Biomechanics of overground vs. treadmill walking in ...cabrr.cua.edu/res/docs/publications/jap2008.pdf · Biomechanics of overground vs. treadmill walking in healthy individuals Song

Biomechanics of overground vs. treadmill walking in healthy individuals

Song Joo Lee1,2 and Joseph Hidler1,2

1Department of Biomedical Engineering, Catholic University, and 2Center for Applied Biomechanics and RehabilitationResearch National Rehabilitation Hospital, Washington, District of Columbia

Submitted 5 December 2006; accepted in final form 21 November 2007

Lee SJ, Hidler J. Biomechanics of overground vs. treadmill walkingin healthy individuals. J Appl Physiol 104: 747–755, 2008. First pub-lished November 29, 2007; doi:10.1152/japplphysiol.01380.2006.—Thegoal of this study was to compare treadmill walking with overgroundwalking in healthy subjects with no known gait disorders. Nineteensubjects were tested, where each subject walked on a split-beltinstrumented treadmill as well as over a smooth, flat surface. Com-parisons between walking conditions were made for temporal gaitparameters such as step length and cadence, leg kinematics, jointmoments and powers, and muscle activity. Overall, very few differ-ences were found in temporal gait parameters or leg kinematicsbetween treadmill and overground walking. Conversely, sagittal planejoint moments were found to be quite different, where during tread-mill walking trials, subjects demonstrated less dorsiflexor moments,less knee extensor moments, and greater hip extensor moments. Jointpowers in the sagittal plane were found to be similar at the ankle butquite different at the knee and hip joints. Differences in muscleactivity were observed between the two walking modalities, particu-larly in the tibialis anterior throughout stance, and in the hamstrings,vastus medialis and adductor longus during swing. While differenceswere observed in muscle activation patterns, joint moments and jointpowers between the two walking modalities, the overall patterns inthese behaviors were quite similar. From a therapeutic perspective,this suggests that training individuals with neurological injuries on atreadmill appears to be justified.

gait; motion analysis; electromyogram

OVER THE LAST DECADE, THERE has been a steady shift in gaittraining strategies in neurorehabilitation clinics where body-weight-supported treadmill training is now considered a viableintervention for treating gait impairments following neurolog-ical disorders such as stroke (12, 13) and spinal cord injury (9).Treadmill training has numerous advantages compared withoverground gait training, where the training can be done insmall area, a larger volume of steps can be achieved, walkingspeed can be well controlled, and because the subject isstationary and often elevated, the positioning of the therapist ismore optimal for providing assistance. Additionally, overheadbody weight-support systems can be utilized, relieving thesubject of a portion of their weight, which allows them to trainsafely and earlier in their recovery period (see Ref. 10 for areview).

Because the goal of all patients is to walk overground andnot on a treadmill, it is important that the motor controlstrategy utilized during each type of walking modality besimilar so that improvements in treadmill ambulation willtransfer to overground walking. In theory, if the belt speed ofthe treadmill is constant, biomechanically, there should be nodifferences between the two walking modalities (31). A num-

ber of previous studies have compared temporal gait parame-ters (1, 5, 24–26, 29, 33), joint kinematics(1, 5, 24, 25, 29), andmuscle activation patterns (2, 24, 25) between overground andtreadmill walking, however, the results are often conflictingand inconclusive. For example, Murray et al. (24) reported nostatistical differences in temporal gait parameters but claimedthat during treadmill walking, subjects demonstrated trends forshorter step lengths, higher cadences, shorter swing phases,and longer double-limb support. They also reported that de-spite no statistical differences in muscle activity except forquadriceps, EMG activity was, in general, higher on the tread-mill. This is in contrast to the reports by Arsenault et al. (2),who reported no differences in muscle activity in the soleus,rectus femoris, vastus medialis, or tibialis anterior between thetwo walking modalities. These investigators did report elevatedactivity in the biceps femoris and the variability was lower inmuscle firing patterns on the treadmill. Alton et al. (1) foundthat during treadmill ambulation, subjects had a shorter stancetime, higher cadence, larger hip range of motion, and greatermaximum hip flexion angle. Unfortunately, the experimentalprocedures in that study were questionable, because the walk-ing speeds between the two conditions were never accuratelymatched, and the method for estimating the joint angles,particularly at the hip, was problematic.

Although the studies described above have looked at tem-poral gait parameters, kinematics, and muscle activation pat-terns, only one has looked at joint moments and powers whilewalking on a treadmill (29). In that study, it was reported thatjoint moments and powers were statistically different, yetbecause these differences were within the variability of thekinetic measures, it was concluded that treadmill and over-ground joint moments and powers are similar. This study onlyreported on the maximum and minimum moments and powersduring stance, and it did not report on muscle activationpatterns for these trials.

Although the study by Riley et al. (29) provides the first lookat kinetic patterns during treadmill walking, it is important tolook at these behaviors across the gait cycle and also toexamine the corresponding muscle activation patterns. Becausemotor coordination, synergy patterns, energy expenditure, andother control strategies can be best observed by looking atkinetics and muscle activity, having a detailed understandingof these behaviors for treadmill and overground gait mayprovide important insight into developing gait training proto-cols for neurological subjects and for understanding how tread-mill-based interventions may be limited. It may also helpexplain the therapeutic benefits of body weight-supportedtreadmill training for individuals with lower limb impairments.

Address for reprint requests and other correspondence: J. Hidler, Dept. ofBiomedical Engineering, Catholic Univ., Pangborn Hall, #104b, 620 MichiganAve., NE, Washington, DC 20064 (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 Appl Physiol 104: 747–755, 2008.First published November 29, 2007; doi:10.1152/japplphysiol.01380.2006.

