Stumbling-Corrective Responses Long-Lasting, Context-Dependent Modification of Stepping in the Cat After Repeated

translational physiology

Infants Adapt Their Stepping to Repeated Trip-Inducing Stimuli

Marco Y. C. Pang1, Tania Lam2, and Jaynie F. Yang1,2

1Centre for Neuroscience and 2Department of Physical Therapy, University of Alberta, Edmonton, Alberta T6G 2G4, Canada

Submitted 23 April 2003; accepted in final form 19 June 2003

Pang, Marco Y. C., Tania Lam, and Jaynie F. Yang. Infants adapttheir stepping to repeated trip-inducing stimuli. J Neurophysiol 90:2731–2740, 2003. First published July 25, 2003; 10.1152/jn.00407.2003.This study examined whether human infants under the age of 12 molearn to modify their stepping pattern after repeated trip-inducingstimuli. Thirty three infants aged from 5 to 11 mo were studied. Theinfants were held over a moving treadmill belt to induce stepping.Occasionally, a mechanical tap was applied to the dorsum of the leftfoot during the early swing phase to elicit a high step. In some trials,the stimulus was applied for only one step. In other trials, the foot wasstimulated for a few consecutive steps. We determined whether theinfants continued to show high stepping immediately after the re-moval of the stimuli. The results showed that after the foot wastouched for two or more consecutive steps, some infants continued todemonstrate high stepping for a few steps after the removal of thestimuli (i.e., aftereffect). Such adaptation was achieved by an increasein hip and knee flexor muscle torque, which led to greater hip andknee flexion during the early swing phase. Aftereffects were morecommonly seen in older infants (9 mo or older). The results indicatedthat before the onset of independent walking, the locomotor circuitryin human infants is capable of adaptive locomotor plasticity. Theincreased incidence of aftereffect in older infants also suggests thatthe ability to adapt to repeated trip-inducing stimuli may be related toother factors such as experience in stepping and maturation of thenervous system.

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

To achieve functional walking, one must be able to adapt tochanges in the external environment, such as changes in theterrain or the presence of obstacles. Ample evidence supportsthe ability of adult humans to adapt to sustained changes in thelocomotor environment. For example, after a period of walkingon a rotating disk, blindfolded human subjects showed curvedwalking trajectories when attempting to walk in a straight line(Earhart et al. 2001; Gordon et al. 1995; Weber et al. 1998).This phenomenon, called aftereffect, revealed the modifica-tions of the motor program after exposure to sustained pertur-bations. Aftereffects have also been reported in other experi-mental conditions. After running forward on a treadmill, blind-folded subjects inadvertently jogged forward when asked to jogin place (Anstis 1995). Similarly, after running on a treadmillthat sloped upward, the horizontal treadmill belt was perceivedas sloping downward (Anstis 1995). Walking on a split-belttreadmill with the belts running at different speeds can alsoinduce aftereffects. After a period of split-belt walking, sub-jects were asked to modify the belt speeds until the two belt

speeds felt matched. A difference in speed between the twobelts remained (Jensen et al. 1998). These data showed that theadult human locomotor system is capable of adaptive plasticityafter sustained perturbations.

The results from the preceding studies are consistent withthose obtained from experiments studying upper limb move-ments (Flanagan et al. 1999; Gondolfo et al. 1996; Martin et al.1996a,b, 2002). With practice, adult humans were able to usea manipulandum to reach visual targets in the presence of aperturbing force field. The trajectory of the reach becamesimilar to the control condition without the force field aftersome training. Interestingly, the movement trajectory becamedistorted immediately after the disturbing force fields wereremoved, revealing an aftereffect. The presence of aftereffectsreflects the ability of adult humans to predict the disturbancesand modify the motor program to cancel the effects of theperturbations (Gandolfo et al. 1996).

Reduced mammalian preparations also showed adaptation torepeated perturbations during walking. In decerebrate ferrets,decerebrate cats and spinal cats, when the swing phase of theforelimb was repeatedly perturbed by a bar (i.e., trip-inducingstimuli), the animal learned to increase the maximum height ofthe limb during the swing phase to avoid the obstacle. Onremoval of the obstacle, the high stepping persisted for severalstep cycles (Bloedel et al. 1991; Edgerton et al. 2001; Hodgsonet al. 1994; Lou and Bloedel 1987). These results suggest thatthis particular form of learning does not require the cerebrum.

Would the developing human locomotor circuitry show sim-ilar phenomenon? Human infants have been used as a model tostudy the control of walking before the descending tracts fromthe motor cortex are fully mature (Yang et al. 1998a). In thisstudy, we examined whether human infants under the age of 1yr adapt to repeated trip-inducing stimuli during the swingphase of stepping. Our data showed that the high steppingpersisted for a few steps after the removal of trip-inducingstimuli, primarily in infants �9 mo of age. The results suggestthat the locomotor circuitry in infants is capable of adaptiveplasticity, but the plasticity is likely dependent on the matura-tion of specific neural structures. Preliminary results have beenpublished in abstract form (Pang and Yang 2002).

M E T H O D S

Subjects

The infants in this study were recruited through three local healthclinics. Ethical approval was obtained through the Health Research

Address for reprint requests and other correspondence: J. F. Yang, 2–50Corbett Hall, University of Alberta, Edmonton, Alberta, Canada T6G 2G4(E-mail: [email protected]).

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

J Neurophysiol 90: 2731–2740, 2003.First published July 25, 2003; 10.1152/jn.00407.2003.

27310022-3077/03 $5.00 Copyright © 2003 The American Physiological Societywww.jn.org

Ethics Board, University of Alberta and Capital Health, Edmonton,Alberta. Parents were contacted by phone and instructed to practicestepping with the infant for 1–2 min daily because practice has beenshown to increase the chance of obtaining good stepping in thelaboratory (Yang et al. 1998a). We contacted the parent/guardianmonthly to determine whether the infant was stepping. The infant wasbrought in for the experiment if the parent or guardian reported thatthe infant can make �10 consecutive steps, with support, at a time.Informed and written consent was obtained from the parent before theinfant participated in the study. The experiments were conducted inaccordance with the Declaration of Helsinki for experiments on hu-man subjects. Thirty-three infants aged 5–11 mo (mean: 9.1 mo) werestudied. All of the infants were born at term. None of the infants couldwalk independently at the time of the experiment.

Recording procedures

After the skin was cleaned with alcohol swabs, Kendall SOFT-E,pediatric (Ag/AgCl) electrodes were applied over four muscle groupsin the left leg: quadriceps (Q), hamstrings (HAMS), tibialis anterior(TA), and gastrocnemius-soleus (GS). An electrogoniometer (Pennyand Giles Computer Products, Biometrics, Blackwood Gwent, UK)was placed over the left knee joint to measure knee motion in thesagittal plane (flexion-extension). The goniometer was placed so thatone arm was aligned with the longitudinal axis of the femur and theother with the lower leg. Adhesive skin markers were placed over theleft side of the trunk just above the superior border of the iliac crest,the greater trochanter, the knee joint line, the lateral malleolus, and thelateral aspect of the fifth metatarsal-phalangeal joint of the left leg.The left view of the infant was recorded (30 frames/s) by using avideo camera (PV-950; Panasonic, Secaucus, NJ).

