (

N)

Time (ms)

R side liftR side-up

( R foot unload)

xy

z

(a) Kinematic set-upFig. 1 a Kinematic set-up with34 markers. Small circles showreflective marker placement(white circles represent posteriortrunk and pelvis markers). Thekinematic coordinate system (xy z) is shown on the side.b Coordinate system for bothvoluntary leg lifts andunexpected surface tilts. Arrowindicates raised leg or surfaceedge. Insets show loading andunloading forces from arepresentative subject (blacklines right force plate, gray linesleft force plate). Solid linescorrespond to the leg-lift task,and dashed lines correspond tosurface tilts. Note the shift fromeven weight distributionfollowing all perturbations

154

to reduce the effects of fatigue. Testing, including set-up,was completed within 2.5 h.

Data recording

A six-camera VICON 512 system (Oxford Metrics) wasused to capture three-dimensional position data at120 Hz from 35 retroreflective markers placed overanatomical landmarks (Fig. 1a). In addition, fourmarkers were placed along two orthogonal axes of themovable platform. A biomechanical model (Plug-InGait, Oxford Metrics) was used in conjunction withkinematic data and anthropometric measures (height,weight, leg length, and joint widths) to define bodysegments and joint angles and to calculate the total-bodyCOM.

Ground reaction forces and moments were acquiredat 1,080 Hz by two adjacent AMTI OR6-7 force platesembedded within the support surface. Resultant COP xand y vectors were calculated as the weighted sums fromthe individual signals, as described previously (Henryet al. 1998). Inertial components in forces and COP datawere corrected due to movement of the support surface(Preuss and Fung 2003).

Four lower limb and trunk agonist-antagonist musclepairs on each subject’s right side were instrumented withbipolar Ag-AgCl disposable surface electrodes (BlueSensor) to record electromyographic (EMG) signalsusing an eight-channel TELEMG system (BTS): tibialisanterior (TA), gastrocnemius medialis (MG), rectusfemoris (RF), semitendinosus (ST), tensor fascia latae(TFL), adductor (AD), rectus abdominus (RA), anderector spinae at the L3 level (ES). EMG signals wereamplified, digitized, and transmitted to the remoteamplifier via optical fibers. These signals were band-passfiltered (10–400 Hz low-pass) and sampled at 1,080 Hz.EMG signals were further full-wave rectified and low-pass filtered at 100 Hz during offline analysis.

Data analysis

All data were time-adjusted to movement onset. Move-ment onset was defined as the time at which the differ-ence between the vertical velocity of the malleolusmarker (experiment 1) or platform markers (experiment2) and the mean 150 ms background velocity value wasequal to or greater than twice the standard deviation(95%) (Mouchnino et al. 1992). For experiment 1, ananticipatory phase and a maintenance phase were alsodesignated. The start of the anticipatory phase wasindicated by the first change in the lateral velocity of theCOP equal to or greater than the mean background plustwice the standard deviation (Mouchnino et al. 1992).The maintenance phase was set as the 1,000-ms periodafter which the vertical velocity of the malleolus markerslowed to 0±5% of its maximum. Because of the elec-tromechanical delay between muscle firing and force

generation, each phase was shifted 30 ms earlier forEMG analysis. An example of these time divisions for asingle subject can be seen in Fig. 2a. The data fromexperiment 2 were subdivided into predetermined timeintervals for analysis: 0–100 ms from stimulus onset,short-latency reflex period; 100–225 ms, balance cor-rection phase I; 225–350 ms, balance correction phaseII; 350–700 ms, stabilizing phase. For EMG analysis,the stretch reflex phase was shortened by 30 ms to end at70 ms. The subsequent time intervals retained the samelength but began 30 ms earlier. An example is shown inFig. 2b.

Three-dimensional kinematic analysis was carried outon the head, trunk, and pelvis segments. It is importantto note that each segment was modeled independently.The global head angle was taken as the angle betweenthe center of the four head markers and the verticalgravity vector. The trunk segment, which represents theupper trunk, was defined as the plane described by theclavicle, sternum, C7, and T10 markers. Trunk inclina-tion was taken as the angle between the center of thetrunk plane and the vertical axis system. The pelvis wasmodeled as a plane that contained the two posterior andtwo anterior superior iliac spine markers. The pelvisangle was taken as the angle between the pelvis and thevertical axis. Total angular excursion for each variablewas calculated, and the relative displacements betweenvariables were examined. Data from different trials ofeach individual were ensemble-averaged across eachdirection. These averages were then pooled to produce apopulation average for each direction.

In addition, segmental coordination patterns of thehead, trunk, and pelvis were evaluated in the sagittal,frontal, and horizontal planes by means of principalcomponent analysis (PCA). These principal componentsare a linear combination of the variance in the data suchthat each component is statistically independent of oneanother and describes decreasing smaller amounts of thevariance. For each trial of each subject, a covariancematrix R, containing the mean centered time-varyingangular displacements of the planar variables, wascomputed. PCA was used to determine the three rank-ordered eigenvectors, u1–u3, and associated scores of Rthat corresponded to the orthogonal directions of max-imum variance in the data set. To further improve sep-aration between variables with intermediate versus verylarge or very small loadings, a Varimax rotation wasperformed (Merkle et al. 1998). Thus, all the axes in thedata set were mathematically rotated such that the firstnew axis corresponded to the direction of the greatestvariance in the data. Each successive axis represented themaximum residual variance. Coordination patterns werethen assessed by using the first two eigenvectors, u1 andu2, to define a plane of angular covariation. The thirdeigenvector, u3, was used to determine the orientation ofthe plane in the data space. A similar method has pre-viously been employed for the analysis of foot, shank,and thigh segment orientations during locomotion(Borghese et al. 1996; Bianchi et al. 1998; Grasso et al.