8750-7587/08 $8.00 Copyright © 2008 the American Physiological Societyhttp://www. jap.org 747

on March 12, 2008

jap.physiology.orgD

ownloaded from

http://jap.physiology.org

With the integration of force-sensing capabilities into split-belttreadmills (3, 22), ground reaction forces and centers of pres-sure can now be measured so that when combined withkinematic data, inverse dynamics calculations of joint mo-ments and powers during treadmill ambulation can now berealized.

The goal of this paper is to provide a comprehensive analysisof the temporal gait parameters, joint kinematics, joint kinetics,and muscle activation patterns utilized when subjects walk ona treadmill compared with overground ambulation. We hypoth-esize that when individuals walk on a treadmill, they willdemonstrate significant differences in joint moments and pow-ers, as well as muscle activation patterns when compared withoverground walking.

METHODS

Subjects

Nineteen healthy individuals with no known gait impairmentsparticipated in the study. Eight subjects (4 men, 4 women) were in theage range of 50–70 yr old (means � SD: 56.0 � 5.6), while 11subjects (5 men, 6 women) were between 18 and 30 yr old (means �SD: 23.5 � 3.3 yr). Our original goal was to determine whether therewere age-related differences in treadmill vs. overground walking.However, with analysis of the data, we found that there were noage-related differences in any of the metrics we examined.1 As such,we collapsed all subject data into the analysis.

Exclusion criteria included cardiac arrhythmia, hypertension, orany known gait abnormality such as an orthopedic injury, lower limbpain, or neurological injury that would bias the results of this study.All subjects were required to provide informed consent approved bythe Institutional Review Boards of Medstar Research Institute. Allexperiments were conducted at the National Rehabilitation Hospital(Washington, DC).

Instrumentation

The primary equipment used in this study was an ADAL3D-F/COP/Mz instrumented split-belt treadmill (TECHMACHINE, An-drezieux, France; see Ref. 3 for detailed description). As shown inFig. 1, each half of the treadmill is mounted on 4 Kistler triaxialpiezoelectric sensors (Winterthur, Switzerland). This allows forground reaction forces to be resolved in the vertical, anterior-poste-rior, and medial-lateral planes while subjects ambulate on the device.These forces could then be used to calculate the center of pressureunder each foot during stance.

Compared with previous studies that compared treadmill withoverground walking (1, 2, 5, 24–26, 33), one of the distinct advan-tages of our experimental setup is that the floor in our laboratory israised to be level with the ADAL treadmill. Thus, for the overgroundexperiments, the treadmill motors were turned off and the subjectssimply walked across the treadmill. Utilizing the same “force plates”for both treadmill and overground walking trials eliminated anyexperimental bias that might have been introduced into the groundreaction force measurements had the overground gait analysis beendone using standard force plates. In that setting, the sensors of thedifferent force plates may have different dynamic response character-istics. More importantly, the surfaces of the force plates in the twowalking modalities would be different (e.g., rubber treadmill vs. hardoverground). Changes in surface stiffness and damping result insignificant changes in muscle EMG activity (23), suggesting theimportance of consistent test conditions. It should be noted that beforethe study, we checked for belt slippage during overground walkingtrials by placing motion analysis markers on the surface of the belt assubjects walked across the treadmill. We did not observe any beltslippage during these overground test trials.

The kinematics of lower limbs were measured using a CodaMotionsystem (Charnwood Dynamics). A single Codomotion CX1048 infra-red camera station was used to capture the three-dimensional coordi-nates of markers on the subject’s lower limbs. To minimize markermovement artifacts that come from direct contact to subjects’ skin,custom-made clusters were used to track the markers. Each clusterconsisted of four Codamotion active markers attached to custom-made Aquaplast shells, where the marker locations could be adjustedfor each subject to minimize marker dropout.

Muscle activity was recorded differentially from the tibialis ante-rior, medial gastrocnemius, medial hamstrings, vastus medialis, rectus

1 It should be noted that the “older” subjects tested in this study wereextremely fit, and they often walk on a treadmill as part of their normalexercise routine.

Fig. 1. A: ADAL treadmill manufactured by Techmachine(Andrezieux Boutheon, France). The split belt treadmill sitson top of 8 tri-axial Kistler force sensors, 4 per half-treadmill,labeled 1–4 in B. See text for detailed description. Notepicture is not to scale.

748 BIOMECHANICS OF TREADMILL WALKING

J Appl Physiol • VOL 104 • MARCH 2008 • www.jap.org

on March 12, 2008

jap.physiology.orgD

ownloaded from


femoris, adductor longus, and gluteus maximus and medius muscleson the left leg using a Bagnoli-8 EMG system (Delsys, Boston, MA).All force and EMG data were antialias filtered at 500 Hz beforesampling at 1,000 Hz using a 16-bit data acquisition board (Measure-ment computing, PCI-DAS 6402, Middleboro, MA) and customsoftware (Mathworks, Natick, MA). Marker position data from Co-damotion were sampled at 100 Hz. Force plate data were furtherlow-pass filtered using a zero-delay fourth-order Butterworth filterwith a 25-Hz cutoff frequency.

Protocol

After the subject signed the informed consent form, EMGelectrodes were attached on the subject’s skin by a trained physicaltherapist, after which the marker clusters were strapped on to thesubject’s legs with neoprene (DuPont, Wilmington, DE) and co-band (3M, St. Paul, MN). The neoprene was wrapped tightlyaround the subject’s shanks and thighs, which helped minimizeskin movement artifacts. Marker clusters were placed on thesubject’s feet, shanks and thighs, while four individual markerswere placed on the subject’s pelvis.