A Gaitway treadmill system (Kistler Instrument, Amherst, NY) wasused for all experiments. Beneath the treadmill belt were two forceplates, one in front of the other, to measure vertical ground reactionforces during walking. The infant was held under the arms by one ofthe researchers or by a parent with one hand on each side of theinfant’s upper trunk. The forearm of the individual holding the infantwas supported to ensure that no movement was imposed on the infant.The infant was allowed to support its own weight as much as possible.To prevent the infant and the adult holding the infant from seeing thebaton that was used to induce tripping, a linen sheet was placed �20cm above the treadmill surface surrounding the infant. The speed ofthe treadmill belt was adjusted to obtain optimal stepping (between0.22 and 0.31 m/s). Several trials of forward stepping were recordedfor each infant.

Trip-inducing stimuli were applied manually. In previous animalstudies (Edgerton et al. 2001; Lou and Bloedel 1987), disturbanceswere elicited by interjecting a bar into the path of the limb on eachsuccessive swing phase. Because the stepping pattern is more variablein infants, placing the bar in the same location would not haveprovided consistent perturbations. For example, the position of thelimb varies mediolaterally with each step and the onset of the swingphase varies. For these reasons, the bar may make contact withdifferent parts of the foot at different times in the swing phase.Therefore we applied the trip-inducing disturbances manually to makesure the perturbations were applied to the correct location of the foot(i.e., dorsum) at the desired part of the step cycle (i.e., early swing).An instrumental baton with a sponge-covered tip was used to brieflytouch the dorsum of the left foot during the early swing phase. Thebaton was instrumented with a force transducer to measure the amountof force applied to the foot during the disturbance. In some trials, onlyone swing phase was disturbed. In other trials, the foot was touchedfor a few consecutive steps (varying from 2 to 6 steps, randomly).Typically, the mechanical stimulus induced a response (see also Lamet al. 2003a), called the stumbling corrective response by Forssberg etal. (1975), who first showed it in spinal cats. We observed whetherhigh stepping persisted after the removal of the trip-inducing stimuli

(i.e., aftereffect). Trials with trip-inducing stimuli were repeated asmuch as possible, depending on the tolerance of the infant.

Throughout the experiment, infants were distracted with games andtoys. Each walking trial was typically 1–2 min long. The wholeexperimental session took �1 h. Electromyography (EMG), signalsfrom the baton, force plates, and knee electrogoniometer were ampli-fied and recorded on VHS tape with a pulse code modulation encoder(A. R. Vetter, Redersburg, PA). All walking trials were videotaped.The video and analog signals were synchronized by a custom-madedigital counter at a rate of 1 Hz. At the end of the session, the massof each infant was recorded (range: 7.3–12.5 kg).

Data analysis

The data were analyzed off-line. The EMG data were high passfiltered at 10 Hz, full-wave rectified, and low-pass filtered at 30 Hz.The signals from the baton, force plates, and knee electrogoniometerwere also low-pass filtered at 30 Hz. All the signals were thenanalog-to-digitally converted at 250 Hz (Axoscope 8; Axon Instru-ments, Foster City, CA).

The video data were reviewed to identify sequences of walking anddisturbances. The corresponding analog data were then identified. Thebeginning of the stance and swing phases were determined by footcontact and toe off, respectively, as indicated by the force plate signalsin conjunction with the video image. To obtain baseline measures ofjoint movements and toe clearance, 10 undisturbed steps were ran-domly chosen to serve as the control for each subject. For selection ofsuccessful disturbances, the following criteria were used: the forcesignal from the baton reached its peak during early swing phase (i.e.,before the knee goniometer signal reversed from flexion to extension),the peak disturbance force exceeded 0.5 N because we found that aforce of 0.5 N was sufficient to elicit a stumbling corrective response,and the sequence of the disturbances was followed by at least fiveconsecutive undisturbed steps because we wished to determine thetime course of any high stepping following the removal of the stimuli.

The peak force value recorded from the baton was used as ameasure of the force applied to the foot during the disturbance. Ifmore than one disturbance was applied in a trial, the peak force ofeach disturbance in that trial was summed and then averaged. Thisvalue served as an estimate of the average peak force applied to thefoot for that particular trial. Because the disturbance was appliedduring the swing phase and the subsequent modifications in locomotortrajectory occurred primarily in the swing phase, the data analysisfocused on the swing phase.

The force plate signals were analyzed to estimate the amount ofbody weight borne by the infant before and after the disturbances wereapplied. This was used to determine whether the person holding theinfant inadvertently changed the amount of weight support during andafter the disturbances. Changes in the amount of body weight supportcould influence the height of toe clearance. For each subject, theaverage force on the right leg during the left swing phase wascomputed for the control steps and the first post-disturbed step.

The relevant video data (i.e., the swing phases of the control steps,disturbed steps and 5 post-disturbed steps) were digitized from thevideotape to the computer (Adobe Systems, Mountain View, CA).The positions of the joint markers were digitized manually usingcustom-written software programs (Frame Analyzer, Garand Interna-tional Telecom). The position data were then filtered using a fourth-order Butterworth, dual-pass filter with a low-pass cut-off frequencyat 4 Hz for the hip, 5 Hz for the knee, and 6 Hz for the ankle and toe(Winter 1990). The maximum toe height (as indicated by the positionof the joint marker on the left 5th metatarsal-phalangeal joint), and theangles of the left hip, knee, and ankle joints were computed withcustom-written software programs (MATLAB; MathWorks, Natick,MA).

We were also interested in determining whether the adaptivechanges in movement pattern after the removal of the stimuli were

2732 M.Y.C. PANG, T. LAM, AND J. F. YANG

J Neurophysiol • VOL 90 • OCTOBER 2003 • www.jn.org

reflected by changes in muscle activity. Mean EMG amplitude of theTA burst for the whole duration of the swing phase was calculated forthe control steps and the five post-disturbed steps. If the TA EMG datashowed considerable artifact or crosstalk with GS EMG, the data werediscarded for this analysis. Concurrent activity in the two channelsduring times when the activity should be reciprocal (such as stanceand swing phase of walking) was defined as probable crosstalk. TheTA EMG data from one subject were eliminated as a result.

It is very difficult to record from the hip and knee flexor musclesfrom the infants because of the deep location of the muscles and theconsiderable amount of fatty tissue in the area, resulting in a smallsignal-to-noise ratio. To estimate the changes in muscle activity at thehip and knee joints during the swing phase, inverse dynamic analysiswas performed to estimate muscle torques at the hip and knee.

