Gait & Posture 37 (2013) 402–407

A comparison of the movement characteristics between the kneeling gait andthe normal gait in healthy adults

Taichi Kurayama a,b,*, Yusuke Tadokoro a, Shuhei Fujimoto a, Zen Komiya b, Susumu Yoshida d,Sudesna Chakraborty b, Daisuke Matsuzawa b,c, Eiji Shimizu b,c, Kunitsugu Kondo a, Yohei Otaka a,e

a Tokyo Bay Rehabilitation Hospital, Chiba, Japanb Research Center for Child Mental Development, Graduate School of Medicine, Chiba University, Chiba, Japanc Department of Cognitive Behavioral Physiology, Chiba University Graduate School of Medicine, Chiba, Japand Division of Physical Therapy, Department of Rehabilitation Sciences, Faculty of Health Care Sciences, Chiba Prefectural University of Health Sciences, Japane Department of Rehabilitation Medicine, Keio University School of Medicine, Tokyo, Japan

A R T I C L E I N F O

Article history:

Received 27 March 2012

Received in revised form 8 August 2012

Accepted 17 August 2012

Keywords:

Hip strategy

Center of mass displacement

Energy efficiency

Stroke rehabilitation

COM trajectory

A B S T R A C T

Background: Trainings of the kneeling position, such as standing exercise on the knees and kneeling gait,

have been anecdotally used in physical therapy to improve postural control of patients with various

pathological conditions. However, clinical evidence is lacking and the movement characteristics of these

kneeling trainings have not been well explored. The purpose of this study is to clarify the movement

characteristics of the kneeling gait compared with the normal gait.

Methods: Twenty healthy volunteers (10 men and 10 women) aged 22–34 years were recruited.

Participants were required to perform the kneeling gait and the normal gait at a self-selected

comfortable speed on the treadmill. Surface electromyograms (EMG) and center of mass (COM)

displacements were measured during each task.

Results: The EMGs of the gait-related proximal muscles during the kneeling gait were greater than

during the normal gait, even at a comfortable speed. The COM displacement to the lateral direction was

longer during the kneeling gait than it was during the normal gait. Furthermore, mechanical energy

efficiency during the kneeling gait was less than that during the normal gait.

Conclusion: The results suggest that the kneeling gait is an effective exercise to strengthen the gait-

related proximal muscles. The increased muscle activities during the kneeling gait were probably due to

the compensatory movements of the trunk and the pelvis.

� 2012 Elsevier B.V. All rights reserved.

Contents lists available at SciVerse ScienceDirect

Gait & Posture

jo u rn al h om ep age: ww w.els evier .c o m/lo c ate /g ai tp os t

1. Introduction

Training on the knees, such as standing on the bilateral knees,standing in the half kneeling position, stepping from the kneelingposition (transition from the kneeling position to the half kneelingposition) and kneeling gait, has been anecdotally used in physicaltherapy. Several textbooks [1–3] introduce these trainings as onetype of the interventions for restoration of the functional mobility.In these textbooks, the kneeling task is listed as an interventionusing developmental sequence postures [1], as an activity toprepare for locomotion [2] and as one of the adjunct treatments fortransitional patterns [3]. Two selective postural strategies areknown for standing balance in normal adult: an ankle strategy

* Corresponding author at: Tokyo Bay Rehabilitation Hospital, 4-1-1 Yatsu,

Narashino-shi, Chiba 275-0026, Japan. Tel.: +81 43 226 2975; fax: +81 43 226 8588.

E-mail address: kurayama@chiba-u.jp (T. Kurayama).

0966-6362/$ – see front matter � 2012 Elsevier B.V. All rights reserved.

http://dx.doi.org/10.1016/j.gaitpost.2012.08.009

(shifts the body’s center of gravity mainly by the ankle jointsmovements) and a hip strategy (repositions the center of gravity bythe hip joints movements.) [4,5]. In kneeling position, however,one cannot use the ankle strategy and has to rely on the hip controlto maintain upright posture. In other words, the kneeling posturesforce one to use the trunk/hip control. This explains the clinicalrelevance of the kneeling trainings for enhancing trunk control andstrengthening the hip stabilizers [1–3].

