human walking along a curved path,ii,gait features and emg patterns

8/10/2019 Human Walking Along a Curved Path,II,Gait Features and EMG Patterns

1/15

Human walking along a curved path. II. Gait features andEMG patterns

Gregoire Courtine1,2 and Marco Schieppati2

1INSERM Motricite & Plasticite, University of Burgundy, Dijon, France2Section of Physiology, Department of Experimental Medicine, University of Pavia, and Human Movement Laboratory (CSAM),

Fondazione Salvatore Maugeri (IRCCS), Scientific Institute of Pavia, Italy

Abstract

We recorded basic gait features and associated patterns of leg muscle activity, occurring during continuous body progression when

humanswalked along a curved trajectory, in order to gain insight into the nervous mechanisms underlying the control of the asymmetric

movements of the two legs. The same rhythm was propagated to both legs, in spite of inner and outer strides diverging in

length(P


2/15

the intrinsic features of spinal networks and the force acting during the

turn. In this paper, we attempted to examine whether locomotor

patterns are fundamentally changed or only finely tuned in order to

exploit a basic functioning of the spinal locomotor system. To this aim,

we investigated basic gait features involved in the production of a

continuous body trajectory along a curve as well as associated patterns

of leg muscle activity. We provide the evidence that commands to walk

along a curved path may exploit the basic mechanisms of the spinal

locomotor generator, thereby limiting the computational cost of

turning.

Materials and methods

Participants

Six healthy male adults volunteered for this experiment (see compa-

nion paper for further subjectsdescription).

Locomotor task

The locomotor task has been described in the companion paper.

Briefly, subjects executed walking trials along two locomotor paths:

straight and curved. The curved path shared the initial 3 m of the

straight path, then followed a 4.6-m curve to the right, ending with a

further 2 m of straight walking. The radius of curvature was constant

(120 cm), thus resulting in an overall change of direction of 2208(see

Fig. 3). Subjects walked barefoot with the arms folded across the chest.

They made 10 repetitions with eyes open (EO) and 10 repetitions

blindfolded (BF) in each walking condition.

Data acquisition

General procedures have also been described in detail in the compa-

nion paper. Both kinematic and electromyographic (EMG) data were

obtained using the integrated ELITE (BTS, Italy) system. Bipolar

surface electrodes (1 cm diameter, electrode separation of 1 cm) were

placed over three muscular groups of each leg: tibialis anterior (belly),

soleus (2 cm below the insertion of the gastrocnemii) and rectus

femoris (belly). A ground electrode was placed on the wrist. In

addition, for three of the six participants, the activity of the peroneuslongus muscle was recorded from both legs. Signals were preamplified

(100), digitized, and transmitted to the remote amplifier via tele-metry. Signals were sampled at 1000 Hz, band-pass filtered (10

500Hz) and full-wave rectified. The resulting EMG signal was

normalized with respect to mean EMG activity measured during a

window of 500 ms selected in the middle of 3 s of maximum voluntary

contraction. Normalized waveforms were smoothed by applying a

moving average of 50 ms width.

Data processing

Kinematic sampling and EMG data recordings were synchronized at

rates of 100 and 1000 Hz, respectively. Data from each trial were

ensemble-averaged after time interpolation over individual gait cycles

to fit a normalized 1000-point time base.

Elaboration of kinematics data

Co-ordinate frames

Co-ordinate frames that describe body motion while moving through

space were defined in a hierarchical manner as described by Imai et al.

(2001). The primary co-ordinate frame was the space-fixed reference

frame of the video system (XS, YS, ZS). Instantaneous heading direction

was defined as theangle of theinstantaneouslinear velocity vector of the

body mid-point (see the companion paper for details) in the horizontal

plane with respect to the space-fixed reference frame. The amplitude of

the locomotor direction change during one gait cycle was calculated as

the difference in heading between two successive heel strikes of one leg.

Subsequently, the co-ordinates of the different markers were converted

in the body-centred reference frame, whose Xaxis matched the heading

direction, through co-ordinate transformation. Following this, angles of

trunk, foot and limb axis segments were computed.

Determination of gait events

The stride cycle was considered the interval between two successive

heel strikes of one leg. In this paper we considered for further analysisonly gait cycles whose heading change was >408. Indeed, whensubjects walked along curved paths, heading change was generally

>408, whereas smaller angles corresponded to the transition fromstraight to curved trajectories. Selected cycles were then expected to

provide more consistent information concerning turn-related gait

organization (in some figures, cycles associated with intermediate

heading changes arealso included, and additions areindicated). Onsets

of swing phases were based on rates of change of foot vertical

translations. A threshold of foot clearance was set at 10% of maximal

peak velocity of foot vertical displacement during the swing phase.

Gait parameters

Gait cycle duration was taken as the time interval between twosuccessive heel strikes of one leg. Stance and swing duration was

computed and converted to percentages of total cycle duration. For

each leg, stride length was measured as the linear translation of the

malleolus between two successive heel strikes. The body displacement

was also computed as the length of the body mid-point path during the

entire time-interval of the cycle:

Body path=cycle Xn1i1

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffixi1 xi

2 yi1 yi2 zi1 zi

2

q

wheren is the number of acquired frames during the gait cycle. The

average velocity of body displacement was computed as the body path

length divided by the duration of the gait cycle. As turning affected the

walking speed (see companion paper), a speed-independent index(Sekiya & Nagasaki, 1998) was used in order to assess the relationship

between gait parameters across the two walking directions and left and

right legs. Awalking ratio was computed as the stride length divided by

the step frequency.

Limb displacements

Inter-limb coordination was calculated as the phase difference (DF)

between angular displacements of left and right limb axes:

DF 360

Dt=T

where T is the duration of the inner (right) limb cycle and Dtis the

difference between the time at which the outer (left) limb reaches its

angular peak (heel strike) and the time halfway between the two

successive angular peaks (heel strikes) of the inner limb (see inset of

Fig. 4). Given this definition, if the left and right limbs move 1808out

of phase, DF would be equal to zero. In turn, when DF is positive,

contralateral heel strike occurred beyond half of the cycle, thus

indicating a phase lag of the outer with respect to the inner limb.

Mean velocity of limb endpoint displacements were evaluated as

average velocities of malleolus displacements during the swing phase.

Elaboration of EMG data

Temporal EMG features

Onsets and ends of EMG burst activity were established at the points at

which muscle activity was, respectively, greater or less than mean

192 G. Courtine and M. Schieppati

2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 18, 191205


3/15

activity plus 1.5 SD recorded during a period when the muscles were

least active(swing phase for the soleus and from 25 to 50% of cycle for

tibialis anterior and 3070% of cycle for rectus femoris). Soleus and

rectus femoris muscles presented a single burst during the gait cycle,

whereas tibialis anterior showed a well-identified double burst. The

latter was divided into two distinct bursts whose separation was set as

the trough occurring around the centre of the double burst. The burst

determination was made automatically (see above) but noise-induced

errors were corrected when necessary by means of interactive custom-

made software. This was done by redefining a temporal window forcomputing EMG background during which noise was absent. The

activity of each muscle during stance and swing phases as well as

during the time interval of the burst was calculated as the integral of the

muscle envelope.

Inter-limb EMG timing

The timing of EMG activation between homonymous muscles of the

two legs during turning was assessed by means of cross-correlation of

averaged EMG waveforms. Averaged waveforms were first normal-

ized by expressing the mean waveform during the entire cycle (T) with

respect to the mean value of the EMG activity during SA, for each

muscle, side and subject individually. Then, the grand averages of

EMG waveforms were computed for each leg and muscle. Finally, thecross-correlation algorithm was applied to the waveform profiles.

Statistical analysis

Means and SD, for each subject in each walking condition (straight or

curved) and for all parameters described above, were calculated. All

the mean values computed during gait cycles were submitted to a

2 (straight-ahead or turning) 2 (limb side) 2 (vision) analysis ofvariance for repeated measures (within-subject ANOVA). Differences

between variables related to the whole trajectory were evaluated using

a 2 (straight or curved path) 2 (normal vision or blindfolded) ANOVAfor repeated measures.Post hocdifferences were assessed by means of

the NewmanKeuls test. In addition, Students t-test and ANOVA/

ANCOVA were, respectively, used to assess the differences between

slopes or intercepts of linear regressions. The software packageStatistica1 was used.