With all of the instrumentation in place, subjects were asked towalk overground for a distance of �5 m at their comfortable speed. Ingeneral, �10 trials were necessary to obtain three acceptable passes

with adequate rest breaks in between. The criteria for an acceptabletrial were that the subject made good contact with the center of theforce plate, and minimal marker dropout was observed over a full gaitcycle. The average walking speeds were calculated from the first threeoverground trials, which were then used in the treadmill trials.

After the overground walking trials, subjects were allowed toacclimate to walking on the treadmill for �3 min. During thetreadmill trials, subjects were not allowed to hold onto the handrailsbecause this could possibly alter their gait pattern. After the acclima-tion phase, the speed of the treadmill was set to the speed obtainedduring the overground walking trials and data were collected for 30 s.

On completion of the walking trials, anatomic landmarks weredigitized using the C-Motion digitizing pointer, which establishedrelationships between landmarks and marker locations. Landmarklocations were obtained from both legs for the lateral malleolus,lateral femoral epicondyle, greater trochanter, and the anterior iliaccrest. Additionally, anatomic measures of the foot, ankle, and kneewere taken for each subject to create subject-specific link segmentmodels in Visual 3D (C-Motion, Rockville, MD) (see below).

Analysis

Subject-specific models were created in Visual 3D using theanatomic measures taken for each subject, along with body mass and

Fig. 2. Average ankle, knee, and hip joint moments in sagittal (A, C, E) and frontal (B, D, F) planes for all subjects. Gray line with shading, mean �1 SD foroverground trials; solid black line with 2 dashed lines, mean �1 SD for treadmill trials. Max Pflexor, maximum plantar flexion; Max Dflexor, maximumdorsiflexion; MaxEV1, maximum eversion in early stance; MaxEV2, maximum eversion in late stance; MaxIN, maximum inversion; MaxEX1, maximumextension in early stance; MaxEX2, maximum extension in late stance; MaxFL, maximum flexion; MaxFL1, maximum flexion in mid stance; MaxFL2, maximumflexion in late swing; MaxIN, maximum inversion; MinVal, minimum valgus; MaxVal1, maximum valgus in early stance; MaxVal2, maximum valgus in latestance; MinAB, minimum abduction; MaxAB1, maximum abduction in early stance; MaxAB2, maximum abduction in late stance.

749BIOMECHANICS OF TREADMILL WALKING


on March 12, 2008

jap.physiology.orgD

ownloaded from


height, which were used to normalize joint moments and joint powers,and to define segment masses and inertial values.

The Visual 3D model assumes that each segment is a rigid body,which is then used to calculate joint angles, joint moments, and jointpowers.

For each trial, individual gait cycles were extracted from a datasequence, considered the interval between successive heel strikes inthe same foot. Since only one heel contact was normally made on theforce plate for the overground trials, the subsequent heel strike wasinferred from the kinematic data. Each stride cycle was resampled foraveraging purposes and time normalized, expressed as a percentage ofthe total gait cycle (e.g., 0–100%). All temporal, kinematic, kinetic,and EMG measures were calculated for each stride, which resulted inthree overground trials and �10–20 treadmill strides. To makeequivalent comparisons between the two walking conditions, 3 strideswere randomly selected from the treadmill trials for each subject,which were then compared with the overground trials (see Statistics).

Temporal gait parameters. Temporal gait parameters were esti-mated for treadmill and overground walking conditions. Measuresincluded stance time, total double-limb support time, swing time,stride time, step time, cadence, and stride length (see Ref. 21 for anexplanation of each gait parameter). Speed, cadence, and stride lengthwere normalized using the procedures described by Hof (15) toaccount for individuals of various heights:

Normalized speed: speed �1

�H � g(1)

Normalized cadence: cadence � �H

g(2)

Normalized stride length: stride length �1

H(3)

where H is the subject’s height in meters, and g is gravity (9.81 m/s2).Kinematic analysis. The angular range of motion in the sagittal

plane for the hip, knee, and ankle was found by subtracting theminimum joint angle from the maximum joint angle for overgroundand treadmill trials (1). Peak ankle flexion and extension, knee flexion,and hip flexion and extension were also identified for each trial. Thesevalues, and the phase of the gait cycle in which they occurred, wereused to evaluate the kinematic differences between treadmill andoverground walking.

Kinetic analysis. Joint moments at the ankle, knee, and hip (11, 21,34, 35) were analyzed in the sagittal and frontal planes. For the ankle,the maximum dorsiflexor and plantar flexor moments were calculatedin the sagittal plane, while maximum eversion and the midstanceminimum moments were determined for the frontal plane. For theknee in the sagittal plane, the maximum extension moments generatedin early stance and late stance, along with the maximum flexionmoment were identified. In the frontal plane, the two peak valgusmoments in early and late stance were identified, along with the minimumvalgus moment in midstance. At the hip, sagittal plane maximumextension moments were identified in early stance and late swing,along with the maximum hip flexion moment. For the frontal plane,peak abduction moments in early and late stance were identifiedas well as the minimum in midstance. These metrics are outlined inFig. 2 and described in RESULTS.

Joint powers (11) were analyzed for the sagittal planes, where theminimum and maximum joint powers were estimated for the ankle,knee, and hip throughout the gait cycle (Fig. 3). Ground reactionforces were examined in the vertical, medial-lateral, and anterior-posterior planes, where the maximum and minimum values wereanalyzed in the gait cycle, particularly at transition points (see Fig. 4).

EMG analysis. Muscle activation patterns (16) were processedusing the technique described by Hidler and Wall (14). Briefly, themean EMG pattern for each muscle was rectified and then smoothed

using a 50-point root-mean-square algorithm (19). For each muscle,the smoothed EMG was normalized to the maximum value observedin that respective muscle across all trials so that intersubject compar-isons could be made. After the average EMG profile for each musclewas calculated, the data were broken up into seven phases as de-scribed by Perry (27). Within each of these phases, the integratedEMG activity was calculated for each muscle.