For the video images, out-of-plane movements were determined bychanges in the apparent lengths of the thigh and lower leg using thevideo data. If the apparent length of the thigh or lower leg varied by�10% during the swing phase, the step was eliminated from kineticanalysis. Only 9% of the steps were excluded as a result. Inversedynamic analysis using a two-segment model (i.e., lower leg and footwere considered as 1 segment) was performed to calculate the hip andknee torques (Hoy and Zernicke 1986). This was reasonable becausethe mass of the foot and the force contribution from the ankle are verysmall in the swing phase (Schneider et al. 1990). The torque valueswere normalized to the mass of the infant (Winter 1991). Muscletorques were evaluated for all control steps and the first post-disturbedsteps only because any adaptive changes in stepping would be mostapparent in the first post-disturbed step. Because there was variabilityin the swing phase duration between and within subjects, the swingphase duration was normalized to allow data averaging and compar-ison between trials and between subjects. Custom-written software(MATLAB) was used for all evaluations.

Control experiments

Control experiments were conducted with five infants to make surethat the aftereffect was not due to heightened excitability of the infantscaused by the repeated stimuli. EMG was not recorded and torqueanalysis was not performed in these infants. The trip-inducing stimuliwere applied in the same way as previously described. In addition,separate trials were recorded in which mechanical taps were applied tothe posterolateral aspect of the left thigh with the same baton. In bothcases, two to six consecutive disturbances were applied. If the after-effect was due to an increase in general excitability caused by repet-itive sensory stimulation, it should not matter where the stimuli wereapplied in the leg.

Statistical analysis

For each trial, one-tailed z tests were used to determine whether themaximal toe height for the first post-disturbed step was significantlyhigher than that from the averaged control steps. z score is a ratio ofthe difference in maximum toe height (between the 1st post-disturbedstep and the averaged control steps) and the SD of the maximum toeheight for the averaged control steps. An aftereffect was defined as az score �1.645 (P � 0.05). If repeated trials were available, the zscore for each trial was summed and averaged. The subject is con-sidered to have an aftereffect if the average z score �1.645. One-wayANOVA was used to determine whether the force applied to the footwas significantly different between trials with different number ofconsecutive disturbances (varying from 1 to 6). �2 test of associationwas used to test the relationship between the incidence of aftereffectand number of perturbed step cycles. Because older (�10 mo) andyounger infants (�8 mo) seemed to show a different response fol-lowing the removal of trip-inducing stimuli, �2 test of association wasalso used to test whether there was an association between the inci-dence of aftereffect and the two different age groups.

Independent sample t-test were used to compare the coefficient ofvariation for the maximum toe height during the swing phase in thecontrol steps for the two age groups. The two age groups comparedwere as defined in the preceding text (i.e., �10 and �8 mo). Thecoefficient of variation was defined as the ratio of SD over the meanfor the 10 control steps.

Two-way ANOVA with a mixed design (i.e., containing bothwithin and between subject factors) was used to determine whetherthere was a difference between the infants showing an aftereffect andthose who did not in a number of measures: 1) the duration of thestance and the swing phase for the control, disturbed, and post-disturbed steps. 2) The average left TA EMG during the swing phasefor the control and the first post-disturbed steps. And 3) the averagevertical ground reaction force from the right leg during the left swingphase for the control and first post-disturbed steps.

To examine the time course of the aftereffect, a one-way ANOVAwith repeated measures was used to compare the maximum toe heightfor the averaged control step and the five post-disturbed steps. Atwo-way ANOVA (completely randomized design) was used to de-termine whether the average peak force of the disturbances and theaverage maximum toe height for the disturbed steps were differentbetween subjects who showed an aftereffect and those who did not.Trials with a single disturbance were not included in the precedinganalyses because they produced very few aftereffects.

The statistical tests were conducted with mean values from eachsubject (i.e., averaged across all successful trials). An alpha value of0.05 was set for all statistical tests. To reduce the probability ofmaking a type I-error, the significance level was adjusted according tothe number of comparisons for all post hoc tests (Glass and Hopkins1996). Post hoc comparisons were made with the Bonferroni t-test.

R E S U L T S

Some infants showed high stepping after removalof trip-inducing stimuli

Some infants continued to show high stepping after theremoval of the trip-inducing stimuli. An example is shown inFig. 1, A–D. The black bars (Fig. 1A, between 3rd and 4thtraces) represent the stance phases of the left leg, whereas thespaces between the bars represent the swing phases. The max-imum toe height during swing phase is indicated (bottom trace)with each data point representing one step. The data pointsbetween the vertical dashed lines indicate the disturbed steps.In this particular example, three consecutive swing phaseswere perturbed (see the corresponding force signal from thebaton, middle bottom). The left leg reacted to the perturbationsby producing high steps as indicated by the maximum toeheight (bottom) and the increase in knee flexion angle (2ndtrace). The duration of the swing phase was concurrentlyprolonged in the disturbed steps. The maximum toe height didnot return to the control value until the third post-disturbedstep. The stick diagrams show the trajectory of the left leg forthe last undisturbed step (pre-disturbed step; Fig. 1B), the firstdisturbed step (Fig. 1C), and the first post-disturbed step (Fig.1D). It is obvious that the movement pattern was modified inthe first post-disturbed step when compared with the pre-disturbed step.

On the other hand, some infants did not show any aftereffectafter repeated trip-inducing stimuli. An individual example isshown in Fig. 1, E–H. In this case, six consecutive swingphases were perturbed (Fig. 1E). Similar to the previous ex-ample, the infant reacted to the disturbances by producing highsteps accompanied by a prolongation of the swing phase.

2733ADAPTATION TO TRIP-INDUCING STIMULI


However, immediately after the removal of the disturbances,no high stepping was observed. The maximum toe clearancefor the first post-disturbed step immediately returned to controlvalue. The stick figures illustrate the whole movement pattern.High stepping was elicited by the disturbance (Fig. 1G). Themovement trajectory for the first post-disturbed step (Fig. 1H)was very similar to that for the undisturbed step (Fig. 1F),indicating the absence of an aftereffect.

The group data indicated that all infants reacted to thedisturbances by producing high steps. The prolonged swingphase of the disturbed limb was accompanied by a concomitantincrease in the stance phase on the contralateral limb so that analternating stepping pattern was preserved (not shown). Be-cause the response of the contralateral limb is consistent withother disturbances we have used to prolong the swing phase

(Pang and Yang 2001; Yang et al. 1998b), no further analysiswas done. The response immediately after the withdrawal ofthe disturbances differed between infants. The maximum toeheight during the swing phase was substantially increased ininfants with an aftereffect (15 subjects) for the first and secondpost-disturbed steps compared with control steps (Fig. 2A). Forinfants without an aftereffect (18 subjects), the maximum toeheight only showed minimal increase for the first and secondpost-disturbed steps. Post hoc comparisons revealed a signifi-cantly higher toe clearance in infants who showed an afteref-fect compared with those who did not for the first post-disturbed step.