Results

The analysis of the results has been deliberately focused here on the

spatio-temporal gait and muscular events occurring after the comple-

tion of the transition between straight-ahead and turning, i.e. when the

turning behaviour had reached a plateau (see Materials and methods).

A total of 337 and 428 gait cycles where analysed during straight-

ahead and turning, respectively. Subjects were similarly represented,

because each contributed a mean of 56 7 and 71 17 cycles for thestraight and curved paths, respectively, to the analysis.

Gait features

Spatial and temporal features of the gait pattern of all subjects across

all trials during both straight-ahead walking (SA) and turning (TU),

both with eyes open (EO) and blindfolded (BF), are detailed in

Fig. 1AC.

During straight-ahead locomotion, cycle duration and stride length

were unchanged between left and right limbs (ANOVA, side effect,

P> 0.8 for all parameters) regardless of visual condition (vision effect,P> 0.7). Duration of cycles was close to 1 s (on the average1.11 0.02 s). Stride length slightly decreased (5.5%) when walkingblindfolded (vision effect, F1,5 9.9, P< 0.05). A significant effectwas found in the walking ratio (length/frequency) (see Materials and

methods) between EO and BF (vision effect, F1,5 13.2, P< 0.05),mainly connected to the decrease in stride length.

During turning, gait cycle duration was predictably unchanged

between left and right limbs (P> 0.3) but also when comparedto straight-ahead (P> 0.01). There was a decrease (18.5%) in thelength of the internal (right) stride compared to the left stride

(direction side-effect,F1,5 204, P < 0.0001; post hoc comparisonP< 0.001). The latter did not change its length with respect to straight-ahead (post hoccomparison,P> 0.2). When turning was executed BF,

stride length was decreased to a similar extent for both limbs (5.6%).This effect has the same size as that observed under the straight-ahead

condition. As a consequence of the limb-specific change in stride

length but not in stride duration, the walking ratio was markedly

diminished (18.3%) for the right leg cycles (side direction,F1,5 286, P< 0.0001; post hoc, P< 0.0005). The walking ratioof the left leg cycles was unaffected during turning compared

to the straight-ahead conditions (post hoc comparison, P> 0.4).Without vision, the walking ratio slightly diminished (4.9%) for both

trajectories of walking and for both limbs (directionvision ordirectionvision side showed no significant effect; P > 0.4).

The histograms in Fig. 1C summarize spatio-temporal features of

body displacements during both straight-ahead and turning conditions.

Mean body velocity significantly decreased when walking alongcurved as compared to straight paths (by 15.4% on average; see also

companion paper) (direction effect,F1,5 15,P 0.01). Blindfoldingresulted in a decrease (5.7%) in mean body velocity during both

straight-ahead and curved walking. The distance of body displacement

along its own trajectory during one cycle (body path) was calculated as

the total length travelled by the body mid-point during the time-

interval of the cycle. The body paths calculated during the left and right

cycles were averaged together for each walking condition because the

duration of left and right cycles was similar within each condition. The

result was that walking along the curved path induced a moderate but

significant decrease in the distance covered by the body during one

cycle (16.3%) (direction effect, F1,5 26, P < 0.005).

Stride length to body velocity relationshipDuring normal walking, a well-known monotonic relationship exists

between stride length and body velocity. This was reproduced in all

subjects, for straight-ahead walking, as shown in the plot of Fig. 2A.

No differences were found between left and right sides. However, for

turning, the two scatter plots for left and right leg were separated from

each other (plot of Fig. 2B) despite there being no significant changes

in the slopes of the lines of best fit. This is also indicated by the equal

slope values (grand means from each subject) reported in Fig. 2C

(side or direction effect, P> 0.2). Figure2D shows that the meanintercept of the line fitting the data points for the right limb during

turning diverged significantly from the other intercepts (ANOVA,

direction side effect, F1,5 50, P< 0.001; P< 0.001 for all right-turning intercepts vs. other intercepts, post hoc comparison).

Spatial organization

From the typical body displacements along the straight and curved

path depicted in Fig. 3A, one can see that the body mid-point trajectory

lay in between the foot prints during straight-ahead walking, but was

much closer to the inner foot during turning (see companion paper

for further details on foot positioning). Positions of foot prints

illustrate the decrease in the stride length of the inner with respect

to the outer limb. The plot of Fig. 3B details these stride length changes

for all subjects. Differences between stride lengths of left and right

limbs and the length of the body path are plotted against the amplitude

of heading change, which corresponds to the tightness of the curvature


Walking along a curved path. EMG 193


4/15

produced during two successive heel strikes of each limb during the

gait cycle. The cluster located around the zero heading change

corresponds to straight-ahead walking. Here, the difference between

the stride length and the body path length is negative because the

body mid-point covers more distance than the foot during the time

interval of the cycle (the body mid-point moves up and down, and

right and left, during normal straight-ahead walking, while stride

length is the spatial distance between two successive heel strikes).

The data point pertaining to curved walking corresponds to the

upper and lower clusters for left and right limb, respectively. Inter-

mediate points connecting straight- and turning-related clusters

correspond to gait cycles executed during turning but at the initial

part of the curve, when the heading change is less pronounced. The

increase in distance shown for the left limb does not result from a true

increase in stride length but depends upon the decreased distance

travelled by the body mid-point during curved cycles. In turn, the

distance travelled by the body mid-point decreases because of a shift of

the body toward the inner foot. The top right diagram attempts to

provide a graphic explanation of this phenomenon. The two feet cover

imaginary paths, their distance depending on the tightness of the

trajectory curvature; the body path (grey dashed line) tends to approach

the inner foot path, thereby loading and unloading the inner and outer

foot, respectively.

Temporal gait features

Turning not only produced opposite effects in the spatial features of the

gait patterns of left and right limbs but was also accompanied by clear-

cut differences in some temporal features of walking. The duration of

the stance phase is indicated in the histogram of Fig. 4A. Duration

decreased when the supporting foot was the left, outer foot, and

increased when the supporting foot was the right, inner one (side -

direction effect, F1,5 44, P 0.001). Compared to straight-aheadcycles, stance duration of left and right limb significantly decreased

(post hoc, P < 0.01) and increased (post hoc, P 0.01), respectively,during curved walking. As a consequence, the mean double stance

duration was hardly affected during turning (9.9 1.3 and 9.6 1.7%of the cycle for SA and TU, respectively; data pooled for the two

limbs). Without vision, the duration of stance significantly increased

(4.5%) for both left and right legs (vision direction effect, F1,5 9,P< 0.05) (post hoc, P< 0.005 for EOBF comparison during turning).

A slight but significant change in the phase lag between the periodic

oscillations of outer and inner limbs was a regular feature of the

progression along the curved path. This was assessed as the phase lag

between left and right limb angular displacements (limb axis), the

positive peak value of which corresponds to the heel strike. As

expected, the calculation of this phase lag gave a value very close

Fig. 1. (A) Mean duration (upper histogram) and mean stride length (lower histogram) of gait cycles performed when walking along straight or curved paths,with eyes open (EO) or blindfolded (BF). (B) Mean values of the walking ratio (stride length divided by step frequency). The drop in the stride length of theright (inner) leg during turning is responsible for the decrease of its walking ratio. (C) Mean velocity (left) and mean path length (right) of the body mid-point.

Mean values from gait cycles of both legs have been averaged because the body motion features is side-independent. Velocity and length of body displacementswere smaller during curved compared to straight-ahead walking. Horizontal lines with an asterisk join conditions between which significant differences(P


5/15

to zero during straight-ahead walking. The distribution of these lags

during straight-ahead walking, cycle per cycle across all subjects, is

reported in the white bars of the histogram of Fig. 4C. When turning

was analysed (grey bars), the mean phase lag shifted toward a positive

value (corresponding to 57% of the cycle), indicating that the right

limb led with respect to the left (direction effect, F1,5 40,P 0.0001).