Statistics

Two parallel statistical analyses were performed on the temporalgait parameters, kinematic measures, joint moments, and joint powers.A repeated-measures ANOVA was used to first compare treadmill andoverground walking. As stated above, the three overground strideswere compared with three randomly selected treadmill strides for eachsubject. Temporal parameters compared included normalized walkingspeed, normalized cadence, normalized stride length, stride time,stance time, swing time, and double-limb support time. Kinematicvariables examined were maximum and minimum flexion and exten-sion angles, and range of motion of the ankle, knee, and hip in sagittaland frontal plane. Kinetic measures were compared by looking atanterior-posterior, medial-lateral, and vertical ground reaction forces,joint moments, and joint powers that have been discussed in Kine-matic analysis. Interactions were also looked at across strides andgroups, as well as within groups and between strides. All statisticaltests were run using STATA (Intercooled Stats 9.2 for Windows,Stata).

Fig. 3. Average ankle (A), knee (B), and hip (C) joint power in the sagittalplane for all subjects. Gray line with shading, mean and �1 SD for overgroundtrials; solid black line with 2 dashed lines, mean and �1 SD for treadmill trials.Max, maximum power; Min, minimum power; Min1, first local minimum;Min2, second local minima; Min3, third local minimum; Max1, maximumpower in early stance; Max2, maximum power in late stance.



on March 12, 2008

jap.physiology.orgD

ownloaded from


Because of the large number of comparisons being made betweentreadmill and overground walking, we also analyzed the data de-scribed above using a seemingly unrelated regression (SUR) model,which is an extension of linear regression model. SUR models aresystems of simultaneous equations in which the variables are notindependent. The SUR model tests the effects of a number of inde-pendent variables on each dependent variable by taking into accountthe potential correlation among the error terms. Thus this modelingtechnique allowed us to look for differences between treadmill andoverground walking by implicitly modeling the similarities betweenthe dependent variables (20). For the SUR model analysis, indepen-dent variables consisted of type of walking and stride number.

To analyze the EMG data, we used a fixed-effects regressionanalysis to account for the interrelatedness between the phases of thegait cycle and to model the unobserved factors that arise from themultiple steps and the individual subjects. This model allowed us totest for differences between the treadmill and the overground for eachphase of the gait cycle, assuming that the phases are part of a completemovement rather than analyzing each phase separately across thetypes of walking. Here, fixed factors in the model included type ofwalking, subject, and phase of the gait cycle, while the dependentmeasure was the mean integrated amount of muscle activity (see EMGAnalysis).

RESULTS

Temporal Gait Parameters and Joint Kinematics

Table 1 lists means SD of the temporal parameters for bothoverground and treadmill ambulation. With the exception ofswing time and stance time, none of the temporal gait param-eters were significantly different between treadmill and over-ground ambulation. Comparing joint kinematics in the sagittalplane, only knee range of motion was significantly differentbetween treadmill and overground walking. A summary ofjoint kinematic parameters is listed in Table 2.

Joint Moments

General observation. With the exception of peak ankleplantar flexion, sagittal plane joint moments during treadmilltrials were significantly different than those utilized duringoverground walking. Conversely, joint moments in the frontalplane were not significantly different between treadmill andoverground walking. A detailed listing of the mean and stan-dard deviations of the sagittal plane joint moment featuresoutlined in Kinetic analysis can be found in Table 3.

Ankle moment. A representative example of the ankle mo-ments for treadmill and overground walking can be seen in Fig.2A. During the loading response of the gait cycle (0–10%),subjects had a tendency to produce larger dorsiflexor momentsduring overground walking. Maximum plantar flexor momentsduring terminal stance phase (30–50%) were not significantlydifferent between overground and treadmill walking. For thefrontal plane, neither ankle inversion nor eversion momentswere significantly different between the walking conditions(Fig. 2B).

Fig. 4. Average ground reaction forces (GRF) forall subjects in the anterior-posterior (A), medial-lateral (B), and vertical (C) axes. Gray line withshading, mean and �1 SD for overground trials;Solid black line with 2 dashed lines, mean and �1SD for treadmill trials. MinAP, minimum anterior-posterior GRF; MaxAP, maximum anterior-posteriorGRF; ML Max1, maximum medial lateral GRF inearly stance; ML Max2 maximum medial lateralGRF in late stance; Vmin, minimum vertical GRF;VMax1, maximum vertical GRF in early stance;VMax2, maximum vertical GRF in late stance.

Table 1. Temporal gait parameters for overgroundand treadmill walking

Parameter Overground Treadmill P Value

Walking speed (dimensionless) 0.27 (0.04) 0.28 (0.04) 0.95Step time, s 0.56 (0.05) 0.54 (0.06) 0.41Stance time, s 0.68 (0.07) 0.65 (0.08)* 0.021Swing time, s 0.45 (0.03) 0.43 (0.04)* 0.0017Double-limb support time, s 0.23 (0.05) 0.22 (0.05) 0.77Cadence (dimensionless) 45.1 (4.0) 46.6 (4.5) 0.28Stride length (dimensionless) 0.73 (0.09) 0.71 (0.08) 0.86

Values are means (SD). *Different from overground walking, P � 0.05 (seeStatistics for details).



on March 12, 2008

jap.physiology.orgD

ownloaded from


Knee moment. Throughout the gait cycle, subjects producedlarger knee extensor moments during overground walkingcompared with treadmill walking. As shown in Fig. 2C andsummarized in Table 3, the peak extensor moments in earlyand late stance (e.g., MaxEX1 and MaxEX2) were both sig-nificantly greater during overground walking than on the tread-mill (P � 0.05). The peak flexor moments in late stance (e.g.,MaxFL1) and in late swing (e.g., MaxFL2) were significantlygreater during treadmill walking. In the frontal plane (Fig. 2D),no statistical differences were found in valgus knee momentsbetween walking conditions.