155

2000). The ‘‘fit’’ of the plane was evaluated by the pro-portion of the variance described by the first two ei-genvectors.

Muscle latencies were determined manually as thefirst burst that exceeded a threshold of two standarddeviations above the background signal and lasting atleast 25 ms. A muscle had to be recruited at least threeout of five trials to be considered to contribute to dy-namic postural response. The mean integral of eachmuscle response was determined by integrating the areaunder the EMG response curve for each time segment ofinterest after filtering the data at 10 Hz (see Fig. 2). Themean background, 150 ms before the start of the antic-ipatory phase (experiment 1) or prior to platform onset(experiment 2), was subtracted. The data were thennormalized to the maximum response for each muscleregardless of perturbation direction or type of pertur-bation. Lastly, the integrals from each subject for eachdirection were averaged so that an ensemble groupprofile could be determined. Because each subject wasinstrumented for EMG only on her right side, all datafrom right leg lifts and right-side-up/right-diagonal-upsurface tilts were termed ‘‘unloaded limb,’’ while left leglifts and left-side-up/left-diagonal-up surface tilt datawere termed ‘‘loaded limb.’’

Comparisons between voluntary and automatic re-sponses were made using a repeated-measure two-way

analysis of variance (ANOVA), the independent vari-ables being task (voluntary leg lifts versus unexpectedsurface tilts) and direction of perturbation (see Fig. 1b).A Kenward-Roger adjustment was performed on thedegrees of freedom to reduce the prevalence of type 1errors. A P-value of 0.05 was accepted to be significantin determining the main and interaction effects, afteradjustment with the Bonferroni test.

Results

COM and COP

Figure 3a shows the relation between peak COM/COPexcursions and leg lift (left column) or surface tilt (rightcolumn) direction. Statistical analysis of peak-to-peakmedial-lateral COM displacements revealed significantmain group effects for task (F[1,15.9]=473.89,P<0.001) and perturbation direction (F[7,57.9]=7.93,P<0.001) as well as a task-versus-direction interaction(F[7,54]=15.87, P<0.001). Medial-lateral COP dis-placements also demonstrated significant main effects(task: F[1,16.4]=233.23, P<0.001; direction: F[7,56.3]=44.29, P<0.001) and an interaction (F[7,55.2]=31.57,P<0.001). Further comparison showed the magnitudeof the displacements to be greater during leg lifting. In

(a) Voluntary leg lift (45° - R diag front lift)

Dis

plac

emen

t(m

m)

Velocity

(mm

/ms)

Anticipatoryphase

Movementphase

Maintenancephase

-200 0 200 400 600 800 100050

0

50

TA

MG

EM

G(m

V)

Time (ms)

0

0

500

0

2000Ankle velocity

Ankle displacement

-300

0

300

-1200

0

1200

COP displacementCOP velocity

(b) Unexpected surface tilt (45°- R diag toes-up tilt)

0

5

10

0

25

50Platformdispalcement

PlatformvelocityD

ispl

acem

ent

(mm

) Velocity

(mm

/ms)

Stabilizingphase

Correction

50

0

50

TA

MG

EM

G(m

V)

0

0 200 400 600

Time (ms)

phase I phase II

Fig. 2 Time integrals.a Voluntary leg lifts. Verticalankle displacement/velocity (z-direction), COP displacement/velocity (medial–lateral), andshank muscle activation duringright diagonal front leg lift (45�)for a representative subject.b Unexpected surface tilts.Platform rotation/velocity(z-direction) and shank muscleactivation during right diagonaltoes-up rotation for the samesubject. Electromyographictime intervals precede kinematicintegrals by 30 ms

156

addition, medial-lateral displacement of the COM andCOP were sensitive to perturbation direction only fol-lowing unexpected surface tilts, in which case the overallexcursions were largest during pure roll tilts (0�), withpeak excursions of 35–40 mm for the COM and 130–140 mm for the COP compared with values half as greatfor diagonal surface tilts (45� and �45�). In the anterior–posterior plane, significant main effects for COM dis-placement (task: F[1,15.1]=92.77, P<0.001; direction:F[7,55.5]=16.54, P<0.001) and a task-versus-directioninteraction (F[7,56.8]=62.37, P<0.001) were found.Anterior–posterior displacement of the COM was twiceas great during front diagonal leg lifts (45–55 mm)compared with the other leg lift directions (20–35 mm)or following multidirectional surface tilts (15–25 mm).As for anterior–posterior COP displacement, no main

effect for task was revealed (F[1,16.1]=6.79, P>0.02)because of the Bonferroni adjustment. However, both amain effect for direction (F[7,112]=16.76, P<0.001) anda task-versus-direction interaction were found(F[7,112]=14.36, P<0.001). The interaction term re-flects the observed directional sensitivity of the anterior–posterior COP displacement associated with multi-directional surface tilting, in which case roll plane tiltingresulted in 50% smaller displacements than perturba-tions in the diagonal planes.