The increased duration of the stance phase of the right, inner, foot

was accompanied by leaning of the trunk towards the inner side of the

walking path (trunk roll). The positive correlation between the increase

in right foot stance duration and the extent of mean trunk roll

(producing increased loading of the inner foot) is shown in the plot

of Fig. 4B. Conversely, left foot stance duration was negatively

correlated with trunk inclination to the contralateral side, a sign of

unloading of the outer limb. In addition, the longer step achieved by the

left leg and the limb girdle rotation during turning significantly

(direction side-effect,F1,5 52, P < 0.001) increased the hip anglerange on the left side (3.0 2.08,P < 0.01), whereas the shorter stepof the right leg decreases its range (3.2 1.88; P < 0.005).

The decreased duration of the stance phase of the outer limb was

probably related to the obligatory overall higher velocity of the left

limb, given the longer path to be travelled with respect to the inner

limb. A further aspect of the asymmetry in the characteristics of right

and left limb displacement was the different velocity of their respective

swing phases. This is shown in the plot of Fig. 4D, where velocity of

the foot swing is depicted for right and left feet as a function of the

velocity of body progression along the curve. Indeed, the intercept of

the best-fit line differed (ANOVA/ANCOVA, P< 0.0001) between the twolimbs, whereas the slope did not (t-test, P > 0.2). An example of thelarger velocity of the outer foot is shown in the Fig. 4E, where two

velocity traces are superimposed for the right (black lines) and left foot

(grey lines) during two successive swing phases taken from the centre

part of the adjacent walking trial (in which the positions of the two

malleoli are drawn every 30 ms).

EMG patterns during turning

Spatial and temporal patterns of thigh and leg muscle EMGs were little

changed between straight-ahead and curved walking for both limbs.

Typical averaged traces from one subject during EO are shown in

Fig. 5. Traces have been normalized in duration on the basis of the

cycle duration and in amplitude with respect to the EMG during

sustained maximum voluntary contraction of the corresponding mus-

Fig. 2. Relationship between stride length and mean body velocity during (A) straight-ahead and (B) curved-gait cycles of the left (open circles) and right (filled

circles) leg. Data from all subjects and trials pooled. Mean values of (C) slope and (D) intercept of individual length-to-speed relationships, independently computedaccording to the direction of walking and side of the body. Change in direction did not modify slopes of speed-to-length relationships but affected the intercept(

P


6/15

cle (see Materials and methods). The shape of the muscular activation

pattern during straight-ahead walking is not fundamentally modified

by walking along a curved path. The soleus muscles of both legs

showed a full-blown burst during the stance phase of gait. The tibialis

anterior muscles exhibited a double-peak burst at the onset of swing,

which lasted up to thefirst third of the subsequent stance phase. The

rectus femoris traces were almost superimposed on the second peak of

the tibialis burst. However, small but systematic changes occurred

between inner and outer limbs in both amplitude of EMG bursts and

their relative phases (temporal lag). The first of these changes was an

increase in the amplitude of the soleus burst during the stance phase in

the left leg and a decrease in the same burst in the right leg. Secondly,

during turning, there was an opposite phase shift in the tibialis EMG

burst between left and right legs. The left was anticipated and the right

was delayed with respect to the same burst in the same leg during the

straight-ahead condition. The activity of the peroneus muscle was

recorded in three subjects only. Nonetheless, its turn-related changes

were consistent across individuals and trials. The amplitude of the

peroneus burst generally decreased during curved walking in both legs.

However, such decreases were weak in the outer leg, but a sizeable

drop in the EMG activity was observed in the inner leg. In addition, the

burst of the right peroneus was delayed during the turn with respect to

the same burst observed during straight-ahead walking (see below).The results of the analysis made on all subjects and trials is

summarized in Fig. 6. The histograms of Fig. 6A and B depict mean

values of normalized EMG activity separated by side, visual and

walking conditions. The normalization was made based on the homon-

ymous burst during straight-ahead walking (data from EO and BF

pooled). Mean values for the soleus burst during stance and for the

tibialis anterior burst during swing are reported in Fig. 6A and B,

respectively. The statistical analysis showed a consistent difference in

EMG activity between left (8.1%;post hoctest,P 0.05) and right(11.1%; P< 0.05) soleus muscles during turning with respect tostraight-ahead walking (side direction effect,F1,5 11.5,P < 0.05).Post hoccomparisons also revealed that the activity increased on the

left side but decreased on the right side (P< 0.05 for left vs. right EMGcomparisons). However, there was a general increase in the tibialisanterior burst in both legs during turning with respect to straight-ahead

(on the average, 16.3% and 12.7% for left and right legs, respectively;

direction effect, F1,5 7, P< 0.05). The visual conditions are notfurther detailed in thefigure, because in no walking condition did lack

of vision modify the EMG features compared to walking EO.

The schematic diagram of Fig. 6C summarizes the main temporal

and amplitude characteristics of bursts of rectus femoris, tibialis and

soleus muscles in the two limbs, during both straight-ahead and curved

walking conditions. Onsets and ends of the boxes are the average

onsets ( SD) and ends ( SD) of muscle EMGs computed from allsubjects and EO trials. The height of theboxes corresponds to the mean

EMG amplitude (SD) normalized with respect to straight-ahead. In

thisfigure, the two components of the tibialis anterior burst have beenanalysed and drawn as two separate but adjacent boxes. This segrega-

tion of the two bursts of the tibialis allowed detecting a further feature

of its activity, namely a turn-specific increase (side direction effect,F1,5 9, P< 0.05) of the first component of the right muscle comparedto the activation during straight-ahead progression (SATU post hoc

comparison for the right limb,P < 0.05). The timing characteristics ofthe bursts of the various muscles can be more fully appreciated in this

representation. First, there was littlevariability of burst onsets and ends

for all subjects and trials, as testified by the short horizontal error bars.

Second, there was a strong consistency in the respective timing

between left limb and right limb, except for the tibialis anterior.

Indeed, the onset of the first component of the tibialis anterior burst

was slightly but significantly advanced for the left limb and delayed for

the right limb during turning with respect to the straight-ahead

conditions.

The consistency of this delay is further evident in the distribution

histograms of Fig. 7. These histograms report the distribution of the

soleus end and of the tibialis onset during the straight-ahead and

curved walking tasks, separately for left and right limbs. The interval

distribution of the end latencies of the soleus is centred around the end

of the stance phase (57% of the cycle). The onset of the tibialis is

delayed by a further 7.5% of the cycle. In addition, the width of the

distribution curve appears to be somewhat larger for the tibialis with

respect to the soleus (see range in the abscissas), regardless of the

walking condition. During turning, the soleus burst end was unaffected

Fig. 3. (A) Typical displacement of the body mid-point during the progressionalong straight or curved paths. Foot prints (the segment joining the malleolusand the foot) of the left and right leg during the stance phases are indicated. Onthe right, imaginary paths along which the body and the legs progress arerepresented (see text); the direction of travel is from bottom up and clockwise.(B) Relationship between the heading change and the difference between stridelength and path travelled by the body during one cycle. In this representation,

the displacement of each leg is referred to the body-centred reference frame.The more the body rotates, the larger the discrepancy between the stride lengthof inner and outer legs. The cycles during thestraight-ahead walking correspondtothe datapointsaround08. Those of thetransition from straight to curved pathshave been added to provide clarity to the figure, and roughly correspond to datapoints between 10 and 408 of heading change.