Despite the fact that kneeling training is frequently used and issupported, there is no clinical evidence that supports theeffectiveness of these training. From these backgrounds, werecently investigated the immediate effects of the kneelingtraining on postural control in stroke patients [6] (article inJapanese) who often have to rely predominantly on hip strategy forcontrolling upright posture due to useless ankle function withdistal weakness [7]. We prescribed 5–10 min kneeling exercises(upright kneeling, stepping exercise in half kneeling and kneelinggait) to fifteen chronic stroke patients and observed the immediate

T. Kurayama et al. / Gait & Posture 37 (2013) 402–407 403

effects on postural control. The results showed that capability ofmoving the center of pressure (COP) to the lateral direction wassignificantly increased after the intervention.

Likewise, there are little reports about the movement analyses ofthe kneeling tasks. Gallagher reported the comparison of trunkmuscle activities between normal standing and the kneelingposition during trunk extension exertions in healthy subjects usingsurface electromyograms (EMG) [8]. The result showed that despiteequivalent trunk muscle activity, reduced extensor capability existsin the kneeling posture compared to normal standing. Mezzaraneand Kohn analyzed the COP in the sagittal plane during standing onknees and during standing on feet [9]. With the model simulation,they found that the differences of COP profile observed between twoconditions in the experiment were resulted from neural processes aswell as the biomechanical factors.

As stated above, clinical evidences as well as basic analyses ofthe kneeling exercises are very limited and have to be explored. Inthis study, we focused on the characteristics of the ‘‘kneeling gait’’and investigated the differences in the muscle activities of the gait-related proximal muscles, center of mass (COM) displacementsand mechanical energy recovery between the kneeling gait and thenormal gait in healthy subjects.

2. Methods

2.1. Subjects

Twenty healthy volunteers (10 men and 10 women), aged 22–34 years (mean 24.0, standard deviation [SD] 4.9), were recruited.None of the participants had any neuromuscular problem asdetermined by a non-structured interview. The study conformedto the Declaration of Helsinki, and informed consent was obtainedfrom all participants according to the procedures of the EthicsCommittee of the Tokyo Bay Rehabilitation Hospital.

2.2. Tasks

Two locomotion tasks on the treadmill were employed: (1) thenormal gait at a self-selected comfortable speed and (2) thekneeling gait at a self-selected comfortable speed (gait on knees,Fig. 1). The participants were randomly divided into two groups.Half of the participants were allocated to the normal gait first and

Fig. 1. Subject on the treadmill during the kneeling gait. Muscle activity was detected b

measurement system was used to measure the center of mass (COM) displacement du

the other half did the kneeling gait first. Before each task, asufficient warm-up period was given to determine a comfortablespeed for each subject. All subjects wore knee pads to reduceexternal knee forces during the kneeling gait.

2.3. Measurements

The gait cycles were determined using the pressure sensors(FSR402, Interlink Electronics Inc, Camarillo, CA) at the samplingrate of 100 Hz. The sensors were attached to the right heel and theright forefoot for the normal gait and to the right tibial tuberosityfor the kneeling gait. Stance phase was defined from the heel striketo the toe off for the normal gait and from the knee on to the kneeoff for the kneeling gait.

Surface EMG on the right side of the erector spinae, the rectusabdominis, the gluteus maximus, the gluteus medius, the rectusfemoris and the semitendinosus were recorded at 1000 Hz by theEMG system (Delsys, Boston, MA) (Fig. 1). Electrodes were placedin accordance with the method described in Perotto (2005) [10].For the erector spinae, the sensor was placed at three-fingers’width laterally to the spinous process 2nd lumbar vertebra. For therectus abdominis, the sensor was placed at two-fingers’ widthlaterally on the abdominal midline at supraumbilical portion. Forthe gluteus maximus, the sensor was placed at the midwaybetween the greater trochanter and the sacrum. For the gluteusmedius, the sensor was placed at one inch distal to the midpoint ofthe iliac crest. For the rectus femoris, the sensor was placed on theanterior aspect of the thigh, midway between the superior borderof the patella and the anterior superior iliac spine. For thesemitendinosus, the sensor was placed midway on a line betweenthe medial epicondyle of the femur and the ischial tuberosity.

The COM displacement was evaluated using an optical markerat the spinous process of the third lumbar vertebra [11,12]. Motionof an optical marker was recorded at 100 Hz with the motioncapture system (NDI Optotrak, Waterloo, Ontario, Canada) (Fig. 1).

All measurements were started at 15 s after beginning each taskand taken for 1 min.