Figure 3b presents the horizontal projection (averageof five trials from nine subjects) of COP and COM forboth the voluntary (left column) and triggered (rightcolumn) tasks from each of the right-sided perturbations(45R�, 0R�, and �45R�). For all directions, COP duringthe leg lift task was predominately characterized by

(b) Average COM and COP horizontal plane trajectories

Voluntary leg lift Unexpected surface tilt

-50

0

50

-150 -100 -50 0 50 100 -150 -100 -50 0 50 100RightLeft

Fro

ntB

ack

RightLeft

COP(mm)

COM(mm)

0

0

A/P

(m

m)

-50

0

50

Fro

ntB

ack

A/P

(m

m)

M/L (mm) M/L (mm)

(a) Peak-to-peak COM and COP displacements

0

50

100 mm

0

150

300 mm

COP (+SE) COM (+SE)

M/L(mm)

A/P(mm)

Voluntary leg lift Unexpected surface tilt

–45° –45°

–45° –45° –45° –45°

–45° –45°

45°

0°

45°

0°

45°

0°

45°

0°

45°

0°

45°

0°

45°

0°

45°

0°

Left Right Left Right

Left Right Left Right

45˚ Diag Front/Diag Toes-up

– 45˚ Diag Back/Diag Toes-down0˚ Side/Side-up

Fig. 3 COM/COP. a AverageCOM/COP peak-to-peakdisplacements over entire trialfor all nine subjects. Blacktriangles refer to COM (+SE);gray triangles refer to COP(+SE). Note scale magnitudes:anterior–posterior displacementis one-third that of medial-lateral displacement.b Ensemble average horizontalplane trajectories of COMand COP for right side leg liftsand right-side-up surface tilts(left side is a mirror image).Displacement begins atzero–zero

157

lateral displacement. As expected, an anticipatory lateralshift away from the support side averaging 96±5.5 mmin magnitude was found. Onset of the lateral displace-ment occurred between 283±115.1 ms (R side lift, 0�)and 365±93.6 ms (L diagonal back, 270L�) beforemovement began. Forward leg lifts also produced aslight forward anticipatory adjustment. During themovement phase, there was a large and backward COPshift in the direction of the support leg (220±9.8 mm).In addition, there was a relatively slow posterior COPdisplacement followed by a quick forward shift. Themaintenance phase included anterior–posterior oscilla-tion and a small reverse lateral shift. COP responsefollowing surface tilts, unlike the voluntary task, wasdirection-specific. COP excursion began in the directionof the perturbation (that is, toward the unloaded limb;see Fig. 1b). Then, during the balance correction phase,the COP made a linear shift reversal away from theunloaded limb. COP displacement during the stabiliza-tion period included a return toward, but never regain-ing, the neutral start position.

During voluntary leg lifting, anterior–posterior COMdisplacement followed a direction-specific path. For-ward leg movements resulted in forward COM dis-placements while backward movements led to a smallforward shift, approximately 10 mm, followed bybackwards displacement as the leg was lifted. Unlike theCOP, the COM continued to move laterally in thedirection of the support leg during the position main-tenance phase to reach a displaced position near that ofthe COP. Following the unexpected surface tilts, theCOM excursion pathway moved away from the stimulusdirection towards the loading side. In summary, adiagonal toes-up surface tilt (45�) resulted in a back-wards shift whereas a diagonal toes-down tilt (�45�) ledto a forward shift. Interestingly, toes-down tilts resultedin a 50% greater overall COM displacement than toes-up tilts (12±9.4 mm versus 23±4.5 mm).

Head, trunk, and pelvis

Head, trunk, and pelvis angular excursions for thesagittal, frontal, and horizontal planes are shown inFigs. 4a and 5a, respectively. These examples are froma left diagonal front leg lift (45L�) and left diagonaltoes-up tilt (45�) from a single subject. The kinematicresponse to diagonal perturbations, whether triggeredby voluntary leg lifts or unexpected surface tilts, was acombination of the corresponding orthogonal re-sponses (for example, front/side, toes-up/side-up, back/side, or toes-down/side-up). In general, the head, trunk,and pelvis followed a set excursion path based onperturbation direction.

As shown in Fig. 4a, during leg lifting the pelvisdisplacement was larger than that of either the trunk orhead and was exemplified by rotation away from thelifting leg in the sagittal and frontal planes and towardthe lifting leg in the horizontal plane. As expected,

maximal excursions were seen during diagonal leg lifts inthe sagittal plane (�15–25�) and lateral leg lifts in thefrontal plane (�25�). Head and trunk displacementswere smaller (�5� and �10�) and involved a schemesimilar to that of the pelvis. During unexpected surfacetilts, as illustrated in Fig. 5a, pelvis displacement wasgreater than that of the head or trunk only in the frontaland horizontal planes. Interestingly, trunk and pelvismovement in the frontal and horizontal planes involvedrotation of the two segments in opposition. The pelvistilted toward the support limb in the frontal plane, whilethe trunk and head tilted toward the unloaded limb. Inthe horizontal plane, the pelvis rotated toward the un-loaded limb while the head and trunk moved in theopposite direction. Maximal displacements were foundduring toes-down diagonal tilts (�45�) in the sagittalplane (�5�) and side/diagonal tilts in the frontal andhorizontal planes (�3–8�).