7/15

whereas the distribution of tibialis onset of left and right limbs shifted

in opposite directions (side direction effect, F1,5 32, P< 0.005).The mean values of these changes are reported in the middle inset of

the same figure. (SATU post hoc comparison, P < 0.01 for left andright legs). An analysis of the temporal coupling of subsequent

activations of the soleus and tibialis anterior was made according to

the trajectory type. However, we found no strong correlation between

the end of the soleus burst and the onset of the tibialis burst, regardless

of the side (r 0.05,ns;r 0.16,P < 0.05 for the left and right limb,respectively), the trajectory followed (r 0.11, P < 0.05; r 0.05, nsfor straight and curved path, respectively), or within left and/or right

limbs during turning (r 0.08, ns; r 0.11, ns for the left and rightlimb, respectively).

A general comparison of the effects of turning across left and right

soleus, tibialis anterior and peroneus longus muscles is given in Fig. 8.

Traces were obtained by averaging all signals from all subjects during

Fig. 4. (A) Mean duration (SD) of the stance phase as a percentage of the gait cycle according to walking direction and body side. During turning, the stanceduration increased and decreased in inner and outer legs, respectively (P< 0.01). (B) Relationship between the relative duration of the stance phase and mean trunkroll, for all gait cycles under all conditions (except for side) pooled. The decrease in stance duration of the left (outer) leg during turning was associated with a

significant leaning of the trunk toward the interior, and vice versa for the right leg. (C) Distribution of the phase lags of the left with respect to the right leg. Thedistribution shifted toward positive values during curved walking, indicating that the inner limb led with respect to the outer. The method of calculation of the phaselag based on limb axis angulardisplacements of both legs is indicatedin the right inset. (D). Therelationship between mean malleolus velocity during theswing phaseand mean body velocity is shown. The velocity of the foot is systematically larger in the outer limb than in the inner limb owing to the larger distance to be covered bythe outer foot during the swing phase. (E) An example of the time course of the velocity of the malleolus of both legs is depicted. A simple illustration of the side-

specific modulation of the limbvelocity during the turn is revealed by the representation at the centre of the panel. The instantaneous position of the malleolus of bothinner and outer leg is displayed every 30 ms. The body mid-point trajectory is also shown. Note the increasing distance between two successive positions of the outerfoot in the central part of the swing phase.




8/15

turning and normalizing them with respect to the mean activation during

straight-ahead walking, each limb individually processed. Figure 8

summarizes EMG changes in timing and amplitude between the corre-

sponding muscles of inner and outer limbs during the progression along

the curved path. The shape of the envelopes, reflecting the activation

profile of muscles, was basically unchanged, as indicated by the very

high value obtained by cross-correlating the two profiles. No phase

shift was observed between right and left soleus muscles. However,

a difference was found in the timing of tibialis and peroneus

muscle bursts. The waveform profiles of the left leg led those of the

right leg by 11 and 40 ms, corresponding to temporal delays of 1.02 and

3.6% of the cycle duration, for the tibialis and peroneus muscles,

respectively.

Discussion

This investigation was an attempt to understand whether and to what

extent the central nervous system modifies its output, or part of it, in

order to produce a continuous curvilinear locomotion. Emphasis has

been put upon the analysis and comparison of straight-ahead with

curved walking.

Gait features

Frequency

Walking along the circular path, either with vision (EO) or blindfolded

(BF), produced no major changes in stepping frequency, thus indicat-

Fig. 5. Representative EMG activity of thigh and leg muscles for left and right limb during straight-ahead and curved walking. Traces were obtained by averagingtime-normalized EMG traces for all trials of one subject walking in the eyes-open condition. EMG amplitude has been normalized in terms of maximal voluntarycontraction, and is expressed as a percentage of that activity. The time course of the ankle angle is traced at the bottom of the figure. Although the general spatial andtemporal patterns of muscle activity were not radically different between the two walking paths, the progression along the curve was accompanied by subtle limb-

specific modifications in muscle activity (see text).




9/15

ing that the transition from a straight to a curved path did not imply a

reorganization of rhythm production, or that the basal rhythm is

adapted to both conditions. There was also no difference in duration

of the right and left cycles during turning. The CNS would indeed

generate a strongly coupled locomotor pattern, during straight-ahead

locomotion as much as during locomotion along a curved path. It

is of interest that crayfish strongly modulate the frequency of each

locomotor limb while they progress along curvedpath (Domenici etal.,

1998), suggesting that temporal coupling across effectors is

not a systematic rule of turning but rather an exception of biped

walking.

The same rhythm was propagated to both legs, in spite of the

obvious but nontrivial consideration that the length of right and left

strides was definitely different when producing the curved trajectory.

With the present radius of curvature, and an average distance between

the feet of some 20cm, the inner foot had to travel along a circle with a

length of15% less than the outer foot. This led to a correspondingdecrease in the length of the right stride (see Fig.3). When the

movement of legs was referenced to body co-ordinates, the length

of left and right strides increasingly diverged with the increase in

heading change. However, the amount of stride length changes with the

increase in velocity obeyed a similar law for both legs. Notably, this

relationship was the same for the left leg during both straight and

curved trajectories, for both slope and intercept. For the right leg the

slope did not change, but the intercept diminished during turning. The

difference between the intercept for the relationships of left and right

Fig. 6. Mean values of EMG activity during straight-ahead and turning gait cycles in the soleus (A, stance phase) and tibialis muscle (B, swing phase) of both legs.Mean valueshave been normalizedwith referenceto themean value of EMGactivity duringthe progression along thestraight path (EOand BF, respectively), foreachlimb and gait phase independently (P


10/15

legs matched the difference in radius of virtual circles along which the

feet moved.

Changes in temporal and spatial features of leg movements were

mirrored in the synthetic measure of gait provided by thewalking ratio,

a velocity-independent index of walking pattern (Sekiya & Nagasaki,1998). This measure confirmed that commands to produce outer leg

forward progression were the same under both straight-ahead and

curved walking. The behaviour of the inner leg was different in

amplitude of progression but not so in cadence. It is easy to conclude

that the CPG distributes its commands to the two legs with the same

frequency, which does not impede the two legs advancing at different

velocities. Others have shown that the two CPGs are normally strongly

entrained (Riek & Carson, 2001) and produce synchronous output to

the two sides of the body (Guadagnoli et al., 2000). Therefore, during

curved walking the CPG would need to distribute a differently

patterned command to the two sides, in a manner appropriate to

perform turning.

Intrinsic temporal features of gait

Not only were spatial features of inner and outer leg movements

different during turning, but also somekey temporal features of thegait

pattern were changed, yet within the cadre of the unchanged frequency.

The duration of the stance phase of gait diminished and increased in

outer and inner legs, respectively. This discrepancy in the temporal

sequences of gait events of the inner and outer legs provoked a

phase lag between the overall limb displacements, which corresponds

to 7% of the total cycle duration. Changes in the phase relationshipof this entity between the two limbs have been observed during

straight-ahead walking in toddlers, and it has been suggested that

they are connected to gait instability (Clark, 1995). Conversely, we

show here that this phase lag is a necessary condition to produce body

rotation.

The longer duration of the swing or transport phase (consequence of

the shorter stance phase) allows more time for the outer leg to travel the

longerdistance imposed by turning. Swing duration during turning waslonger than during straight walking, because the path travelled by the

foot was not linear during turning but bent because of the concomitant

body rotation. This explains the different vertical distance between the

lines bestfitting velocity of foot swing against body progression for the

right and left foot. In other words, the difference in both stance

duration and travelling distance between left and right feet imposed

different swing velocities. In this regard, turning-related gait features

were reminiscent of those associated with the adaptation of walking on

a treadmill with split belts adjusted at different speeds. Indeed, Dietz

and collaborators (Dietz et al., 1994; Prokop et al., 1995; Zijlstra &

Dietz, 1995; Jensen et al., 1998) have shown that, under such condi-

tions, stride length was larger on the fast side (as for the outer leg

during turning, in our case) with respect to the slow side. At the same

time, the stance duration was shortened on the fast leg whereas it

increased on the slow leg, and inversely for swing phase duration.