Hip moment. Sagittal plane hip extensor moments in earlystance (e.g., MaxEX1) and late swing (MaxEX2) were bothgreater during treadmill walking than during overground trials(see Fig. 2E), whereas the peak hip flexion moment (MaxFL)was significantly higher (e.g., more negative) during treadmillwalking. In the frontal plane, no statistical differences werefound in joint moments for treadmill or overground walking(Fig. 2F).

Joint Powers

Joint powers for the sagittal plane are reported. As illustratedin Fig. 3A and listed in Table 4, the maximum and minimumjoint powers at the ankle for treadmill and overground trialswere not different. Conversely, significant differences werefound for knee power curves throughout the gait cycle. Figure3B illustrates the mean knee power curves for both treadmilland overground walking, along with the metrics used for

comparison. The first two local minima (e.g., Min1 and Min2)were found to be significantly less for overground walking(e.g., more negative), whereas the third local minimum (Min3)was not different for the two walking conditions. The maxi-mum knee power (Max) was also not significantly differentbetween treadmill and overground walking. For the hip, thefirst maximum (Max1) was significantly greater for treadmillwalking while the local minimum (Min) was significantlygreater (e.g., more negative) for overground walking. Nodifferences were found for the second maximum hip power(Max2) between walking conditions.

Ground Reaction Forces

Because the joint kinematics were similar between treadmilland overground walking yet numerous joint moments andpowers were different, we examined the ground reaction forcesfor each condition in hopes of better understanding thesedifferences. Figure 4 illustrates the ground reaction forces inthe anterior-posterior (A), medial-lateral (B), and verticalplanes (C). We found that the maximum braking forces (la-beled MinAP in Fig. 4A) was significantly greater duringoverground walking than during treadmill walking. No otherground reaction force metrics shown in Fig. 4 were signifi-cantly different.

In addition to comparing absolute measures of ground reac-tion forces, we also examined the magnitude of the variousground reaction forces at the time when metrics used todescribe sagittal plane joint moments were observed (e.g.,those listed in Table 3). When peak dorsiflexor moments wereobserved, the braking forces and the vertical ground reactionforces were significantly higher during overground walking(P � 0.05). We also observed that the propulsion forces weresignificantly lower during overground walking at kneeMaxFL1 and significantly higher during overground walking atknee MaxEX2 (P � 0.05).

Muscle Activity

During treadmill walking, EMG activity in the tibialis ante-rior was lower throughout stance (P � 0.05). Similarly, activ-ity in the gastrocnemius was also lower throughout much ofstance yet slightly higher in terminal swing (P � 0.05). Aninteresting pattern among hamstrings, vastus medialis, andadductor longus emerged between the two conditions. Here,

Table 2. Sagittal plane kinematic measures for treadmilland overground walking


Maximum ankle flexion 13.9 (4.2) 12.2 (4.4) 0.15Maximum ankle extension �12.8 (7.3) �15.3 (7.0) 0.091Ankle range of motion 26.6 (6.2) 27.5 (5.6) 0.84Maximum knee flexion 69.1 (4.3) 67.7 (4.7) 0.48Knee range of motion 67.7 (3.2) 65.6 (3.3)* 0.0037Maximum hip flexion 31.5 (4.0) 31.4 (4.1) 0.97Maximum hip extension �12.6 (4.2) �12.6 (3.5) 0.81Hip range of motion 44.1 (3.6) 44.0 (4.5) 0.96

Values are means (SD) given in degrees. *Different from overgroundwalking, P � 0.05 (see Statistics for details).

Table 3. Sagittal plane joint moments during overgroundand treadmill walking


Maximum dorsiflexor �0.20 (0.09) �0.11 (0.12)* 0.0004Maximum plantar flexor 1.35 (0.22) 1.38 (0.17) 0.46Knee MaxEX1 0.63 (0.27) 0.39 (0.25)* 0.0005Knee MaxFL1 �0.10 (0.16) �0.25 (0.24)* 0.0016Knee MaxEX2 0.34 (0.16) 0.22 (0.18)* 0.0010Knee MaxFL2 �0.23 (0.06) �0.28 (0.06)* 0.0000Hip MaxEX1 0.40 (0.15) 0.57 (0.23)* 0.0000Hip MaxFL �0.75 (0.26) �0.62 (0.30)* 0.0023Hip MaxEX2 0.21 (0.10) 0.31 (0.11)* 0.0013

Values are means (SD) given in N �m/kg. MaxEX1, maximum extension inearly stance; MaxEX2, maximum extension in late stance; MaxFL, maximumflexion; MaxFL1, maximum flexion in early stance; MaxFL2, maximumflexion in late stance. *Different from overground walking, P � 0.05 (seeStatistics for details).

Table 4. Sagittal joint powers during overgroundand treadmill walking


Ankle Minimum �0.84 (0.36) �0.85 (0.32) 0.93Ankle Maximum 3.13 (1.00) 3.24 (1.05) 0.62Knee Maximum 0.56 (0.35) 0.46 (0.19) 0.22Knee Min1 �0.62 (0.44) �0.38 (0.35)* 0.04Knee Min2 �1.74 (0.80) �1.12 (0.87)* 0.001Knee Min3 �0.94 (0.33) �1.14 (0.31) 0.07Hip Max1 0.54 (0.38) 0.79 (0.39)* 0.03Hip Minimum �0.62 (0.27) �0.55 (0.31)* 0.02Hip Max2 0.97 (0.38) 0.95 (0.42) 0.58

Values are means (SD) given in W/kg. Min1, first local minimum; Min2,second local minimum; Min3, third local minimum; Max1, first local maxi-mum; Max2, second local maximum. *Different from overground walking, P� 0.05 (see Statistics for details).



on March 12, 2008

jap.physiology.orgD

ownloaded from


throughout early and midswing, there was higher activityduring overground walking in each of these muscles yet atterminal swing, this relationship reversed (e.g., significantlymore activity during treadmill walking). For the rectus femoris,there was significantly higher muscle activity when subjectswalked on the treadmill during the transition from stance toswing (e.g., phases 4 and 5) as well as at terminal swing. Table 5summarizes integrated muscle activation parameters for tread-mill and overground walking conditions.