The swing phase was prolonged significantly during thedisturbed steps for both groups of infants (Fig. 2B). For the firstpost-disturbed step, the swing phase was slightly prolonged

FIG. 1. Single subject data showing 2types of responses. A: data from subject AAP.Signals from the left tibialis anterior (TA)electromyogram (EMG; top), left knee goni-ometer (middle top), instrumental baton (mid-dle bottom) are shown. Bottom: the left stepcycle (SC). The black bars represent thestance phase, whereas the spaces between thebars represent the swing phase. Bottom: themaximum toe clearance for the pre-disturbedsteps (pre), disturbed steps (demarcated byvertical dashed lines) and 5 post-disturbedsteps (post). Each data point represents 1 sin-gle step. In this particular example, 3 consec-utive swing phases were disturbed. High step-ping persisted for another 2 step cycles beforethe maximum toe height returned to controlvalue [statistical tests (z score) were only per-formed for the post-disturbed steps; doubleasterisk: P � 0.01]. B–D: stick diagrams forthe same subject during swing phase of thenormal step immediately before the 1st dis-turbance (pre-disturbed step; B), the 1st dis-turbed step (C), and the 1st post-disturbedstep (D). Each stick diagram corresponds to 1video frame, from left toe off to left footcontact (swing phase). The stick figures areequally spaced horizontally to minimize over-lap. The infant reacted to the disturbance bygenerating a high step (C). Immediately afterthe removal of the trip-inducing stimuli, theinfant continued to demonstrate high step-ping, indicating the presence of an aftereffect(D). E: data from subject KIB. Six consecu-tive swing phases were disturbed. However,the maximum toe height immediately re-turned to control value after the removal ofthe stimuli. F–H: the infant reacted to thetrip-inducing stimuli by generating a high step(G), similar to the example shown in C. How-ever, the limb trajectory for the 1st post-dis-turbed step (H) resembled that for the pre-disturbed step (F), indicating the absence ofan aftereffect.



and the stance phase slightly shortened in both groups ofinfants. There were no differences in the stance or swing phasedurations between the two groups for any of the steps.

Number of consecutive disturbances affected the appearanceof an aftereffect

Figure 3A shows the relationship between the number ofconsecutive disturbances applied in a given trial and the per-centage of trials in which an aftereffect was successfully in-duced. Data from all infants are included. When a single swingphase was perturbed, aftereffects were rarely seen. As thenumber of consecutive disturbances was increased to two,aftereffects were observed more frequently. The success rate ofinducing an aftereffect remained more or less the same as thenumber of consecutive disturbances was further increased up tosix. There was a significant association (�2 test of association,P � 0.05) between the number of perturbed step cycles and theincidence of an aftereffect. The average peak force applied tothe foot was the same regardless of the number of consecutive

disturbances (1-way ANOVA), so the ineffectiveness of induc-ing aftereffects by a single perturbation is not due to differ-ences in force (Fig. 3B).

Age affected the incidence of obtaining an aftereffect

The most interesting finding was a correlation between thepresence of an aftereffect and age. Data were pooled for alltrials with two or more disturbances in an infant because therewas no difference in the response between disturbances of twoor more (number of trials per subject: median: 5, mean: 6). Thepooled data were plotted so that the horizontal axis representsthe age of the infants while the vertical axis indicates theaverage z score for maximum toe height in the first step afterthe disturbance was removed (Fig. 4). Each data point repre-sents the averaged data from one infant; - - - indicates thez-score value of 1.645. Data points above the line indicate asignificant aftereffect. Despite variability between subjects, thez scores showed a tendency to increase with age. Most of theinfants �9 mo of age (71%) and few of the infants �9 mo ofage (20%) demonstrated an aftereffect.

The infants were further divided into two groups: those �8mo and those �10 mo (Fig. 5). For each group, the number of

FIG. 2. Group data showing 2 types of responses. The maximum toe heightduring the swing phase (A) was significantly higher in the infants who showedan aftereffect (F) compared with those who did not (E) for the 1st post-disturbed step (post 1). The error bar represents 1 SE. Both groups of infantsshowed a slightly prolonged left swing phase (B) and a slightly shortenedstance phase (C) in the 1st post-disturbed step. There were no differences in theduration of the stance and swing phases between the 2 groups of infants in anyof the steps.

FIG. 3. Incidence of aftereffect and the number of perturbed step cycles. A:the percentage of trials with an aftereffect was relatively low when a singlestep cycle was perturbed. When �2 consecutive swing phases were disturbed,the incidence of aftereffect was increased. B: there was no significant differ-ence in force regardless of the number of consecutive disturbances applied.

FIG. 4. Distribution of z score for maximum toe height. The horizontal axisrepresents age while the vertical axis represents the z score for maximum toeheight. Each data point indicates the z score for 1 subject. A score �1.645(- - -) indicates a significant difference in maximum toe height between the 1stpost-disturbed step and the averaged control step. It is apparent that olderinfants (�9 mo) tend to show a higher z score than the younger infants.



infants with and without an aftereffect was plotted against thenumber of consecutive disturbances applied in a given trial(number of trials per subject for each category of disturbance:median: 2, mean: 2). For the younger infants, aftereffects wererarely seen regardless of the number of disturbances applied(Fig. 5A). In contrast, for the older infants, the probability ofobtaining an aftereffect was increased if more than one distur-bance was applied (Fig. 5B). There was a significant differencein the incidence of obtaining an aftereffect between the twogroups of infants for trials in which three or more consecutivedisturbances were applied (�2 test of association).

Possible confounding factors

Could the difference in obtaining an aftereffect result fromthe difference in the amount of force applied to the foot? Thepooled data (Fig. 6A) showed that there was no significantdifference in force between the infants with an aftereffect (■ )and those without an aftereffect (�) regardless of the number ofconsecutive disturbances applied (2-way ANOVA). Is it pos-sible that the two groups of infants responded to the distur-bances differently and as a result exhibited differences in theaftereffect? The maximum toe height achieved was comparedbetween infants that showed an aftereffect and those that didnot. There was no systematic trend in this comparison, al-though in two comparisons (i.e., 4 and 5 consecutive distur-bances) the two groups were significantly different (Fig. 6B).

Did the experimenter impose changes to the infant’s weightsupport that might account for differences in the post-distur-bance steps? For the infants without an aftereffect, the average

force borne by the right leg during the swing phase of the leftleg was 36.3 � 3.1% body wt (means � SE) during the controlsteps, which was not significantly different from that during thefirst post-disturbed steps (34.7 � 2.8% body wt). Similarfindings were obtained in infants with an aftereffect (controlsteps � 38.0 � 3.1% body wt, 1st post-disturbed steps �36.5 � 3.2% body wt). There was also no difference in thewalking speed between the two groups (average speed of 0.26m/s for infants without an aftereffect and 0.25 m/s for infantswith an aftereffect). Therefore neither changes in body-weightsupport nor treadmill speed could account for the differentresponses in the two groups of infants.