These observations were quantified by computing theprincipal components of the angles for each plane. In allthree planes, for all subjects in all perturbation direc-tions, the first two eigenvectors describe more than 95%of the variance. In most cases, that number is closer to98%. Figs. 4b and 5b show the variable ‘‘loadings’’ forthe first two eigenvectors for each perturbation directionin each plane, and Figs. 4c and 5c illustrate averagescores for a left diagonal front leg lift or a left diagonaltoes-up surface tilt.

Regardless of perturbation direction, the first princi-pal component of the leg lifting data began on or slightlybefore movement onset and leveled around the start ofthe maintenance phase (Fig. 4c). Angular displacementin both the sagittal and frontal planes was dominated bypelvis rotation. In the sagittal plane, front diagonal leglifts resulted in backwards rotations while the pelvisrotated forward during back diagonal and lateral leglifts. The pelvis rotated up in conjunction with the liftingleg in the frontal plane. Back diagonal leg lifts (�45�)also involved some trunk displacement in the samedirection as the pelvis in the frontal plane and inopposition in the sagittal plane (Fig. 4b). The secondprincipal component began after the first componentand generally ended by the onset of the maintenancephase. In the sagittal and frontal planes, the trunk was

Fig. 4 Voluntary leg lifts. a Three-dimensional plot of head, trunk,and pelvis displacement for the sagittal, frontal, and horizontalplanes. Data are from a left diagonal front leg lift (45�) from asingle representative subject. The gray diamond indicates thebeginning of displacement. The data have been fit by a planedescribed by the first two eigenvectors. More than 98% of thevariance was described by the planes. b First and secondeigenvector loadings for the sagittal, frontal, and horizontal planes.Sign (+/-) indicates displacement direction.White circles representthe pelvis, dark gray rectangles the trunk, and light gray trianglesthe head. Bars indicate standard error. c Average scores andstandard errors for the first and second principal components fromall nine subjects in the left diagonal front leg lift direction (45�). Theblack line surrounded by dark gray represents the first score; theblack line surrounded by hatching indicates the second

c

158

responsible for most of the second eigenvector. Thedisplacement pattern was similar to that of the pelvis inthe first component. The head was relatively stable in thesagittal plane and accompanied the trunk in the frontalplane. Analysis of peak-to-peak displacements foundthat trunk displacement patterns with respect to per-turbation direction were similar to those of the pelvis butwere one-third the magnitude.

In response to platform tilting, pelvis displacementwas small in the sagittal plane, with overall head

movement greater in all directions. Examination of thefirst two principal components revealed a loading pat-tern in the frontal plane similar to that found during thevoluntary task (Fig. 5b). The first component was pre-dominately pelvis rotation in agreement with the plat-form rotation, while the second component wascomprised of trunk and head rotation in the samedirection. Timing of the principal components variedbetween the three angular planes. In the frontal plane,component one began near the start of correction phase

159

I, and component two began around the onset of cor-rection phase II (Fig. 5c). However, angular responsewas delayed in the sagittal and horizontal planes. Per-haps the most striking difference between the leg lift taskand surface tilts can be seen in the horizontal plane.During surface tilts, the first eigenvector shows that thehead trunk and pelvis all rotated toward the unloadedlimb (raised edge). The second component involvedreversal of the pelvis (Fig. 5b).

EMG characteristics

Figure 6a illustrates the EMG activation patterns foreight right-sided muscles from a single subject duringfront diagonal leg lifting (same trial as in Figs. 4a, c).The TA was the only muscle to show any statisticaldifferences due to leg lift direction (F[9,54]=3.22,P<0.05). Further analysis confirmed that the differencein latency (�50 ms) was a result of loading versusunloading of the limb. For all perturbation directions,the anticipatory phase was exemplified by an earlyactivation of the shank muscles (MG and TA). The TAwas recruited on the loaded leg at approximately 240 msbefore movement onset, while the MG on the unloadedleg was recruited at about the same time (�245 ms). Theearly shank muscle activation corresponded to the firstlateral COP shift. In the unloaded leg, MG activationwas followed 50–60 ms later by the AD and TA. RFrecruitment began after another 30 ms, followed by theST and TFL at 125 ms before movement onset. Whenthe leg was to be loaded, 100 ms after anticipatory TArecruitment, the AD and RF were activated. MG, ST,and TFL were not consistently recruited. Fewer thanhalf of the subjects, often only two, had anticipatoryactivation of the trunk muscles. Of those who did, ESlatencies were close to movement onset in the back leglift directions (�30 ms), whereas the RA was activatedconsiderably earlier (�183 ms).

The amplitude of the EMG response was evaluatedby calculating the integrated area during the specifiedtime epochs (see Methods). Data were normalized tothe maximal activation for each muscle independent ofperturbation direction or task for each subject. Polarplot representations of the average integrated EMGfrom all nine subjects for the leg lifting task are shownin Fig. 6b. The spatial patterns for the shank and thighmuscles were reasonably consistent across subjects, asillustrated by the direction of maximum activation foreach subject (note placement of small gray and blackcircles in Fig. 6b). For the shank, during the anticipa-tory period the TA was maximally active on the stanceleg during back diagonal leg lifts (�45�). In contrast,the MG was maximally active when the leg was liftedto the side or diagonal front (0� and 45�). Typically,subjects ‘‘kicked’’ rather than ‘‘lifted’’ their legs in theforward lift directions, as was confirmed by an antici-patory plantar flexion of the foot and flexion of thelifting limb knee. During both the movement and