Concomitant with the increased duration of the stance phase of the

right, inner, leg, the trunk leant towards the interior. This was con-

firmed by a significant relationship between trunk roll and stance

duration. There is therefore an association between trunk leaning to the

inner side, increase of walking speed during turning (see companion

paper) and increased stance duration of the inner limb. The displace-

ment of the trunk could contribute to the modulation of the temporal

features of the gait pattern, because loading the inner leg (and

unloading the outer leg) might produce a phase modulation through

re-afferent input from the evolving movement (see below).

Fig. 7. Distribution of the ends of soleus EMG burst and onsets of thefirst burst of tibialis anterior in the left (upper panel) and right (lower panel) leg, separatedaccording to the direction of walking. No difference occurred in the end of the soleus burst, regardless of side and trajectory conditions (compare upper and lowergraphs of the left panel). With respect to straight-ahead walking, the distributionof tibialis anterior burst onsets measured during turning wasshifted toward lower andhigher values on the left and right leg, respectively. Limb-specific mean changes ( SD) in the occurrence of the onset of tibialis anterior burst is indicated in the insethistogram. Significant differences are indicated as in Fig. 3.




11/15

EMG pattern

Subtle but consistent changes, which appear to be associated with the

above described motor behaviour, occurred in the leg muscles

recorded. The similarity between the patterns of muscular activation

responsible for the body progression along straight or curved paths

indicates that turning did not induce dramatic changes in the organiza-

tion of the efferent commands to these muscles. Indeed, EMG pattern

presented similar shape- and phase-dependent modulation during the

two types of motion. Nevertheless, amplitudes and timing changes

were consistently observed.

Soleus

The overall pattern of soleus activity was similar during turning andstraight-ahead walking. However, the amplitude of the soleus burst

decreased during stance in the inner leg (11%) and increased during

turning in the outer leg (8%). Overall, the difference between the bursts

of inner and outer legs during walking was 18%. At the same time,left and right strides showed a discrepancy in their length (left stride

longer than right stride) of about the same amount. The precise origin

of extensor muscle modulation during walking along the curve is not

readily explained.

The relationship between EMG amplitude and stride length does not

seem to be generalisable. On the one hand, the decrease in soleus burst

amplitude during turning on the right with respect to left side can be

related to the shortening of the right stride with respect to the left. The

soleus burst of the right leg also diminished during turning with respectto its burst during straight-ahead walking. However, in the two cases

there was no correspondence between average EMG changes and

average stride length changes. However, the left soleus increased not

only with respect to the right muscle during turning, but also with

respect to the same left soleus during straight walking. However,

during straight walking the stride length of the left leg did not change

with respect to its stride length on the curved path.

We feel that any attempt to explain these discrepancies should take

into account the fact that body orientation changes during turning. This

can create nonoptimal biomechanical conditions for the propulsive

action of the leg muscles, owing to the different spatial relationship

between the force vector produced by soleus muscle action and the

direction of the moving body. In turn, any disadvantage would be

differently shared by the two limbs. In the companion paper weshowed that foot yaw in relation to body heading was different for

the two feet during turning with respect to straight walking. In

particular, the left foot tended to approach the body trajectory while

the right foot tended to turn to the inner side of the trajectory. Within

this general trend, however, the relationship between foot positioning

and body trajectory changed during the completion of the stance phase.

At heel strike the left foot was almost aligned with heading direction

but progressively rotated with respect to body direction, thereby

diminishing the mechanical efficiency of the muscle action. This

might therefore require a stronger muscle action to support body

propulsion along the curved trajectory. Soleus activity was also larger

with respect to that recorded during straight-ahead walking,

where stride length was not different from that observed during

turning. Such an enhancement of the outer extensor muscle activity

probably contributes to the generation of the torque necessary for

producing the body turn. Moreover, because the stance phase duration

is shortened, a larger soleus activity would help create the torque for

turning.

An inverse trend was observed for the right foot, which was placed

on thefloor at a wide angle with respect to heading, and became almost

parallel to body direction at the time of toe-off. This might explain the

need for less muscle activity in the right than left soleus muscle during

turning, and in the same right soleus during turning compared to

straight walking. Two further effects of turning, liable to require

modulation in opposing senses of right soleus activity, are the smaller

Fig. 8. Comparison of EMG envelopes of inner and outer legs, recorded duringwalking along the curved trajectory. The mean profiles have been obtained byaveraging the mean traces of all subjects after normalization of their amplitudebased on muscular activity during straight-ahead cycles. The discrepancy inleftright amplitudes is revealed by the hatched area, the pattern of whichindicates the leg for which the muscle activity increases compared to the otherleg (see legend). The quantification of the phase shift between EMG envelopeshas been obtained by cross-correlating thewaveforms of the inner and outer leg.The value of the phase lag is expressed as a percentage of the cycle and inabsolute duration, and is reported close to the corresponding traces. A positivevalue means that the EMGprofile of the outer leg leads thatof the inner leg.Ther-value is also indicated.




12/15

length of the inner stride, thereby requiring less propulsive force for the

right leg, and the upper body displacement toward the inner side,

possibly requiring a stronger supportive action of the right foot

extensor muscles (Fouad et al., 2001).

The EMG profile of the soleus did not shift in time (for either onset

or end time) with the change in the trajectory, for either the left or right

side muscles. One would thus conclude that central commands for

turning almost selectively modulate the amplitude but not the timing of

the command delivered to the two muscles.

Tibialis anterior

There were subtle but significant differences in tibialis anterior muscle

activity. Thefirst component of its double burst began at the onset of

the swing phase and produced foot dorsiflexion. The second compo-

nent anticipated heel strike and persisted for the first part of stance,

braking the foot drop before the burst of the soleus muscle. The tibialis

anterior was almost silent during the burst of the antagonist soleus

during mid stance, and there was no effect of walking type in this

phase. During the swing phase of curved walking, the amplitude of

tibialis anterior activity significantly increased in the left (16%) and

right (13%) legs with respect to the activity recorded during straight-

ahead walking.

The increase in the left tibialis EMG burst was partly due to the factthat swing duration was longer during turning than during straight-

ahead walking. In addition, the burst was earlier during turning (see

below). As a consequence, a larger share of the second component of

the tibialis burst occurred during the swing phase, thereby producing

the overall increase of the tibialis activity during the swing phase of the

gait cycles along the curved path. This event may be connected with

the need to enhance left tibialis stiffness during the swing phase, to

avoid the foot plantar flexion possibly favoured by increased foot

velocity, and to ease foot rotation towards the interior of the curve.

The case of the inner (right) leg was different, and the major increase

in the right tibialis muscle activity was limited to the burst beginning at

the onset of the swing phase. This event can be explained by the

temporal constraint acting on the inner foot. Indeed, owing to body

position being leant towards the interior of the curve, the shorter lengthof the stride and the decreased duration of the swing phase, the foot

must be dorsiflexed more rapidly than during straight walking. As a

corollary, the rate of change of ankle dorsiflexion occurring at the

beginning of the swing phase was increased during turning (see the

ankle angle traces in Fig. 5). It has been shown elsewhere that

modulation of the amplitude of the tibialis burst (as during continuous

soleus or tibialis vibration during walking) is accompanied by a

corresponding modulation of the speed and amplitude of dorsiflexion

(Courtineet al., 2001; Verschueren et al., 2002).

Turning implied a slight but consistent side-specific modulation in

the timing of the tibialis burst: it was advanced in the left leg and

delayed in the right, during curved with respect to straight-ahead

walking. This was possibly responsible for the shorter duration of

the stance phase on the left and the longer duration of the stance phase

on the right (Duysens & Pearson, 1980; Rossignol, 1996). However, the

correlation between onset of tibialis burst and onset of the swing phase

(as a percentage of cycle duration), although mildly significant, was not

consistent across subjects, preventing us from making strong conclu-

sions about this point. Similarly, no correlation between end of

soleus and onset of tibialis was observed across limbs or trajectory

types. The absence of this correlation suggests that the reciprocal

inhibition between ankle antagonists was not responsible for the

modulation of either burst timing change during the turn. If anything,

the mild correlation present during straight trajectories vanished during

turning.