2.4. Data analysis

The basic gait parameters for speed [m/min], step length [m],cadence [steps/min], stance time and swing time [s] were

y a surface electromyogram (EMG) connected to an acquisition system. An optical

ring the task.

T. Kurayama et al. / Gait & Posture 37 (2013) 402–407404

obtained from the settled treadmill speed and gait cycles. TheEMG and COM data were analyzed using MATLAB software(R2011b, The Math Works Inc. Natick, MA). The raw EMG datawere low-pass filtered at 450 Hz and high-pass filtered at 10 Hzto eliminate movement artifacts, using Butterworth fourth-orderzero-lag filters. The signal was then full-wave rectified and low-pass filtered again at 10 Hz to obtain a linear envelope. Aftereach wave was normalized to the percentage of the gait cycle,the grand average waveform of 30 steps from all participants ineach of the 6 muscles was calculated to evaluate the muscleactivation patterns. For the statistical analyses, the root meansquare (RMS) of the EMG signal of 30 gait cycles for each subjectwas obtained to quantify the EMG signal. Then, the obtainedRMS for each muscle was divided for the stance phase and theswing phase. The raw COM data were low-pass filtered at 30 Hzusing Butterworth fourth-order zero-lag filters to detect themaximum and the minimum values of each vertical and lateraldirection. Using averaged COM data of 30 time-normalizedconsecutive strides, the mechanical energy exchange ratio(%recovery) was calculated as follows.

(1) The total instantaneous kinetic energy (Ek; in J) of the COM wascalculated by the following equation:

Ek ¼1

2mV2 (1)

where m is the body mass (kg) and V is the resultant COM

velocity (m/s) determined from its vertical, anteroposterior,

and mediolateral components.(2) The potential energy (Ep; in J) of the COM was calculated as

follows:

Ep ¼ mgh (2)

where g is the gravitational constant and h is the vertical

position of the COM. The total mechanical energy at a specific

time (Etot; in J) was calculated as the sum of the Ek and Ep.(3) The amount of %recovery due to pendulum-like energy transfer

between the gravitational potential energy and the kineticenergy was calculated according to Cavagna et al. [13], Peyrotet al. [12], and Schepens et al. [14].

Table 1The mean values (� [SD]) of the basic gait parameters, RMS, COM displacement and %rec

Gait speed [m/min]

Step length [m]

Cadence [steps/min]

Stance time [s]

Swing time [s]

RMS [mV] stance phaseswing phase Rectus abdominis

Erector spinae

Gluteus maximus

Gluteus medius

Semitendinosus

Rectus femoris

COM displacement [cm] Vertical range

Lateral range

%recovery [%]

RMS [mV]: root mean square of the EMG signal. COM: center of mass. n.s.: no significa* Statistically significant differences between the kneeling gait and the normal gait (

%recovery ¼ 100 �Wk þ W p � Wext

Wk þ W p(3)

where Wext (J kg�1 m�1) indicates the external mechanical workperformed to lift and accelerate the COM and was calculated as thesum of the positive increments in Etot, Wk (J) and Wp (J) are the sumof the positive increments in Ek and Ep, respectively.

2.5. Statistical analysis

The variables of the basic gait parameters, the COM displace-ments and the %recovery were analyzed using the paired t-test. TheRMS values in the stance and in the swing phase were analyzedusing the Wilcoxon signed-rank test. The level of statisticalsignificance was set at 0.05. Statistics were performed using SPSS19.0 software (SPSS Inc., Chicago, IL).

3. Results

3.1. Basic gait parameters

Table 1 shows the values of the basic gait parameter in thekneeling gait and the normal gait. The gait speed and the steplength in the kneeling gait were about one-third of those in thenormal gait, and there was a significant statistical difference(p < 0.001). The stance time, the swing time and the cadence of thekneeling gait were almost equal to that of the normal gait.

3.2. Muscle activity

Fig. 2 shows the raw data of the EMG and the foot sensors of onesubject during the normal gait (a) and the kneeling gait (b). TheRMS values of all muscles, both in the stance and the swing phases,were greater during the kneeling gait than those in the normal gait(Table 1). In the stance phase, statistically significant differenceswere observed between the kneeling gait and the normal gait inthe erector spinae (p = 0.011), the semitendinosus (p < 0.001) andthe rectus femoris (p < 0.001). In the swing phase, statisticallysignificant differences between the tasks were observed in all themuscles except for the gluteus maximus (the rectus abdominis,p = 0.028; the erector spinae, p < 0.001; the gluteus medius,p = 0.002; the semitendinosus, p = 0.006; rectus femoris,

overy in the kneeling gait and the normal gait.