maintenance phases, the lift direction of maximumactivation for the MG switched from the unloaded tothe loaded limb. For the thigh muscles, the magnitudeof EMG activity was relatively small during theanticipatory phase. The RF was maximally active be-fore lifts to the back (�45�), whereas the ST wasmaximally active before forward lifts (45�). During themovement and maintenance phases, the RF was usedto control balance at the support limb during side andback diagonal leg lifts (0� and �45�). The lift directionof maximum activation for the ST during the move-ment and maintenance phases was diagonal back onthe unloaded limb for all subjects. The ST was alsoactivated, probably for balance control, during forwardlifts (45�). Hip and trunk muscle activity spatial pat-terns were more variable across subjects. The RAshowed directional specificity only during the antici-patory phase, in which the maximal EMG amplitudewas found on the loaded side prior to diagonal backleg lifts (�45�). During the anticipatory phase, no ESactivation pattern was discernable. However, as ex-pected, leg lifting to the back led to maximum EMGamplitudes during movement and position maintenancephases on the unloaded side (�45�).

The EMG activation patterns elicited by a diagonaltoes-up surface tilt for eight right-sided muscles from asingle subject are shown in Fig. 7a (same trial as inFigs. 5a, c). Latencies of only three muscles demon-strated sensitivity to perturbation direction: TA –F(7,49)=3.91, P<0.002; MG—F(7,42)=3.12, P<0.05;RF—F(7,49)=4.27, P<0.001. Both the MG and RFwere found to have significantly different latencies fordiagonal toes-up tilts (45�) versus diagonal toes-downtilts (�45�). However, muscle recruitment generallyfollowed a distal-to-proximal strategy. The TA wasactivated first between 100 ms and 120 ms after plat-form onset. In the 0� and 45� directions (i.e., side andtoes-up), TA recruitment was accompanied by coacti-vation of the MG. Approximately 20 ms later, both theRF and AD were activated and were followed 20–40 ms later by ST, TFL, RA, and ES. Toes-downrotations (�45�) began with early recruitment of theTA at 105 ms followed by coactivation of the RF, AD,RA, and TFL an average of 20 ms later. The first MG,ST, and ES bursts occurred approximately 175 ms afterplatform onset.

Figure 7b shows the polar plot representations of theaverage integrated EMG from multidirectional surfacetilts for all nine subjects. Again, the data have beennormalized to the maximal activation for each muscleregardless of perturbation direction or task. Duringcorrection phase I (70–195 ms), the shank muscles dis-played sensitivity to perturbation direction. That is, theTA demonstrated larger activation amplitudes followingtoes-up/unloading tilts (0R� and 45�). As for the MG,the maximum amplitude for half of the subjects was inaccordance with the TA, while the maximum activationdirections for the other four subjects were in response totoes-down/loading tilts (0L� and �45�). As time

160

advanced, the direction of maximum activity for the TAbecame more focused in the pure toes-up direction.After the initial correction phase, the MG behaved in

tandem with the TA in that the direction of maximaloutput was directly opposite (0L� and �45�). The thighmuscles, RF and ST, as well as the TFL and ES were lesswell tuned. However, during both the secondary cor-rection phase and the stabilizing phase, the thigh mus-cles acted in a fashion similar to that of the shankmuscles. RF amplitude was greater following loadeddiagonal toes-up and lateral perturbations (45� and 0L�)as opposed to unloaded diagonal toes-down and lateralperturbations (0� and �45L�) for the ST. The TFLresponse was variable across subjects, while the AD

Fig. 5 Unexpected surface tilts. a Three-dimensional plot of thehead, trunk, and pelvis displacement for the sagittal, frontal, andhorizontal planes. Data are from a left diagonal toes-up surface tilt(45�) from the same subject as in Fig. 4a. Gray diamond indicatesbeginning of displacement. Conventions as in Fig. 4a. b First andsecond eigenvector loadings for the sagittal, frontal, and horizontalplanes for all perturbation directions. c Average scores andstandard errors for the first and second principal componentsfrom all nine subjects in the left diagonal toes-up tilt direction (45�)

161

output was greatest following lateral unloading (0�) forboth correction phase I and correction phase II. Thetrunk muscle, RA, demonstrated directional specificity

with regard to activation amplitude only during cor-rection phase I. As expected, it was maximally activefollowing toes-down rotations.

Discussion

Our results confirm that when stance is disrupted by asudden unloading of one foot, minimization of COMdisplacement is paramount for maintaining equilibrium.Minimization of COM displacement is achieved pri-marily through active control of trunk orientation.When the trunk is adequately controlled, the appropri-ate use of the COP under the feet contributes to therestoration of balance during weight shifting as inducedby leg lifting or surface tilting.

Fig. 6 Muscle activation during voluntary leg lifts. a ExampleEMG data from shank (dark TA, light MG), thigh (dark RF, lightST), hip (dark TFL, light AD), and trunk (dark RA, light ES)muscles during a diagonal front leg lift (45�) from a typical subject.Left column contains data from the loaded limb, and right columncontains data from the unloaded limb. b Muscle tuning curvesrepresenting integrated EMG over anticipatory and movementphases. Maintenance phase is similar to the movement phase. Datahave been normalized to lift cycles and maximum activation asdescribed in the Methods section. Small black and gray circles showdirection of maximum activation for each subject. Outside circlerepresents 30,000 U. Note: Left half of curve (light gray) containsunloaded limb data, and right half of curve (white) shows data fromthe loaded limb

Loaded side

Anticipatoryphase

Movementphase

Maintenancephase

Unloaded side(a)