Could the relationship between the z score and age be due toa less variable stepping pattern in the older infants? Presum-ably, a less variable stepping pattern would result in a smallerSD and thus a higher z score. The first post-disturbed stepshowed an average increase in toe clearance from the controlsteps of 4.5 cm for the older infants, which was two timeshigher than the younger infants (2.1 cm). The mean SD of themaximum toe height for the control steps was 1.4 cm [coeffi-cient of variation (CV) � 0.39] and 1.7 cm (CV � 0.42),respectively, for the infants �10 mo and those �8 mo. Thusthe stepping of the younger infants was only slightly morevariable than that of the older infants. The coefficient ofvariation was not significantly different between the twogroups (independent sample t-test). Therefore the difference invariability of stepping could not explain the different responsein the two age groups.

Time course of the aftereffect

Pooled data across subjects who showed an aftereffect indi-cated that the aftereffect was relatively short-lived, varyingfrom one to two step cycles (ANOVA with repeated measures)regardless of the number of perturbed step cycles. Figure 7illustrates the maximum toe clearance for the infants whoshowed an aftereffect following the application of two and sixdisturbances. The averaged control steps, disturbed steps, andthe five post-disturbed steps are shown. After the trip-inducingstimuli were removed, high stepping persisted for another oneto two step cycles before the toe clearance returned to thecontrol value.

Kinematics

Kinematic analysis revealed that the high stepping after theremoval of the trip-inducing stimuli was produced by an in-crease in hip and knee flexion. Pooled data for the trials in

FIG. 5. Relationship between age and incidence of aftereffect. The infantswere divided into 2 groups: those �8 mo (A) and those �10 mo (B). �, theinfants without an aftereffect; ■ , the number of infants who demonstrated anaftereffect. An aftereffect tended to occur more often in older infants especiallywhen the number of the perturbed step cycles was �2.

FIG. 6. Force applied and maximum toe height during thedisturbed steps. A: force applied during the disturbed steps. ■ ,those infants with an aftereffect whereas the open bars rep-resent those without an aftereffect. There was no significantdifference in force between the 2 groups of infants regardlessof the number of consecutive disturbances. B: maximum toeheight during the disturbed steps. F, those infants with anaftereffect; E, those without an aftereffect. — and - - -, theaveraged control value and 1 SE, respectively. The 2 groupsof infants responded to the disturbances similarly (*P �0.05).



which two to six consecutive swing phases were disturbed isillustrated in Fig. 8. Hip angle (Fig. 8, A and B), knee angle(Fig. 8, C and D) and toe height (Fig. 8, E and F) are plottedagainst the percentage of normalized swing phase duration (0%represents toe-off, whereas 100% represents foot-floor con-tact). For those infants who did not demonstrate an aftereffect(Fig. 8, left), the hip and knee angle and toe height for the firstpost-disturbed steps (thick lines) were quite similar to those forthe control steps (thin lines). Slight increase in knee flexion andtoe clearance is noted, however, probably because a number ofinfants in this group showed an increase in maximum toeheight but barely missed the significance level of 0.05 (see Fig.4). In contrast, for those infants who demonstrated an afteref-fect, the movement during the swing phase of the first post-disturbed step was characterized by an increase in hip and kneeflexion (Fig. 8, right). Note that the knee flexion angle peakedslightly earlier than the hip flexion angle.

Muscle torques

The aftereffect was also seen as an increase in hip and kneeflexor muscle torques during swing phase. The average hip andknee muscle torque profiles during the swing phase are shownin Fig. 9. For infants without an aftereffect, the hip and kneetorque profiles for the first post-disturbed steps (thick lines)were quite similar to those for the control steps (thin lines;Fig. 9, A and C). In contrast, for infants with an aftereffect,there was a large increase in the hip flexor torque during thewhole swing phase, accompanied by a small increase in the

knee flexor torque in the early part of the swing phase (Fig.9, B and D).

The mean amplitude of the left TA EMG burst during thewhole swing phase was also measured. Overall, there was nosignificant increase in TA EMG amplitude between the aver-aged control step and the first post-disturbed step for infantswith an aftereffect and those without (not shown).

Control experiments

Five infants participated in the control experiments. All ofthe five infants reacted to the trip-inducing stimuli to thedorsum of the foot by increasing their toe clearance during theswing phase (z score �7.1 � 1.5, means � SE). In contrast,none of the infants responded to the thigh stimulation byproducing high steps (z � 0.2 � 0.1). More importantly, whilefour of five infants showed an aftereffect in the first post-disturbed step after the trip-inducing stimuli were removed(z � 5.8 � 2.0), none of these infants showed an aftereffectfollowing the removal of thigh stimulation (z � 0.5 � 0.3).The average force applied to the dorsum of the foot (3.4 � 0.4

FIG. 8. Kinematic changes associated with an aftereffect. Left: the data forthose infants who did not show an aftereffect; right: the data for those whoshowed an aftereffect. The changes of hip angle (A and B), knee angle (C andD), and toe height (E and F) are plotted against the time during the normalizedswing phase (0% represents toe off; 100% represents toe contact). In eachdiagram, the thin solid line indicates the control step (with the light gray shaderepresenting 1 SE). The thick solid line represents the 1st post-disturbed step(with the dark gray shade representing one standard error). For the infants whodid not show an aftereffect, the hip and knee motion as well as toe trajectoryfor the 1st post-disturbed step only showed minor changes compared with thecontrol steps. In contrast, for the infants who demonstrated an aftereffect, therewas a large increase in hip and knee flexion, accompanied by an increase in toeclearance for the 1st post-disturbed step.

FIG. 7. Time course of aftereffect. A: data from the infants who demon-strated an aftereffect after 2 consecutive disturbances are shown (n � 10). Thecontrol step (C), disturbed steps (- - -) and post-disturbed steps (post) areshown. The trip-inducing stimuli clearly caused an increase in maximum toeheight during the swing phase. The maximum toe height remained significantlyabove the control value for 1 step (**P � 0.01). B: data from the infants whoshowed an aftereffect after 6 consecutive disturbances (n � 7). The maximumtoe height remained above the control value for another 2 steps after thewithdrawal of the disturbances.



N, means � SE) was not significantly different from thatapplied to the thigh (3.2 � 0.8 N).

D I S C U S S I O N

Our main finding is that many infants �9 mo learned toadapt to repeated trip-inducing stimuli by generating high stepseven after the stimulus was removed. The results indicate thatthe locomotor circuitry in human infants is capable of adaptiveplasticity, particularly after the age of 9 mo.