Peroneus longus

During straight-ahead walking, there was a coactivation of peroneus

and soleus muscles. Such activity was not due to crosstalk between

soleus and peroneus muscles, because it was checked that pure foot

plantarflexion did not induce activity in the peroneus (see also Hunt

et al., 2001, for straight walking). Both soleus and peroneus muscles

were active during the stance phase and their bursts were similar in

shape and timing. The peroneus assists in plantar flexion and produces

abduction and eversion of the foot, which counteracts the internal

rotation action of the triceps muscle, and helps to produce the lateral

displacement of the body weight from theactual stance foot to the next,

contralateral, foot (Hunt et al., 2001). During turning, relatively large

changes occurred in the timing and amplitude of the peroneus burst

with respect to straight-ahead. First, the amplitude of the burst

decreased in both legs, but almost two times more in the inner than

in the outer leg. This general decrease in peroneus activity may be due

in part to the decrease in the body velocity during turning. Never-

theless, side-specific modification of muscle activity rather suggests

that these changes are directly connected to the achievement of body

rotation. This would be not particularly astonishing because peroneus

action produces a medio-lateral displacement of the body in addition to

helping progression. Although peroneus muscle action ensures lateral

body equilibrium, regardless of walking conditions, its functionalimplication should be different during turning than during straight-

ahead walking.

Turning implies an opposite, asymmetric, action of the two limbs in

order to produce the force pulling the body to the ever-changing

direction of the curved trajectory. In the companion paper, we reported

that the change in heading direction is associated with a narrowing of

stance width of the left leg. More precisely, the more the body rotates

the closer the distance between the line joining two successive left

(outer) footprints and the right foot print. As a consequence, the body

need not be displaced toward the inner foot because the latter is already

positioned underneath the body weight, as also indicated by the

decrease in the distance of the right foot to the body trajectory, thereby

requiring less action of the right peroneus muscle. The peroneus of the

left side, in turn, need not increase its activity during stance, either,

because its pushing action toward the inner side of the trajectory is

enhanced by the mechanical moment created by the increased distance

between the stance foot and the position of the body. During turning,

the combined muscular action of the peroneus and of the soleus of the

outer leg is then mainly devoted to assisting body rotation.

However, at the beginning of the swing phase the peroneus burst

appears to behave in a different way in the two legs: on the right side,

there is prolonged activity and relative increase of the burst amplitude.

We suggest that it is because of the relative increase in duration of the

stance phase on the right with respect to the left side, and because of

the need for the right feet to be extra-rotated to prepare the subsequent

foot heel strike.

Rectus femoris

On average, the rectus femoris was active around the transition

between swing and stance phases. In general, its activity was of

smaller amplitude than that of the other recorded muscles. No sig-

nificant difference could be found in the amplitude of the bursts of the

same muscle (left or right) between straight and curved walking. No

differences were found, either, when the two limbs were compared

during turning. This was equally true in terms of timing. Increase in

rectus femoris burst has been shown to occur during rapid complete

turning and it has been explained by the need to brake body progres-

sion and allow a quick spin turn (Hase & Stein, 1999). Absence of any




13/15

rectus EMG modulation, in our opinion, can therefore be taken as a

further indication of the smoothness of progression along the curved

trajectory.

Neural substrates for motor implementation of the turn

The actual neural substrates involved in the production of continuous

steering along a curved path cannot be identified from the present

results. Nevertheless, a tentative explanation can be based on the

reconsideration of the known sensory and motor mechanisms that

modulate the muscular pattern for walking.In most animal species, locomotion is controlled by central pattern-

generating networks (CPGs) located within the spinal cord, which are

under the continuous influence of supraspinal signals (Grillner, 1981;

Pearson, 1993; Rossignol, 1996; Burke, 2001). In addition, the sensory

feedback from the moving body parts is integrated within spinal

networks and may even be part of them, directly participating in

the activation of muscles for walking (Duysens & Pearson, 1980;

Gurfinkel et al., 1998; Nielsen & Sinkjaer, 2002). In this section, we

considerfirst the sensory inputs, which may modulate spinal networks

in order to produce and/or accompany the body turn. Then we will deal

with the neural structures possibly involved in re-programming the

appropriate motor output for steering along the curve.

Sensory inputs

During the performance of the turn, the body becomes increasingly

asymmetric in the motion of body parts. Consequently, the asymmetric

inflow of movement-related feedback might contribute to shaping the

motor commands directed to the muscles. From where and which

sensory modality would any asymmetric input come? The role of vision

may be excluded, at least under the present condition where subjects

reproduced a memorized trajectory, because the motor program for

turning was not obviouslymodified by the presence or absence of visual

cues. In normal subjects, the role of vision may be limited to informing

motor centres about the orientation and location of the body in the

earth-based reference frame, thereby simplifying navigation and

dynamic equilibrium. If anything, it is remarkable that, when vision

was allowed, the production of the curved path was not greatlyimproved with respect to BF (see companion paper).

Neck and vestibular receptors might be crucial during turning

(Fitzpatricket al., 1999; Bent et al., 2000; Bove et al., 2001). When

a change of heading is achieved, the head is normally turned towards

the interior (Grasso et al., 1996; Grasso et al., 1998a; Grasso et al.,

1998b; Patla et al., 1999; Hollandset al., 2001; Imaietal., 2001; Vallis

et al., 2001; Glasaueret al., 2002; Hollands et al., 2002). Under our

experimental conditions, the absolute range of periodic oscillations of

the head increased with the trajectory tightness (see companion paper).

Therefore, the pattern of head movement produced two superimposed

asymmetrical components for vestibular and lateral neck propriocep-

tive inputs. Thefirst component lasts throughout the whole duration of

the gait cycle. The second component is phasic and leads a change in

heading by 200 ms. Tonic head orientation may help in monitoringbody orientation with respect to the inertial force produced by the

rotation, and in detecting body deviation from the required equilibrium

(Grassoet al., 1996; Pozzo et al., 1998; Imai et al., 2001). The phasic

input, in turn, may be directly involved in the motor command

producing the rotation of the body. We have previously shown that

lateralized stimulation of neck muscle receptors by vibration provoked

involuntary body deviation during walking (Bove et al., 2001) or

rotation during stepping-in-place (Boveet al., 2002). In the same vein,

galvanic-induced asymmetric vestibular input causes subjects to turn

from their planned trajectory (Fitzpatrick et al., 1999; Bent et al.,

2000; Jahn et al., 2000; Dietz et al., 2001). This hypothesis is

consistent with the observation made by Hollands et al. (2001) that

subjects showed difficulties in achieving changes in walking direction

when the head was immobilized with respect to shoulders. Note,

however, that asymmetrical gait is produced during split-belt walking

despite a neutral head position (Dietz et al., 1994; Zijlstra & Dietz,

1995). Whether the neural mechanisms underlying the split-belt

simulation of turning condition may be similar to those actually used

to perform curved walking is an open question.

Both neck and vestibular inputs project onto vestibular nuclei of the

brainstem (Gdowski & McCrea, 1999, 2000), the discharge of whichincreases the level of extensor muscle activity during gait (Orlovsky,

1972; Matsuyama & Drew, 2000a). Matsuyama & Drew (2000b,a)

showed in the freely walking cat that head movement induces phasic

modulations of vestibular nuclei, which contribute to tuning the level

of EMG activity in leg muscles. During curved walking, asymmetric

vestibular and neck sensory inputs can thus possibly facilitate the

respective increase and decrease of the outer and inner ankle extensor

muscles observed in our experiment.