DISCUSSION

Although numerous studies have compared treadmill andoverground walking, there still exists significant debate as tothe differences between the two walking modalities. Our re-sults suggest that when individuals walk on a treadmill, theymodify their muscle activation patterns and subsequently jointmoments and powers while maintaining relatively constantlimb kinematics and spatiotemporal gait parameters.

We did not find any statistical differences in the peakvertical ground reaction forces between treadmill and over-ground walking. This is different than the reports by Whiteet al. (33), who found that for normal and fast walking speeds,the magnitude of the vertical ground reaction forces weregreater on the treadmill during midstance yet lower in latestance. Because those authors evaluated the same events as wedid in this study (e.g., VMax1, VMin, and VMax2 in Fig. 4C),

one possible explanation for the differences in results is thatthey collected their overground data on a different system thanthe treadmill data. We believe that a key strength of this studywas that both overground and treadmill data were collectedfrom the same force plates, which were mounted underneatheach half of the treadmill. This resulted in consistent sensorsbeing used for both walking modalities, as well as the samesurface, both of which could introduce errors into the record-ings.

We also did not find significant differences in medial-lateralground reaction forces. Riley et al. (29) reported significantlydifferent medial-lateral forces at self-selected speeds, whichled to statistical difference in frontal plane joint moments.They claimed that the differences in these forces were in therange of variability. Again, in that study, overground gaitassessment was done in a motion analysis laboratory ratherthan on the treadmill. Because the surfaces of the force plateswere different for the two conditions in their study, whereas inthis study we had a consistent and continuous force platesurface, this may explain the differences with the resultspresented in this study.

When subjects walked on the treadmill, we did find thebraking ground reaction forces at heel contact were less thanoverground walking (Fig. 4, MinAP), which led to smallerankle dorsiflexor moments and smaller knee extensor mo-ments. One possible explanation for this would be if the

Table 5. Means and standard deviations of EMG integration per each phase of gait cycle during overgroundand treadmill walking

Phase Overground Treadmill Phase Overground Treadmill

Tibialis anterior Gastrocnemius1 141.9 (56.7) 105.3 (27.2)* 1 27.3 (20.6) 24.9 (19.7)*2 74.2 (31.9) 66.1 (38.8)* 2 99.9 (79.9) 101.6 (60.3)3 58.6 (19.1) 47.2 (15.6)* 3 226.9 (49.2) 214.7 (67.1)*4 41.8 (18.1) 29.3 (22.3)* 4 17.2 (15.7) 17.0 (12.8)5 132.6 (32.7) 107.1 (24.5)* 5 27.3 (26.4) 22.2 (17.7)6 57.6 (18.2) 66.0 (19.5) 6 19.4 (21.0) 21.6 (23.6)7 162.0 (59.7) 154.9 (46.5) 7 23.0 (17.6) 25.5 (19.0)*

Hamstrings Vastus medialis1 63.4 (34.9) 44.6 (30.4)* 1 119.0 (56.5) 125.4 (42.5)2 47.5 (23.9) 48.5 (17.8) 2 87.4 (55.3) 77.9 (28.1)3 65.8 (43.8) 44.6 (8.6) 3 46.8 (21.4) 39.0 (18.9)4 24.5 (27.3) 19.9 (12.1) 4 26.4 (14.8) 24.7 (10.5)5 23.0 (18.6) 15.2 (7.1)* 5 24.9 (10.8) 23.5 (11.0)*6 46.7 (18.9) 36.8 (15.1)* 6 27.3 (16.1) 23.6 (12.6)*7 105.5 (17.4) 123.9 (24.4)* 7 65.1 (32.2) 109.0 (44.0)*

Rectus femoris Gluteus maximus1 111.5 (78.5) 120.8 (58.1) 1 128.4 (50.5) 124.3 (23.6)2 100.5 (86.5) 93.2 (73.1) 2 84.0 (38.1) 96.6 (40.6)3 54.8 (49.6) 52.2 (45.5) 3 60.6 (37.8) 61.2 (29.6)4 37.1 (20.7) 47.6 (25.3)* 4 32.6 (16.0) 33.8 (16.8)5 39.4 (20.2) 66.0 (40.2)* 5 39.4 (17.6) 49.8 (25.8)*6 26.9 (20.1) 27.2 (18.9) 6 30.9 (17.6) 32.5 (16.6)7 56.7 (46.7) 72.9 (39.5)* 7 110.9 (33.7) 112.0 (23.0)

Gluteus medius Adductor longus1 120.7 (38.2) 119.1 (22.9) 1 111.4 (87.6) 92.6 (42.2)*2 86.6 (36.5) 94.5 (44.8) 2 85.9 (38.1) 106.8 (54.1)*3 52.7 (37.0) 48.4 (25.4) 3 77.3 (61.7) 77.7 (63.0)4 25.4 (18.0) 24.2 (16.7)* 4 67.0 (36.8) 65.8 (56.2)5 25.7 (17.0) 28.1 (19.6)* 5 125.7 (116.5) 59.5 (37.0)*6 25.6 (18.0) 25.5 (18.5) 6 86.2 (47.5) 62.4 (35.2)*7 87.0 (31.5) 79.1 (22.1) 7 127.1 (65.6) 152.2 (77.8)*