Methodological considerations

The trip-inducing perturbations were applied manually.Variability between disturbances was inevitable. Efforts weremade to ensure that we were as consistent as possible inapplying the disturbances. For example, the same researcherapplied all the disturbances in this study, the force applied tothe foot was quantified by a force transducer, and criteria wereset to guide our selection of successful disturbances (see METH-ODS). With these controls, the data showed that the variabilityin the disturbance force between subjects was small (Fig. 3B).There were no significant differences in force between theinfants with an aftereffect and those without (Fig. 6A). The twogroups of infants also responded similarly to the disturbances(Fig. 6B) with only two of the five comparisons being different.Thus it is unlikely that the two groups responded differently tothe touch stimuli. Moreover, we have also ruled out anychanges in body-weight support or treadmill speed as con-founding factors.

The results also showed that the difference in variability ofstepping between the older and younger infants cannot accountfor the difference observed in aftereffects in the two groups.

The coefficients of variation of the maximum toe trajectorywere very similar between the two age groups. Finally, thecontrol experiments indicated that it is unlikely that the after-effect was due to an increase in general excitability producedby repetitive mechanical stimuli. In summary, the presence ofaftereffects in infants cannot be attributed to methodologicalproblems.

Modifications in the movement pattern in response torepeated stimuli

All infants reacted to the disturbances by enhancing flexion,resulting in high steps (i.e., stumbling corrective response) (seealso Lam et al. 2003a). However, the response differed be-tween infants immediately after the disturbances were with-drawn. Some infants modified their motor program with re-peated stimuli to the dorsum of the foot (i.e., aftereffect). Themodified movement pattern is very similar to the high stepsgenerated during the disturbed steps (Fig. 1, C and D). More-over, the movement pattern resembles the “elevating strategy”in adult humans when the dorsum of the foot strikes an obstacleduring early swing phase (Eng et al. 1994). Similar to adulthumans, the response in human infants is characterized by anincrease in hip and knee flexion in the swing phase (Fig. 8, Band D), thereby increasing the foot clearance (Fig. 8F). Thekinematic changes result from an increase in hip and kneeflexor torque (Fig. 9, B and D). The modification in the move-ment pattern is functionally appropriate in that it increases thefoot clearance to avoid the obstacle (Eng et al. 1994).

Are lower centers of the human CNS capable of adaptivelocomotor plasticity?

Decerebrate ferrets and cats showed high stepping immedi-ately after the removal of repeated trip-inducing stimuli(Bloedel et al. 1991; Lou and Bloedel 1987), indicating thatthis form of learning does not require the cerebrum. When thecerebellum was also removed from decerebrate ferrets, theyshowed the same learning behavior, although the movementwas reported to be more disorganized (Bloedel et al. 1991).This finding suggests that the cerebellum, while necessary forcoordination of smooth movements, is not essential for thistype of learning either. In addition, it has been reported thatspinal cats show the same phenomenon (Edgerton et al. 2001;Hodgson et al. 1994). Unfortunately, only single-subject datawere presented in their reports. It is thus difficult to determinewhether aftereffects can be consistently obtained in all spinalcats. Whether the mature spinal cord is capable of this form oflearning remains an open question.

It has been well known that the isolated spinal cord iscapable of other forms of learning as demonstrated in experi-ments studying habituation and sensitization of spinal reflexesand classical conditioning (reviewed in Patterson and Grau2001; Wolpaw and Tennissen 2001). Operant conditioning ofspinal reflexes in intact animals also involves plastic changes atthe spinal level (reviewed in Wolpaw and Tennissen 2001).Moreover, the spinal cord is also able to learn specific func-tional tasks (i.e., stepping, standing) depending on the specifictraining regimen (De Leon et al. 1998a,b; Edgerton et al. 1992,1997; Viala et al. 1986). After peripheral nerve injury, spinalcats are capable of significant locomotor recovery (Bouyer and

FIG. 9. Hip and knee muscle torque profiles. Left: the data for those infantswho did not show an aftereffect; right: the data for those who showed anaftereffect. The average hip (A and B) and knee (C and D) torque profiles areillustrated. The figure convention is the same as Fig. 8. For the infants whoshowed an aftereffect, there was a large increase in hip flexor muscle torqueduring the whole swing phase accompanied by a smaller increase in kneeflexor muscle torque during the early part of the swing phase.



Rossignol 1998; Bouyer et al. 2001; Carrier et al. 1997). Morerecently, Timoszyk et al. (2002) demonstrated evidence oflearning in spinal rats subjected to sustained loading. Thereforespinal learning can occur under many different experimentalconditions.

It has generally been assumed that the stepping response inhuman infants is largely controlled by the brain stem and thespinal cord (Forssberg 1985; Peiper 1963). The cerebrum andits descending motor pathways to the spinal cord are notmature before the age of 1 yr as demonstrated by histological(Altman and Bayer 2001; Brody et al. 1987; Kinney et al.1988; Yakovlev and Lecours 1967), electrophysiological(Crum and Stephens 1982; Evans et al. 1990; Eyre et al. 1991;Issler and Stephens 1983; Koh and Eyre 1988; Muller et al.1991; Nezu et al. 1997; O’ Sullivan et al. 1981; Vecchierini-Blineau and Guiheneuc 1981) and radiological studies (Bark-ovich et al. 1988; Dietrich et al. 1988; Holland et al. 1986).Therefore based on our results, it is reasonable to suggest thatthe human subcortical locomotor circuitry is also capable ofadaptive plasticity just as in lower mammals.

What is intriguing in our results is the relationship betweenthe incidence of aftereffects and age. Infants �10 mo of agewere much more likely to show aftereffects than infants �8 mo(Fig. 4 and 5). We have demonstrated that methodologicalfactors are highly unlikely to have accounted for these differ-ences (see preceding text). Thus we are left with the possibilitythat other factors that change with age such as maturation ofthe nervous system, changes in body dimensions, or experiencewith stepping facilitate this form of learning. Previous expo-sure to similar repetitive perturbations is extremely unlikelybecause none of the infants could walk independently at thetime of the experiment. Infants �9 mo, however, are morelikely to be able to walk while holding onto furniture and thusto be exposed to situations where tripping could occur.Whether this experience affects the learning reported hereremains unknown.