Normal gait Kneeling gait p value

58.83 � 7.84 22.83 � 3.43 <0.001*

0.56 � 0.06 0.22 � 0.04 <0.001*

103.54 � 8.34 102.20 � 12.30 0.545 n.s.

0.71 � 0.08 0.74 � 0.14 0.559 n.s.

0.44 � 0.07 0.46 � 0.08 0.568 n.s.

4.20 � 0.42 6.90 � 1.17 0.156 n.s.

3.74 � 0.34 6.88 � 1.40 0.028*

8.60 � 1.22 18.64 � 2.57 0.011*

6.31 � 0.89 16.71 � 2.45 <0.001*

7.42 � 1.46 10.71 � 1.73 0.167 n.s.

5.55 � 0.83 7.98 � 1.08 0.145 n.s.

19.77 � 2.85 22.63 � 3.35 0.391 n.s.

9.54 � 1.13 17.45 � 2.78 0.002*

9.23 � 1.27 21.08 � 2.53 <0.001*

13.15 � 1.90 23.56 � 2.68 0.006*

7.20 � 1.02 12.05 � 1.70 <0.001*

6.38 � 0.96 10.61 � 1.78 0.007*

5.6 � 2.1 5.1 � 1.9 0.178 n.s.

9.7 � 1.6 17.3 � 3.1 <0.001*

65.9 � 4.9 60.3 � 7.0 0.009*

nt differences.

p < 0.05).

Fig. 2. Raw EMG data and foot sensor data of one individual subject during the normal gait (a) and the kneeling gait (b). Longitudinal scales were equal in both conditions in

each muscle.

T. Kurayama et al. / Gait & Posture 37 (2013) 402–407 405

p = 0.007). The patterns of the muscle activity in each muscle areshown in Fig. 3. The rectus abdominis, the gluteus maximus, therectus femoris and the semitendinosus showed similar peakpatterns in both tasks. However, there were large peaks of muscleactivity in the erector spinae, the gluteus medius and thesemitendinosus from the terminal stance to the initial swingphase.

3.3. Motion analysis

Fig. 4 shows the grand average of the COM displacements of 30normalized gait cycles during the normal gait (Fig. 4a) and thekneeling gait (Fig. 4b). The COM displacement to the lateraldirection was significantly greater in the kneeling gait than that inthe normal gait (p < 0.001). The %recovery during the kneeling gaitwas significantly less than that in normal gait (p = 0.009) (Table 1).

Fig. 3. Grand average waveforms of the EMG signals in each 6 muscles. This figure depicts

line). Vertical lines on each waveform indicate the standard deviations (�SD). The averag

indicate the beginning of the swing phase.

4. Discussion

The muscle activity of the gait-related proximal muscles inthe kneeling gait was significantly greater than that in thenormal gait, despite the fact that the comfortable speed of thekneeling gait was about one-third of the normal gait. Followingfactors are hypothesized to be the reasons for increased muscleactivity during the kneeling gait. First, muscle activities relatedto keeping the clearance are thought to be a major factor. It maybe much more difficult to keep clearance in kneeling gait than inthe normal gait. In fact, the increased COM displacement to thelateral direction observed in the kneeling gait might reflect thestrong efforts to keep clearance by compensatory trunk bendingand/or rotating the pelvis. Subjects had to lift their legs withflexed knees to obtain clearance in the swing phase without theankle and the knee joint motion. Strong contractions of the

the mean EMG curves in both the kneeling gait (dark line) and the normal gait (thin

ed EMG was time normalized to the percentage of the gait cycle. Vertical dotted lines

Fig. 4. The center of mass (COM) trajectory of all 20 subjects in the normal gait (a) and the kneeling gait (b). Each trajectory represented an average of the COM trajectory in

each subject. The numbers beside the axis stand for the length of scale [m]. The position of each trajectory was positioned so that the center points in the vertical direction,

lateral direction and anteroposterior direction is ‘‘0’’.