Values are means (SD). 1, loading response (0–10%); 2, midstance (10–30%); 3, terminal stance (30–50%); 4, preswing (50–60%); 5, initial swing (60–75%);6, midswing (73–85%); 7, terminal swing (85–100%); Different from overground walking, P � 0.05 (see Statistics for details).



on March 12, 2008

jap.physiology.orgD

ownloaded from


treadmill belt slowed briefly at heel contact due to excessiveloads placed on the treadmill motors. In looking at the outputof the treadmill motor encoders, we did observe speed de-creases that reached a maximum of 2.5% of the referencespeed. Although this speed fluctuation may appear to be small,it could result in the slight decreases in braking forces observedin the study. In addition, this speed fluctuation could also beresponsible for the attenuation in vertical ground reactionforces when subjects produced a maximum dorsiflexor mo-ment.

In contrast to Riley et al. (29), our results demonstrate thatthere are in fact significant differences in sagittal plane jointmoments and powers between treadmill and overground walk-ing. They found 15 of 18 joint moment maximums and 3 of 6power maximums, each in stance, were different but concludedthat because the differences were in the range of repeatabilityof kinematic measures, the differences should not be consid-ered meaningful. We too found that there was significantvariability in joint moments; however, many of the metrics wecompared (e.g., see Figs. 2 and 3) were still significantlydifferent. In fact, these differences in joint moments andpowers are further supported by the differences we observed inmuscle activation patterns in numerous muscles (see Statis-tics).

Our original aim was to determine whether there are differ-ences between walking on a treadmill compared with walkingoverground, in terms of kinematics, kinetics, and muscle acti-vation patterns. While we found that there were differences injoint moments, joint powers, and EMGs, the key question nowis why are these behaviors different? Based on the study byVan Ingen Schenau (31), if the treadmill belt speed remainsconstant, there should, in theory, be no differences in thedynamic behavior between the two conditions. We believe thatthere are a number of factors that probably violate this theo-retical law. First, most studies have compared walking on thetreadmill with overground walking using different force plates(e.g., the force plates under the treadmill and the gait labora-tory force plate). This could introduce mechanical errors thatthe subject may need to compensate for. However, in ourstudy, this was controlled for because the same surface wasused in both walking conditions. Second, if the treadmill beltspeed does in fact change, then the acceleration patterns of thelimbs under the two conditions would change, requiring adifferent control strategy to maintain constant kinematics.Similar to previous studies (29, 30), we did observe a slightdrop in belt speed at heel contact (e.g., �2.5%) which canexplain the differences in the dorsiflexor moments we observedin early stance. However, the differences in knee and hipmoments observed later in the gait cycle would presumably notbe affected by this issue.

Perhaps the most likely reason we observed differencesbetween the two walking conditions is that optic flow subjectsreceive on the treadmill is in stark contrast to what they receivewhile walking overground. Studies have shown that visionplays a role in gait (32) such that different optic flow patternsmay alter locomotor control strategies. When walking on thetreadmill, subjects do not receive the same optic flow as theydo when walking overground, which may alter their balanceand stability or their perception of where they are on thetreadmill or the speed at which they are ambulating. Thesefactors are supported by previous studies, because Regnaux

et al. (28) found that treadmill ambulation was not an auto-mated task. Since the optic flow patterns are different betweentreadmill and overground walking, small perceptual changes atone stage of the gait cycle could lead to altered joint momentsand consequently to a cascade of changes during the remainderof the gait cycle if the goal is to preserve their kinematicpatterns, which is what we hypothesize. This hypothesis isfurther supported by Carollo and Matthews (6) and Ivanenkoet al. (17, 18), who suggested that during human locomotion,the kinematics appear to be the desired control variable. Sat-isfying kinematic demands during functional tasks has alsobeen reported in primates (4, 7).

Our findings suggest that while temporal gait parameters andkinematic patterns are similar between treadmill and over-ground walking, the muscle activation patterns and joint mo-ments and powers used to achieve these movement patterns areoften different. Although such effects may lead to differencesin motor strategies, from a therapeutic perspective, the overallkinematic and muscle activation patterns appear to be similarenough that training individuals with neurological injuries(e.g., stroke and spinal cord injury) on a treadmill appearsjustified. Because walking at home and society often requiresindividuals to navigate obstacles and alter their strategies, itcould be viewed that the slight differences we observed herecould be beneficial to the individual being able to adapt todifferent environments.

ACKNOWLEDGMENTS

We thank Mihriye Mete for helping with statistical analysis, Scott Selbie atC-Motion for suggesting a visual 3D model and all subjects who volunteeredfor the study.

REFERENCES

1. Alton F, Baldey L, Caplan S, Morrissey MC. A kinematic comparisonof overground and treadmill walking. Clin Biomech 13: 434–440, 1998.

2. Arsenault AB, Winter DA, Marteniuk RG. Treadmill versus walkwaylocomotion in humans: an EMG study. Ergonomics 29: 665–676, 1986.

3. Belli A, Bui P, Berger A, Geyssant A, Lacour JR. A treadmill ergometerfor three-dimensional ground reaction force measurement during walking.J Biomech 34: 105–112, 2001.

4. Bizzi E, Tresch MC, Saltiel P, d’Avella A. New perspectives on spinalmotor Systems. Nat Rev Neurosci 1: 101–108, 2000.

5. Boda WL, Tapp W, Findley TF. Biomechanical comparison of treadmilland overground walking. Eighth Biennial Conf, Canadian Soc for Biome-chanics, Calgary, 1994, p. 88–89.