The maturation of certain neural pathways may also beessential for this type of learning. Although animal studieshave shown that the cerebrum is not required for this type oflearning (Edgerton et al. 2001; Hodgson et al. 1994; Lou andBloedel 1987), the specific brain stem or spinal pathwaysinvolved in this type of learning have not been identified. Somestudies suggested that the cerebellum is important for motorlearning during perturbed locomotion (Earhart et al. 2002;Yanigihara and Kondo 1996). In contrast, Blodel et al. (1991)found that decerebrate animals could still adapt to trip-inducingstimuli 2 mo to 1 yr after a cerebellectomy. Although thisfinding shows that the cerebellum is not absolutely essential foradaptation to repeated trip-inducing stimuli, it does not rule outthe possible involvement of the cerebellum in intact animals.As infants approach 1 yr of age, many neural structures arematuring, including the cerebrum, the corticospinal tract andthe lateral cerebellar hemispheres (Barkovich et al. 1988;Brody et al. 1987; Kinney et al. 1988; Yakovlev and Lecours1967). It is possible that the maturation of these structurescontribute to the learning effects reported here. Moreover, wedo not know whether there are other maturational changes inthe spinal cord that are important for this type of learning. So,while it is clear there are age-related changes in an infant’sability to demonstrate this form of learning, we cannot addressthe cause for these changes in our study. Further animal work

would be one way to determine the neural substrate for thistype of learning.

In an earlier study from this laboratory, Lam et al. (2003b)used a different protocol to determine if learning occurred inyoung infants. In that study, sustained loading of a lower limbwas used. Contrary to the current findings, the previous one didnot show an age-dependent effect. We feel that this differenceis likely related to the difference in protocol of the experi-ments. When a weight was attached to an infant, the responseto the weight is less certain (i.e., he/she could respond vigor-ously by using more flexor activity, or respond minimally anddrag his/her foot more). This option may have resulted ingreater variability. In this study, the infants had no choice butto respond to the disturbance because the stumbling-correctiveresponse is a reflex. Thus all infants responded and indeedresponded in a similar way. The current protocol therefore ismuch more robust at showing learning if it was present.

In summary, the results suggest that the locomotor circuitryin humans is capable of short-term adaptive plasticity, beforethe onset of independent walking. The increased incidence ofaftereffect with increasing age indicates that neural maturationor experience may be some of the contributing factors thatenable the adaptation.

We thank Drs. M. A. Gorassini and V. Mushahwah for reviewing earlierversions of this manuscript. We also thank C. Wolstenholme for technicalassistance.

D I S C L O S U R E S

This work was supported by a grant from the Canadian Institutes of HealthResearch and the Canadian Neurotrauma Research Program to J. F. Yang.M.Y.C. Pang was supported by a scholarship from the Alberta HeritageFoundation for Medical Research and the Physiotherapy Foundation of Canadathrough an Ann Collins Whitmore Memorial Award.

REFERENCES

Altman J, and Bayer SA. Development of the Human Spinal Cord. NewYork: Oxford, 2001.

Anstis S. Aftereffects from jogging. Exp Brain Res 103: 476–478, 1995.Barkovich AJ, Kjos BO, Jackson DE, and Norman D. Normal maturation

of the neonatal and infant brain: MR imaging at 1.5T. Radiology 166:173–180, 1988.

Bloedel JR, Bracha V, Kelly TM, and Wu J-Z. Substrates for motorlearning. Does the cerebellum do it all? Ann NY Acad Sci 627: 305–318,1991.

Bouyer LJG and Rossignol S. The contribution of cutaneous inputs tolocomotion in the intact and the spinal cat. Ann NY Acad Sci 860: 508–512,1998.

Bouyer LJG, Whelan PJ, Pearson KG, and Rossignol S. Adaptive locomo-tor plasticity in chronic spinal cats after ankle extensors neurectomy. J Neu-rosci 21: 3531–3541, 2001.

Brody BA, Kinney HC, Kloman AS, and Gilles FH. Sequence of centralnervous system myelination in human infancy. I. An autopsy study ofmyelination. J Neuropathol Exp Neurol 46: 283–301, 1987.

Carrier L, Brustein E, and Rossignol S. Locomotion of the hindlimbs afterneurectomy of ankle flexors in intact and spinal cats: model for the study oflocomotor plasticity. J Neurophysiol 77: 1979–1993, 1997.

Crum JE and Stephens JA. Cutaneous reflex responses recorded in the lowerlimb of the newborn infant. J Physiol 332: 30–31P, 1982.

De Leon RD, Hodgson JA, Roy RR, and Edgerton VR. Locomotor capacityattributable to step training versus spontaneous recovery after spinalizationin adult cats. J Neurophysiol 79: 1329–1340, 1998a.

De Leon RD, Hodgson JA, Roy RR, and Edgerton VR. Full weight-bearinghindlimb standing following stand training in the adult spinal cat. J Neuro-physiol 80: 83–91, 1998b.



Dietrich RB, Bradley WG, Zaragoza IV EJ, Otto RJ, Taira RK, WilsonGH, and Kangarloo H. MR evaluation of early myelination patterns innormal and developmentally delayed infants. Am J Neuroradiol 9: 69–76,1988.

Earhart GM, Fletcher WA, Horak FB, Block EW, Weber KD, Sucho-wersky O, and Melvill Jones G. Does the cerebellum play a role inpodokinetic adaptation? Exp Brain Res 146: 538–542, 2002.

Earhart GM, Jones GM, Horak FB, Block EW, Weber KD, and FletcherWA. Forward versus backward walking: transfer of podokinetic adaptation.J Neurophysiol 86: 1666–1670, 2001.

Edgerton VR, de Leon RD, Tillakaratne N, Recktenwald MR, HodgsonJA, and Roy RR. Use-dependent plasticity in spinal stepping and standing.In: Advances in Neurology: Neuronal Regeneration, Reorganization andRepair, edited by Seil FJ. Philadelphia: Lippincott-Raven, 1997.

Edgerton VR, Roy RR, and de Leon RD. Neural Darwinism in the mam-malian spinal cord. In: Spinal Cord Plasticity: Alterations in Reflex Func-tion, edited by Patterson MM and Grau JW. Norwell: Kluwer Academic,2001.

Edgerton VR, Roy RR, Hodgson JA, Prober RJ, De Duzman CP, and deLeon RD. Potential of adult mammalian lumbrosacral spinal cord to executeand acquire improved locomotion in the absence of supraspinal input.J Neurotrauma 9: S119–128, 1992.

Eng JJ, Winter DA, and Patla AE. Strategies for recovery from a trip in earlyand late swing during human walking. Exp Brain Res 102: 339–349, 1994.

Evans AL, Harrison LM, and Stephens JA. Maturation of the cutaneomus-cular reflex recorded from the first dorsal interosseous muscle in man.J Physiol 428: 425–440, 1990.

Eyre JA, Miller S, and Ramesh V. Constancy of central conduction delaysduring development in man: investigation of motor and somatosensorypathways. J Physiol 434: 441–452, 1991.

Flanagan JR, Nakano E, Imamizu H, Osu R, Yoshioka T, and Kawato M.Composition and decomposition of internal models in motor learning underaltered kinematic and dynamic environments. J Neurosci 19: RC34(1–5),1999.

Forssberg H. Ontogeny of human locomotor control. I. Infant stepping,supported locomotion and transition to independent locomotion. Exp BrainRes 57: 480–493, 1985.

Forssberg H, Grillner S, and Rossignol S. Phase dependent reflex reversalduring walking in chronic spinal cats. Brain Res 85: 103–107, 1975.