R Shank

R Thigh

R Hip

R Trunk

TA

MG

RF

ST

AD

TFL

RA

ES

TA

MG

RF

ST

AD

TFL

RA

ES

500 ms

Anticipatoryphase

Movementphase

(b) Shank Thigh Hip Trunk

unloadingR foot

loading45°45°

0° 0°

RFST

ADTFL

RAES

TAMG

Anticipatoryphase

Movementphase

Maintenancephase

162

Control of postural equilibrium through minimizationof head and trunk motion

Our data support the concept that the CNS activelycontrols displacement of the trunk during voluntary leglifting or in response to unexpected surface tilting. Thefunctionality of the displacement is dependent both

upon perturbation source and direction. Principal com-ponent analysis revealed the important relationship be-tween pelvis, trunk, and head excursions in themaintenance of balance. Unexpected surface tiltingelicits an automatic response strategy that focuses oncontrol of trunk orientation, and subsequently the head,with respect to the vertical gravity vector through pelvisrotation. On the other hand, during fast leg lifting, trunkverticality is compromised for movement generation andthe recovery of postural equilibrium while the head re-mains stabilized in space.

In general, during multidirectional surface tilting, thehead and trunk behaved as a single ‘‘locked’’ unit. Evi-dence of this linkage is clearly demonstrated by the closerelationship between the head and trunk as described bythe first and second principal components (Fig. 5b).

Stabilizingphase

Loaded side

Correction Stabilizingphase

Unloaded side(a)

Correctionphase I phase II phase I phase II

R Shank

R Thigh

R Hip

R Trunk

TA

MG

RF

ST

AD

TFL

RA

ES

TA

MG

RF

ST

AD

TFL

RA

ES

100 ms

Correctionphase I

Correctionphase II

(b) Shank Thigh Hip Trunk

unloadingR foot

loading45°45°

0°0°

RFST

ADTFL

RAES

TAMG

Fig. 7 Muscle activation during unexpected surface tilts. a Exam-ple EMG data from shank, thigh, hip, and trunk muscles during adiagonal toes-up surface tilt from the same subject as in Fig. 6a.For conventions, refer to Fig. 6a. b Muscle-tuning curves repre-senting integrated EMG over correction phase I (70–195 ms) andcorrection phase II (195–320 ms). Stabilizing phase is similar tocorrection phase II. Small black and gray circles show direction ofmaximum activation for each subject. Outside circle represents3,000 U, one-tenth the magnitude of voluntary leg lifts

163

These findings are consistent with the concept of areduction in system redundancy. Another advantage inlocking the head to the trunk is that through joint head-trunk displacement, vestibular receptors can be used toestimate trunk motion and regulate its position withrespect to vertical (Mergner et al. 1993). Our results wereaffected by perturbation direction only with regard tosagittal plane displacements. Surface tilts involving apitch-down component produced a backwards dis-placement of the trunk followed by displacement of thehead some 30–50 ms later. This may be an effort bythe CNS to stabilize the head position with respect to theearth’s horizon, thus providing a reference frame forpostural stability. These results are in keeping with pastresearch that demonstrated that when stance is com-promised by an unstable surface such that somatosen-sory information from the feet is less reliable, changes inhead and trunk orientation are smaller than when thesupport surface is stable (Nashner et al. 1988; Pozzoet al. 1992). We believe that this constitutes a controlstrategy whereby the CNS minimizes COM displace-ment through controlling head and trunk orientation.

Further evidence of stabilization of the head in spacewas found during voluntary leg lifting. In all lift direc-tions, head movement was minimized (�5�), whereas thetrunk was inclined between 5� and 15� away from thelifting limb. Our findings are similar to those reported ina study by Mouchnino et al. (1992) in which both naı̈vesubjects and trained dancers maintained a vertical headposition during lateral leg lifting. The naı̈ve subjects alsoinclined their trunks away from the moving leg while thedancers maintained a vertical trunk position. It wasproposed that the trunk displacement was required formovement generation as well as postural stability. Byinclining the trunk away from the moving limb, COMdisplacement is minimized in the same way as a boardcan be balanced on a fulcrum. Our subjects also begandisplacing the trunk on or after the focal movement forlifts in the front diagonal and lateral directions. How-ever, during back diagonal lifts, the onset of trunk dis-placement preceded the leg lift and subsequent pelvisrotation by as much as 75 ms. We suggest that becauseof the biomechanical constraints inherent in the direc-tionality of the ankle and knee joints, the anticipatorytrunk movement is a strategy used by the CNS to gen-erate thrust force for the upcoming back lift.

Control of postural equilibrium through COPadjustment

One means by which the CNS can counter undesiredCOM displacement, given that the trunk is adequatelycontrolled, is through adjustments made at the feet-floorinterface. Roll et al. (2002) and Kavounoudias et al.(2001, 1998) have shown that the perception of bodyposture and orientation are highly dependent on cuta-neous afferents in the feet. In addition, mechanorecep-tors in the feet are known to provide important

information concerning both perturbation velocity (Di-ener and Dichgans 1988) and direction (Macpherson1994), as illustrated by EMG activity. During self-initi-ated movement, it has been suggested that end-positionbalance rather than body orientation has been prepro-grammed by the CNS (Do et al. 1991).