6. Carollo JJ, Matthews D. Strategies for clinical motion analysis based onfunctional decomposition of the gait cycle. Phys Med Rehabil Clin N Am13: 949–977, 2002.

7. d’Avella A, Saltiel P, Bizzi E. Combinations of muscle synergies in theconstruction of a natural motor behavior. Nat Neurosci 6: 300–308, 2003.

8. Dingwell JB, Cusumano JP, Cavanagh PR, Sternad D. Local dynamicstability versus kinematic variability of continuous overground and tread-mill walking. J Biomed Eng 123: 27–32, 2001.

9. Dobkin B, Apple D, Barbeau H, Basso M, Behrman A, Deforge D,Ditunno J, Dudley G, Elashoff R, Fugate L, Harkema S, Saulino M,Scott M. Weight-supported treadmill vs over-ground training for walkingafter acute incomplete SCI. Neurology 66: 484–493, 2006.

10. Field-Fote EC. Spinal cord control of movement: implications for loco-motor rehabilitation following spinal cord injury. Phys Ther 80: 477–484,2000.

11. Harris GF, Smith PA. Human Motion Analysis: Current Applicationsand Future Directions. Piscataway, NJ: IEEE Press, 1996.

12. Hesse S, Bertelt C, Schaffrin A, Malezic M, Mauritz KH. Restorationof gait in nonambulatory hemiparetic patients by treadmill training withpartial body weight support. Arch Phys Med Rehabil 75: 1087–1093,1994.



on March 12, 2008

jap.physiology.orgD

ownloaded from


13. Hesse S, Helm B, Krajnik MS, Gregoric M, Mauritz KH. Treadmilltraining with partial body weight support: Influence of body weight releaseon the gait of hemiparetic patients. J Neurol Rehabil 11: 15–20, 1997.

14. Hidler JM, Wall AE. Alterations in muscle activation patterns duringrobotic-assisted walking. Clin Biomech 20: 184–193, 2005.

15. Hof AL. Scaling gait data to body size. Gait Posture 4: 222–223, 1996.16. Inman VT, Ralston HJ, Todd F. Introduction in Human Walking.

Baltimore, MD: Williams & Wilkins, 1981.17. Ivanenko YP, Poppele RE, Lacquaniti F. Five basic muscle activation

patterns account for muscle activity during human locomotion. J Physiol556: 267–282, 2004.

18. Ivanenko YP, Poppele RE, Lacquaniti F. Motor control programs andwalking. Neuroscientist 12: 339–348, 2006.

19. Kenney JF, Keeping ES. Root mean square In: Mathematics of Statistics.Princeton, NJ: Van Nostrand, 1962, pt. 1, p. 59–60.

20. Kennedy PA. Guide to Econometrics (4th ed.), Cambridge: MIT Press,1998.

21. Kirtley C. Clinical Gait Analysis: Theory and Practice. London:Churchill Livingston, 2006.

22. Kram R, Griffin TM, Donelan JM, Chang YH. A Force-treadmill formeasuring vertical and horizontal ground reaction forces. J Appl Physiol85: 764–769, 1998.

23. Moritz CT, Farley CT. Passive dynamics chance leg mechanics for anunexpected surface during human hopping. J Appl Physiol 97: 1313–1322,2004.

24. Murray MP, Spurr GB, Sepic SB, Gardner GM, Mollinger LA.Treadmill vs floor walking: kinematics, electromyogram, and heart rate.J Appl Physiol 59: 87–91, 1985.

25. Nymark JR, Balmer SJ, Melis EH, Lemair ED, Millar S. Electromyo-graphic and kinematic nondisabled gait differences at extremely slow

overground and treadmill walking speeds. J Rehabil Res Dev 42: 523–534,2005.

26. Pearce ME, Cunningham DA, Donner AP, Rechnitzer PA, FullertonGM, Howard JH. Energy cost of treadmill and floor walking at self-selected paces. Eur J Phys 52: 115–119, 1983.

27. Perry J. Gait Analysis: Normal and Pathological Function. Thorofare,NJ: Slack, 1992.

28. Regnaux JP, Robertson J, Smail DB, Daniel O, Bussel B. Humantreadmill walking needs attention. J Neuroeng Rehabil 3: 2006.

29. Riley PO, Paolini G, Croce UD, Paylo KW, Kerrigan DC. A Kinematicand kinetic comparison of overground and treadmill walking in healthysubjects. Gait Posture 2006;doi:10.1016/j. gaitpost.2006.07.003.

30. Savelberg HHCM, Vorstenbosch MATM, Kamman EH, WeijerJGW, Schambardt HC. Intra-stride belt-speed affects treadmill locomo-tion. Gait Posture 7:26–34, 1998.

31. Van Ingen Schenau GJ. Some fundamental aspects of the biomechanicsof overground versus treadmill locomotion. Med Sci Sports Exerc 12:257–261, 1980.

32. Warren WH, Kay BA, Zosh WD, Duchon AP, Sahuc S. Optic flow isused to control human walking. Nat Neurosci 4: 213–216, 2001.

33. White SC, Yack HJ, Tucker CA, Lin HY. Comparison of verticalground reaction forces during overground and treadmill walking. Med SciSports Exerc 30: 1537–1542, 1998.

34. Winter DA. Biomechanical motor patterns in normal walking. J MotorBehav 15: 302–330, 1983.

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

36. Zellner A. An efficient method of estimating seemingly unrelated regres-sions and tests for aggregation bias. J Am Stat 57: 348–68, 1962.



on March 12, 2008

jap.physiology.orgD

ownloaded from


Documents

Biomechanics of overground vs. treadmill walking in ...cabrr.cua.edu/res/docs/publications/jap2008.pdf · Biomechanics of overground vs. treadmill walking in healthy individuals Song