Glass GV and Hopkins KD. Statistical Methods In Education and Psychology(3rd ed.). Needham Heights: Allyn and Bacon, 1996.

Gondolfo F, Mussa-Ivaldi FA, and Bizzi E. Motor learning by field approx-imation. Proc Natl Acad Sci USA 93: 3843–3846, 1996.

Gordon CR, Fletcher WA, Melvill Jones G, and Block EW. Adaptiveplasticity in the control of locomotor trajectory. Exp Brain Res 102: 540–545, 1995.

Hodgson JA, Roy RR, de Leon R, Dobkin B, and Edgerton VR. Can themammalian lumbar spinal cord learn a motor task? Med Sci Sports Exerc 26:1491–1497, 1994.

Holland BA, Haa DK, Norman D, Brant-Zawadzki M, and Newton TH.MRI of normal brain maturation. Am J Neuroradiol 7: 201–208, 1986.

Hoy MG and Zernicke RF. The role of intersegmental dynamics during rapidlimb oscillations. J Biomech 19: 867–877, 1986.

Issler HC and Stephens JA. The maturation of cutaneous reflexes studied inthe upper limb in man. J Physiol 335: 634–654, 1983.

Jensen L, Prokop T, and Dietz V. Adaptational effects during humansplit-belt walking: influence of afferent input. Exp Brain Res 118: 126–130,1998.

Koh TH and Eyre JA. Maturation of corticospinal tracts assessed by elec-tromagnetic stimulation of the motor cortex. Arch Dis Child 63: 1347–1352,1988.

Kinney HC, Brody BA, Kloman AS, and Gilles FH. Sequence of centralnervous system myelination in human infancy. II. Patterns of myelination inautopsied infants. J Neuropathol Exp Neurol 47 :217–234, 1988.

Lam T, Wolstenholme C, Van der Linden M, Pang MYC, and Yang JF.Stumbling corrective responses during treadmill-elicited stepping in humaninfants. J Physiol In press.

Lam T, Wolstenhome C, and Yang JF. How do infants adapt to loading ofthe limb during the swing phase of stepping? J Neurophysiol 89: 1920–1928, 2003b.

Lou J-S and Bloedel JR. A new conditioning paradigm: conditioned limbmovements in locomoting decerebrate ferrets. Neurosci Lett 84: 185–190,1987.

Martin TA, Keating JG, Goodkin HP, Bastian AJ, and Thach WT.Throwing while looking through prisms. I. Focal olivocerebellar lesionsimpair adaptation. Brain 119: 1183–1198, 1996a.

Martin TA, Keating JG, Goodkin HP, Bastian AJ, and Thach WT.Throwing while looking through prisms. II. Specificity and storage ofmultiple gaze-throw calibrations. Brain 119: 1199–1211, 1996b.

Martin TA, Norris SA, Greger BE, and Thach WT. Dynamic coordinationof body parts during prism adaptation. J Neurophysiol 88: 1685–1694, 2002.

Muller K, Homberg V, and Lenard HG. Magnetic stimulation of motorcortex and nerve roots in children. Maturation of cortico-motoneuronalprojections. Electroencephalogr Clin Neurophysiol 81: 63–70, 1991.

Nezu A, Kimura S, Uehara S, Kobayashi T, Tanaka M, and Saito K.Magnetic stimulation of motor cortex in children: maturity of corticospinalpathway and problem of clinical application. Brain Dev 19: 176–180, 1997.

O’Sullivan MC, Eyre JA, and Muller S. Radiation of phasic stretch reflex inbiceps brachii to muscles of the arm in man and its restriction duringdevelopment. J Physiol 439: 529–543, 1981.

Pang MYC and Yang JF. Interlimb coordination in human infant stepping.J Physiol 533: 617–625, 2001.

Pang MYC and Yang JF. Infants learn to adapt their walking to trip-inducingstimuli. Soc Neurosci Abstr 28: 267.10, 2002.

Patterson MM and Grau JW. Spinal Cord Plasticity: Alterations in ReflexFunction. Norwell: Kluwer Academic, 2001.

Pearson KG, Fouad K, and Misiaszek JE. Adaptive changes in motoractivity associated with functional recovery following muscle denervation inwalking cats. J Neurophysiol 82: 370–381, 1999.

Peiper A. Cerebral Function in Infancy and Childhood. New York: Consult-ants Bureau, 1963.

Schneider K, Zernicke RF, Ulrich BD, Jensen JL, and Thelen E. Under-standing movement control in infants through analysis of limb intersegmen-tal dynamics. J Mot Behav 22: 493–520, 1990.

Timoszyk WK, De Leon RD, London N, Roy RR, Edgerton RV, andReinkensmeyer DJ. The rat lumbosacral spinal cord adapts to roboticloading applied during stance. J Neurophysiol 88: 3108–3117, 2002.

Vecchierini-Blineau MF and Guiheneuc P. Excitability of the monosynapticreflex pathway in the child from birth to four years of age. J NeurolNeurosurg Psychiatry 44: 309–314, 1981.

Viala D, Viala G, and Fayein N. Plasticity of locomotor organization in infantrabbits spinalized shortly after birth. In: Development and Plasticity of theMammalian Spinal Cord, edited by Goldberger ME, Gorio A, and MurrayM. Padova, Italy: Liviana Press, 1986.

Weber KD, Fletcher WA, Gordon CR, Melvill Jones G, and Block EW.Motor learning in the podokinetic system and its role in spatial orientationduring locomotion. Exp Brain Res 120: 377–385, 1998.

Winter DA. Biomechanics and Motor Control of Human Movement. NewYork: Wiley, 1990.

Winter DA. The Biomechanics and Motor Control of Human Gait: Normal,Elderly and Pathological. Waterloo: University of Waterloo Press, 1991.

Wolpaw JR and Tennissen AM. Activity-dependent spinal cord plasticity inhealth and disease. Annu Rev Neurosci 24: 807–843, 2001.

Yakovlev PI and Lecours AR. The myelogenetic cycles of regional matura-tion of the brain. In: Regional Development of the Brain in Early Life, editedby Minkowski A. Oxford, UK: Blackwell, 1967, p. 3–70.

Yanagihara D and Kondo I. Nitric oxide plays a key role in adaptive controlof locomotion in cat. Proc Natl Acad Sci USA 93: 13292–13297, 1996.

Yang JF, Stephens MJ, and Vishram R. Infant stepping: a method to studythe sensory control of human walking. J Physiol 507: 927–937, 1998a.

Yang JF, Stephens MJ, and Vishram R. Transient disturbances to one limbproduce coordinated, bilateral responses during infant stepping. J Neuro-physiol 79: 2329–2937, 1998b.



Stumbling-Corrective Responses Long-Lasting, Context-Dependent Modification of Stepping in the Cat After Repeated

Documents