If postural equilibrium takes priority over segmentalorientation, then the interface of the feet with the sup-port surface becomes a significant source of balancecontrol. The predominance of pelvis rotation in the firstprincipal component combined with the temporal rela-tionship between COM and COP during voluntary leglifts demonstrated the use of advanced changes in COPposition to control COM displacement. Additionally,once the target position was reached (position mainte-nance phase), the COM was encompassed by an oscil-lating COP. Other researchers have reported similarfindings in relation to voluntary movement. Clear evi-dence of the central control of COM positioningthrough COP shifts has been shown in tasks involvingadaptation of the support configuration (Kaminski andSimpkins 2001; Lyon and Day 1997; Mille andMouchnino 1998). Lifting one foot will decrease the baseof support substantially and induce a fall if the COM isnot repositioned over the stance foot. An increase in theground reaction forces under the lifting leg, which servesto propel the COM towards the stance limb, has beenfound to precede leg lifting and is scaled in order toreestablish posture and equilibrium (Lyon and Day1997; Mille and Mouchnino 1998). This anticipatoryresponse is not seen in supine leg lifting (McIlroy et al.1999) or during leg lifting in microgravity (Mouchninoet al. 1996), when there is no interface between the feetand the environment. We found a similar relationshipbetween COM and COP following unexpected surfacetilts. Again, the first principal component was domi-nated by pelvis rotation, which led to displacement ofthe COP. Also, as we saw during leg lifting, the COPcorralled the COM.

Voluntary versus automatic adjustment

Through the comparisons of voluntary versus unex-pected postural responses, we can explore how the CNSintegrates feedforward and feedback information. Asystem whose control mechanisms are adaptable as wellas context-dependent would have more flexibility tocompensate for a variety of instabilities whether theyarose from self-generated or externally imposed move-ments.

We found many similarities in the postural responsestrategies following voluntary leg lifting versus unex-pected surface tilting. Perhaps the most significant refersto the reduction of joint segment redundancy as illus-trated by the head, trunk, and pelvis coordination.During lateral and diagonal surface tilts, as well as frontand front diagonal leg lifts, pelvis rotation, in oppositionto head or trunk movement, dominates the postural

164

response. However, during back diagonal and side leglifts, the trunk and pelvis move in harmony as if thebody were an inverted pendulum. The externally rotatedfoot position that our protocol required favored medial-lateral stabilization by increasing the width of the baseof support, while hip control was still apparent. Wefound the AD muscle maximally activated during limbunloading and the TFL tuned with limb loading. This isnot unlike the pairing of the AD in the unloaded limbwith the TFL in the loaded limb during voluntary legflexion (Rogers and Pai 1995). However, during multi-directional leg lifting, we found cocontraction of the ADand TFL in all lift directions. This discrepancy may bedue to different task goals and biomechanical con-straints (that is, stiffening the pelvis to accommodate thelong lever arm and to control medial-lateral stability). Inaddition, most of our subjects relied on flexion of thesupport limb knee to lower their COM and enable betteruse of the thigh and hip muscles. Also, the use of thearms as a balance tool undoubtedly reduced the mag-nitude of the postural responses.

In conclusion, our study of postural response strate-gies during multidirectional leg lifts and unexpectedsurface tilts found that balance corrections were sensi-tive to the perturbation direction. Furthermore, we haveprovided evidence of redundancy reduction throughCNS recruitment of a specific kinematic strategy tooppose postural instability. We suggest that, at leastduring weight shifting as induced by voluntary leg liftingand unexpected surface tilting, the CNS dynamicallycontrols COM displacement through trunk stabilization.

Acknowledgements We would like to thank Rumpa Boonsinsukhand Margerita Rapagna for their assistance with data collection,Dr. Patricia McKinley for her helpful suggestions with the manu-script, and Dr. David Burns for his invaluable contribution to theuse of PCA. This study was supported by the Canada Foundationfor Innovation, CRIR, and the JRH Foundation.

References

Allum JH, Honegger F (1998) Interactions between vestibular andproprioceptive inputs triggering and modulating human bal-ance-correcting responses differ across muscles. Exp Brain Res121:478–494

Allum JH, Gresty M, Keshner E, Shupert C (1997) The control ofhead movements during human balance corrections. J VestibRes 7:189–218

Andersen RA, Snyder LH, Li CS, Stricanne B (1993) Coordinatetransformations in the representation of spatial information.Curr Opin Neurobiol 3:171–176

Assaiante C, Amblard B (1993) Ontogenesis of head stabilization inspace during locomotion in children: influence of visual cues.Exp Brain Res 93:499–515

Assiante C, McKinley PA, Amblard B (1997) Head-trunk coordi-nation during hops using one or two feet in children and adults.J Vestib Res 7:145–160

Baroni G, Ferrigno G, Rabuffetti M, Pedotti A, Massion J (1999)Long-term adaptation of postural control in microgravity. ExpBrain Res 128:410–416

Bianchi L, Angelini D, Orani GP, Lacquaniti F (1998) Kinematiccoordination in human gait: relation to mechanical energy cost.J Neurophysiol 79:2155–2170

Borghese NA, Bianchi L, Lacquaniti F (1996) Kinematic deter-minants of human locomotion. J Physiol 494:863–879

Clement G, Pozzo T, Berthoz A (1988) Contribution of eye posi-tioning to control of the upside-down standing posture. ExpBrain Res 73:569–576

Diener H-C, Dichgans J (1988) On the role of vestibular, visual andsomatosensory information for dynamic postural control inhumans. Prog Brain Res 76:253–262