Information conveyed by proprioceptors embodied in the asymme-

trically moving legs may also be important for the turning-related

regulation of both timing and amplitude of muscular activation

patterns. Displacements of body weight toward the inner leg modify

the input from load receptors on both body sides (Duysens & Pearson,1980; Dietz & Duysens, 2000; Duysens etal., 2000). Consequently, the

onset of the swing phase would be delayed on the inner limb owing to

the persistent input from leg muscles and foot sole receptors (Duysens

& Pearson, 1980; Stephens & Yang, 1999; Pang & Yang, 2000; Fouad

et al., 2001). The contrary would occur on the left, outer, leg. In

addition, the longer step achieved by the left leg and the limb girdle

rotation during turning increases the hip angle range on the left side,

whereas the opposite occurs on theright. The resultant mismatch might

be associated with a delay in the end signal (the inputs from the hip

muscles and joint receptors) for the phase change of the respective gait

cycle (Grillner & Rossignol, 1978; Pang & Yang, 2000), or with the

modification of the relative coupling between the CPG centres of both

sides, otherwise driven at the same gait frequency on both sides by the

supraspinal tonic input. Recently, examination of the linkage betweenpatterns of activity in several hind-limb motor pools and the modula-

tion of cutaneous reflex pathways duringfictive locomotion in cats has

allowed the notion that some aspects of the locomotor pattern forma-

tion can be separated from rhythm generation (Burke etal.,2001)to be

forwarded. This might imply that the two CPG functions may be

embodied in distinct neural organization.

Supraspinal motor output

Obviously, under the present conditions turning is a deliberate action,

both for its initiation (not addressed here) and during its execution

during the steady-state phase of steering. To what extent, or by means

of which mechanisms, higher centres participate in producing walking

along a curved trajectory cannot be thoroughly discussed on the basis

of the present data. The curved trajectory produces inertial forces that

modify the context of dynamic equilibrium. Whereas straight-ahead

walking mainly requires antero-posterior equilibrium, greater

demands in lateral equilibrium emerge during body rotation. Drew

and colleagues showed that reticulo-spinal neurons contribute to the

selection of patterns of postural activity while intact cats walk along

level (Matsuyama & Drew, 2000a) and sloping (Matsuyama & Drew,

2000b) surfaces, or cross obstacles (Prentice & Drew, 2001). Similar

results have been found when lateralfictive turns are generated in the

swimming lamprey (Fagerstedtet al., 2001). Dynamic postural adjust-

ments accompanying body rotation may probably also be derived from

similar mechanisms in humans (Massion, 1992). Further, how much




14/15

the motor cortex contributes to the ultimate motoneuron output during

turning is hard to estimate. It has been recently shown that the motor

cortex has access to leg muscles during locomotion (Capaday et al.,

1999; Christensen etal., 1999; Schubert etal., 1999; Christensen etal.,

2001; Petersenet al., 2001), and that ample areas of the fronto-parietal

cortex are activated during walking (Fukuyama et al., 1997; Miyai

et al., 2001). Admittedly, much more effort is necessary to expand the

story of human turning and its control, given that turning can only

complicate the control of human bipedal walking, which is already

inherently unstable in the straight-ahead condition (Capaday, 2002).In particular, the dual integration within the locomotor spinal

networks of the processes underlying navigation and those controlling

equilibrium appears to pose both theoretical and methodological

challenges.

Acknowlegments

This research was supported by grants from the University of Pavia, the Italian

Ministry of Health and the Fondazione Salvatore Maugeri (IRCCS). G.C. wassupported by a grant from theFrenchMinisterof Research.We aremost gratefulto Dr Paul Stapley, who edited the last version of the manuscript.

Abbreviations

ANCOVA, analysis of covariance; ANOVA, analysis of variance; BF, blindfolded;CPG, central pattern-generating network; EMG, electromyography; EO, eyesopen; SA, straight ahead; TU, turning.

References

Alexander, R.M. (1989) Optimization and gaits in the locomotion of verte-brates. Physiol. Rev., 69, 11991227.

Bent, L.R., McFadyen, B.J., Merkley, V.F., Kennedy, P.M. & Inglis, J.T. (2000)Magnitude effects of galvanic vestibular stimulation on the trajectory ofhuman gait. Neurosci. Lett., 279, 157160.

Berthoz, A. & Viaud-Delmon, I. (1999) Multisensory integration in spatialorientation. Curr. Opin. Neurobiol., 9, 708712.

Bove, M., Courtine, G. & Schieppati, M. (2002) Neck muscle vibration andspatial orientation during stepping in place in humans. J. Neurophysiol., 88,

22322241.Bove, M., Diverio, M., Pozzo, T. & Schieppati, M. (2001) Neck muscle

vibration disrupts steering of locomotion. J. Appl. Physiol., 91, 581588.Burke, R. (2001) The central pattern generator for locomotion in mammals .

Adv. Neu rol., 87, 1124.Burke, R., Degtyarenko, A. & Simon, E. (2001) Patterns of locomotor drive to

motoneurons and last-order interneurons: clues to the structure of the CPG.J. Neurophysiol., 86, 447462.

Bussel, B., Roby-Brami, A., Nerris, O. & Yakovleff, A. (1996) Evidence for aspinal stepping generator in man. Paraplegia, 34, 9192.

Capaday, C. (2002) The special nature of human walking and its neural control.Trends Neurosci., 25, 370376.

Capaday, C., Lavoie, B.A., Barbeau, H., Schneider, C. & Bonnard, M. (1999)Studies on the corticospinal control of human walking. I. Responses to focaltranscranial magnetic stimulation of the motor cortex. J. Neurophysiol., 81,129139.

Christensen, L.O., Andersen,J.B., Sinkjaer,T. & Nielsen, J. (2001) Transcranialmagnetic stimulation and stretch reflexes in the tibialis anterior muscleduring human walking. J. Physiol. (Lond.), 531 , 545557.

Christensen, L.O., Morita, H., Petersen, N. & Nielsen, J. (1999) Evidencesuggesting that a transcortical reflex pathway contributes to cutaneousreflexes in the tibialis anterior muscle during walking in man . Exp. Brain

Res., 124, 5968.Clark, J.E. (1995) On becoming skillful: Patterns and constraints. Res. Q.

Exercise Sport., 65, 173183.Courtine, G.,Pozzo, T., Lucas,B. & Schieppati, M. (2001)Continuous, bilateral

Achilles tendonvibration is not detrimental to human walk.Brain Res. Bull.,55, 107115.

Dietz, V., Baaken, B. & Colombo, G. (2001) Proprioceptive input overridesvestibulo-spinal drive during human locomotion. Neuroreport, 12,27432746.

Dietz, V. & Duysens, J. (2000) Significance of load receptor input during

locomotion: a review. Gait Posture, 11, 102110.Dietz, V., Zijlstra, W. & Duysens, J. (1994) Human neuronal interlimb

coordination during split-belt locomotion. Exp. Brain Res., 101, 513520.Dimitrijevic, M., Gerasimenko, Y. & Pinter, M. (1998) Evidence for a spinal

central pattern generator in humans. Ann. NY Acad. Sci., 860, 360376.Domenici, P., Jamon, M. & Clarac, F. (1998) Curve walking in freely moving

crayfish (Procambarus clarkii). J. Exp. Biol., 201, 13151329.Duysens, J., Clarac, F. & Cruse, H. (2000) Load-regulating mechanisms in gait

and posture: comparative aspects. Physiol. Rev., 80, 83133.Duysens, J. & Pearson, K.G. (1980) Inhibition offl exor burst generation by

loading ankle extensor muscles in walking cats. Brain Res., 187, 321332.Fagerstedt, P., Orlovsky, G.N., Deliagina, T.G., Grillner, S. & Ullen, F. (2001)

Lateral turns in the Lamprey. II. Activity of reticulospinal neurons during thegeneration offictive turns. J. Neurophysiol., 86, 22572265.

Fitzpatrick, R.C., Wardman, D.L. & Taylor, J.L. (1999) Effects of galvanicvestibular stimulation during human walking. J. Physiol. (Lond.), 517,931939.

Fouad, K., Bastiaanse, C.M. & Dietz, V. (2001) Reflex adaptations duringtreadmill walking with increased body load.Exp. Brain Res., 137, 133140.

Fukuyama, H., Ouchi, Y., Matsuzaki, S., Nagahama, Y., Yamauchi, H., Ogawa,M., Kimura, J. & Shibasaki, H. (1997) Brain functional activity during gait innormal subjects: a SPECT study. Neurosci. Lett., 228, 183186.

Gdowski, G.T. & McCrea, R.A. (1999) Integration of vestibular and headmovement signals in the vestibular nuclei during whole-body rotation.

J. Neurophysiol., 82 , 436449.Gdowski, G.T. & McCrea, R.A. (2000) Neck proprioceptive inputs to primate

vestibular nucleus neurons. Exp. Brain Res., 135, 511526.Glasauer, S., Amorim, M.A., Viaud-Delmon, I. & Berthoz, A. (2002) Differ-

ential effects of labyrinthine dysfunction on distance and direction duringblindfolded walking of a triangular path. Exp. Brain Res., 145, 489497.

Grasso, R., Assaiante, C., Prevost, P. & Berthoz, A. (1998a) Development ofanticipatoryorienting strategies duringlocomotor tasks in children.Neurosci.

Biobehav. Rev., 22, 533539.Grasso, R., Glasauer, S., Takei, Y. & Berthoz, A. (1996) The predictive brain:

anticipatory control of head direction for the steering of locomotion.Neuroreport, 7, 11701174.

Grasso, R., Prevost, P., Ivanenko, Y.P. & Berthoz, A. (1998b) Eyeheadcoordination for the steering of locomotion in humans: an anticipatorysynergy. Neurosci. Lett., 253, 115118.

Grillner, S. (1981) Control of locomotion in biped, tetrapod and fish. In:Brooks, V.B. (ed), Handbook of Physiology. The Nervous System II. Am.Physiol. Soc., Bethesda, MD, pp. 11791236.

Grillner, S. & Rossignol, S. (1978) On the initiation of the swing phase of

locomotion in chronic spinal cats. Brain Res., 146, 269277.Guadagnoli, M.A., Etnyre, B. & Rodrigue, M.L. (2000) A test of a dual central

pattern generator hypothesis for subcortical control of locomotion. J. Elec-tromyogr. Kinesiol., 10, 241247.

Gurfinkel, V.S., Levik, Y.S., Kazennikov, O.V. & Selionov, V.A. (1998)Locomotor-like movements evoked by leg muscle vibration in humans.Eur. J. Neurosci., 10, 16081612.

Hase, K. & Stein, R.B. (1999) Turning strategies during human walking. J.Neurophysiol., 81, 29142922.

Hollands, M.A., Patla, A.E. & Vickers, J.N. (2002) Look where youre going!:gaze behaviour associated with maintaining and changing the direction oflocomotion. Exp. Brain Res., 143, 221230.

Hollands, M.A., Sorensen, K.L. & Patla, A.E. (2001) Effects of head immo-bilization on the coordination and control of head and body reorientation andtranslation during steering. Exp. Brain Res., 140, 223233.

Hunt, A.E., Smith, R.M. & Torode, M. (2001) Extrinsic muscle activity, footmotion and ankle joint moments during the stance phase of walking . Foot

Ankle Int., 22, 3141.Imai, T., Moore, S.T., Raphan, T. & Cohen, B. (2001) Interaction of the body,

head, and eyes during walking and turning. Exp. Brain Res., 136, 118.Jahn, K., Strupp, M., Schneider, E., Dieterich, M. & Brandt, T. (2000)

Differential effects of vestibular stimulation on walking and running. Neu-roreport, 11 , 17451748.

Jensen, L., Prokop, T. & Dietz, V. (1998) Adaptational effects during humansplit-belt walking: influence of afferent input. Exp. Brain Res., 118, 126130.

de Leon, R.D., Roy, R.R. & Edgerton, V.R. (2001) Is the recovery ofstepping following spinal cord injury mediated by modifying existing neuralpathways or by generating new pathways? A perspective. Phys. Ther, 81,19041911.

Massion, J. (1992) Movement, posture and equilibrium: interaction and coor-dination. Prog. Neurobiol., 38, 3556.

2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 18 , 191205



15/15

Matsuyama, K. & Drew, T. (2000a) Vestibulospinal and reticulospinal neuronal

activity during locomotion in the intact cat. I. Walking on a level surface.J. Neurophysiol., 84, 22372256.

Matsuyama, K. & Drew, T. (2000b) Vestibulospinal and reticulospinal neuronalactivity during locomotion in the intact cat. II. Walking on an inclined plane .

J. Neurophysiol., 84, 22572276.Miyai, I., Tanabe, H.C., Sase, I., Eda, H., Oda, I., Konishi, I., Tsunazawa, Y.,

Suzuki, T., Yanagida, T. & Kubota, K. (2001) Cortical mapping of gait inhumans: a near-infrared spectroscopic topography study. Neuroimage, 14,11861192.

Nielsen, J.B. & Sinkjaer, T. (2002) Afferent feedback in the control of human

gait. J. Electromyogr. Kinesiol., 12 , 213217.Orlovsky, G.N. (1972) Activity of vestibulospinal neurons during locomotion .

Brain Res., 46, 99112.Pang, M.Y. & Yang, J.F. (2000) The initiation of the swing phase in human

infant stepping: importance of hip position and leg loading. J. Physiol.(Lond.), 528, 389404.

Patla, A.E., Adkin, A. & Ballard, T. (1999) Online steering: coordination andcontrol of body center of mass, head and body reorientation.Exp. Brain Res.,129, 629634.

Pearson, K.G. (1993) Common principles of motor control in vertebrates andinvertebrates. Annu. Rev. Neurosci., 16, 265297.

Petersen, N.T., Butler, J.E., Marchand-Pauvert, V., Fisher, R., Ledebt, A., Pyndt,H.S., Hansen, N.L., Nielsen, J. & B. (2001) Suppression of EMG activity bytranscranial magnetic stimulation in human subjects during walking.

J. Physiol. (Lond.), 537, 651656.Pozzo, T., Papaxanthis, C., Stapley, P. & Berthoz, A. (1998) The sensorimotor

and cognitive integration of gravity. Brain Res. Brain Res. Rev., 28,92101.

Prentice, S.D. & Drew, T. (2001) Contributions of the reticulospinal system to

the postural adjustments occurring during voluntary gait modifications.J. Neurophysiol., 85, 679698.

Prokop, T., Berger, W., Zijlstra, W. & Dietz, V. (1995) Adaptational andlearning processes during human split-belt locomotion: interaction betweencentral mechanisms and afferent input. Exp. Brain Res., 106, 449456.

Riek, S. & Carson, R.G. (2001) Let your feet do the walking: constraints on thestability of bipedal coordination. Exp. Brain Res., 136, 407412.

Rossignol, S. (1996) Neural control of stereotypic limb movement. In Rowell,L.B. & Shepherd, J.T.(Eds),Handbook of Physiology. Exercise: Regulationand Integration of Multiple Systems.Am. Physiol. Soc., Bethesda, MD, pp.

173216.Schubert, M., Curt, A., Colombo, G., Berger, W. & Dietz, V. (1999) Voluntary

control of human gait: conditioning of magnetically evoked motor responsesin a precision stepping task. Exp. Brain Res., 126, 583588.

Sekiya, N. & Nagasaki, H. (1998) Reproducibility of the walking patterns ofnormal young adults: testretestreliability of the walk ratio (step-length/step-rate). Gait Posture, 7, 225227.

Stephens, M.J. & Yang, J.F. (1999) Loading during the stance phase of walkingin humans increases the extensor EMG amplitude but does not change theduration of the step cycle. Exp. Brain Res., 124, 363370.

Vallis, L.A., Patla, A.E. & Adkin, A.L. (2001) Control of steering in thepresence of unexpected head yaw movements. Influence on sequencing ofsubtasks. Exp. Brain Res., 138, 128134.

Verschueren, S.M., Swinnen, S.P., Desloovere, K. & Duysens, J. (2002) Effectsof tendon vibration on the spatiotemporal characteristics of human locomo-tion. Exp. Brain Res., 143, 231239.

Zijlstra, W. & Dietz, V. (1995) Adaptability of the human stride cycle duringsplit-belt walking. Gait Posture, 3, 250257.



human walking along a curved path,ii,gait features and emg patterns

Documents