Do MC, Nouillot P, Bouisset S (1991) Is balance or posture at theend of a voluntary movement programmed? Neurosci Lett130:9–11

Fung J, Johnstone E (1998) Lost in space: multi-axial and multi-dimensional surface perturbations delivered by a novel motionbase device. Soc Neurosci Abs 24:1158

Grasso R, Zago M, Lacquaniti F (2000) Interactions between pos-ture and locomotion: motor patterns in humans walking withbent posture versus erect posture. J Neurophysiol 83:288–300

Henry SM, Fung J, Horak F (1998) Control of stance during lat-eral and anterior/posterior surface translations. IEEE TransRehab Eng 6:32–42

Hlavacka F, Mergner T, Krizkova M (1996) Control of the bodyvertical by vestibular and proprioceptive inputs. Brain Res Bull40:431–434

Hodges P, Cresswell A, Thortensson A (1999) Preparatory trunkmotion accompanies rapid upper limb movement. Exp BrainRes 124:69–79

Kaminski TR, Simpkins S (2001) The effects of stance configura-tion and target distance on reaching I. Movement preparation.Exp Brain Res 136:439–446

Kavounoudias A, Roll R, Roll J-P (1998) The plantar sole is a‘‘dynamometric map’’ for human balance control. Neuroreport9:3247–3252

Kavounoudias A, Roll R, Roll J-P (2001) Foot sole and anklemuscle inputs contribute jointly to human erect posture regu-lation. J Physiol 532:869–878

Lyon IN, Day B (1997) Control of frontal plane body motion inhuman stepping. Exp Brain Res 115:488–507

Macpherson JM (1994) The force constraint strategy for stance isindependent of prior experience. Exp Brain Res 101:397–405

Massion J (1992) Movement, posture and equilibrium: interactionand coordination. Prog Neurobiol 38:35–56

Massion J, Fabre JC, Mouchnino L, Obadia A (1995) Body ori-entation and regulation of the center of gravity during move-ment under water. J Vestib Res 5:211–221

Massion J, Popov K, Fabre JC, Rage P, Gurfinkel V (1997) Is theerect posture in microgravity based on control of trunk orien-tation or center of mass position? Exp Brain Res 114:384–389

Mauer C, Mergner T, Bolha B, Hlavacka F (2000) Vestibular, vi-sual, and somatosensory contributions to human control ofupright stance. Neurosci Lett 281:99–102

McIlroy WE, Bent LR, Potvin JR, Brooke JD, Maki BE (1999)Preparatory balance adjustments precede withdrawal responseto noxious stimulation in standing humans. Neurosci Lett267:197–200

Mergner T, Siebold C, Schweigart G, Becker W (1991) Humanperception of horizontal trunk and head rotation in space duringvestibular and neck stimulation. Exp Brain Res 85:389–404

Mergner T, Hlavacka F, Schweigart G (1993) Interaction of ves-tibular and proprioceptive inputs. J Vestib Res 3:41–57

Merkle LA, Layne CS, Bloomberg JJ, Zhang JJ (1998) Using factoranalysis to identify neuromuscular synergies during treadmillwalking. J Neurosci Methods 82:207–214

Mille ML, Mouchnino L (1998) Are human anticipatory posturaladjustments affected by a modification of the initial position ofthe center of gravity? Neurosci Lett 242:61–64

Mouchnino L, Aurenty R, Massion J, Pedotti A (1992) Coordi-nation between equilibrium and head-trunk orientation duringleg movement: a new strategy built up by training. J Neuro-physiol 67:1587–1598

Mouchnino L, Aurenty R, Massion J, Pedotti A (1993) Is the trunka reference frame for calculating leg position? Neuroreport4:125–127

165

Mouchnino L, Cincera M, Fabre JC, Assaiante C, Amblard B,Pedotti A, Massion J (1996) Is the regulation of the center ofmass maintained during leg movement under microgravityconditions? J Neurophysiol 76:1212–1223

Nashner LM, Shupert CL, Horak FB (1988) Head-trunk move-ment coordination in the standing posture. Prog Brain Res76:243–251

Pozzo T, Berthoz A, Lefort L (1990) Head stabilization duringvarious locomotor tasks in humans. I. Normal subjects. ExpBrain Res 82:97–106

Pozzo T, Levik Y, Berthoz A (1992) Head stabilization in thefrontal plane during complex equilibrium tasks in humans. In:Woollacott M, Horak F (eds) Posture and gait: control mech-anisms. University of Oregon Books, Eugene OR, pp 118–121

Pozzo T, Levik Y, Berthoz A (1995) Head and trunk movements inthe frontal plane during complex dynamic equilibrium tasks inhumans. Exp Brain Res 106:327–338

Preuss R, Fung J (2004) A simple method to estimate force plateinertial components in a moving surface. J Biomech 37:1177-1180

Rogers MW, Pai YC (1990) Dynamic transitions in stance supportaccompanying leg flexion movements in man. Exp Brain Res81:398–402

Rogers MW, Pai YC (1995) Organization of preparatory posturalresponses for the initiation of lateral body motion during goaldirected leg movements. Neurosci Lett 187:99–102

Roll R, Kavounoudias A, Roll J-P (2002) Cutaneous afferent fromhuman plantar sole contribute to body posture awareness.Neuroreport 13:1957–1961

Winter DA (1990) Biomechanics and control of human movement,2nd edn. Wiley, New York

166