T. Kurayama et al. / Gait & Posture 37 (2013) 402–407406

erector spinae, the semitendinosus and the gluteus mediusmight be required for this movement from the terminal stance tothe initial swing phase. Second, a forced leg swing was requiredbecause the pendulum-like movement of the lower limb wasreduced by the shortened pendulum length and by the frictionbetween the leg and the treadmill. Third, the decreased%recovery suggested that the loss of energy exchange betweenthe potential energy to the kinetic energy is greater in thekneeling gait than that in the normal gait. Therefore, subjectsneeded to produce stronger driving force in the kneeling gaitthan in the normal gait. Fourth, according to Gallagher [8], bothisometric and isokinetic back strengths are significantly reducedin the kneeling posture compared with that achieved in astanding position. Nevertheless, equivalent trunk muscle activi-ty was detected simultaneously by the EMG recording. Thisresult suggests that the trunk muscles have to produce astronger activation in the kneeling position than in the standingposition to obtain an extension torque equivalent to that of thestanding position. Finally, in the kneeling position, the proprio-ceptive inputs from the soles and the ankle are useless forcontrolling balance. Mezzarane and Kohn [9] indicated thatpostural control in the kneeling gait has to rely on the availablesomatosensory inputs coming from the structures associatedwith the knee joint (e.g., thigh muscles) in addition to the visualand vestibular inputs. The kneeling posture is an unnaturalposition, meaning that balance control is poorer than that of thestanding gait. Hence, more strong muscle activation seems to berequired for posture control in the kneeling posture.

The COM displacement in the vertical direction was similarfor both tasks, although the COM displacement in the lateraldirection was greater in the kneeling gait as stated above. Gait ischaracterized by a pendulum-like mechanism in which potentialenergy is transformed into kinetic energy and vice versa [13,14].The up-and-down motion of the COM during gait seems to storemechanical energy and to contribute to the suppression ofenergy consumption [15]. The action of the ankle is important

for the smooth up-and-down motion of the COM during gait[16–18]. Since the ankle motion was not usable and the length ofthe pendulum shortened during the kneeling gait, the motion ofthe COM in the vertical direction was assumed to be disturbed.However, no significant difference between the tasks wasobserved in this study. This finding suggests that subjects seemto make some compensatory movements to control the height ofthe COM. One possible explanation is that the compensatoryefforts to keep the leg clearance by bending the trunk and/orrotating the pelvis, resulted in increased lateral COM displace-ment, eventually raised the COM adequately.

Increased muscle activities observed in the keeling gait is notdue to the sustained increases in the baseline muscle activities,but rather is due to the increased peak activity pattern. Thestance time, swing time and cadence of the kneeling gait areapproximately equal to those of the normal gait. Moreover, peakpatterns of muscle activities in terms of timing are approxi-mately similar between two tasks. These results suggest that thekneeling gait has the potential to be employed as an effectivemuscle strength training that reflects a gait muscle activitypattern.

There are several advantages to the kneeling gait. First, one canremove the undesirable effects from the impaired ankle strategythat inevitably influence the other joints. Second, the risk of fallingis lower, since the speed is slower and the height is less in thekneeling gait than in the normal gait. However, some risk also existin kneeling gait. Jensen et al. [19] pointed out that a strong externalforce is applied to the knees during working in the kneelingposition. Continuing the upright kneeling position for more than15 min can cause low back pain [20]. These risks should becarefully before clinical applications.

There are some limitations in the present study. A moredetailed movement analysis is needed to reveal the relationshipsbetween the muscle activation and the trunk motion. In thisstudy, gait tasks are performed on the treadmill. It is importantto note that some biomechanical differences have been reported

T. Kurayama et al. / Gait & Posture 37 (2013) 402–407 407

between the treadmill gait and the ground gait [21–23]. Inaddition, we did not measure the EMG activities duringmaximum voluntary contraction (MVC) of each muscle, as theaim of this study is simply to compare the EMG activitiesbetween the normal gait and the kneeling gait. Furtherinvestigation on %MVC is needed to assess the magnitude ofthe relative increase in muscle activities.

5. Conclusion

We investigated the differences in the muscle activity of thegait-related proximal muscles and the COM displacement betweenthe kneeling gait and the normal gait in healthy subjects. Themuscle activity during the kneeling gait was greater than that ofthe normal gait even at a comfortable speed. This result indicatedthat the kneeling gait is an effective exercise for strengthening thewalk-related proximal muscles.

Acknowledgment

This project was funded by a grant from the KAKENHI(23931048).

References

[1] Dreeben O. Physical therapy clinical handbook for PTAs. Jones & Bartlett; 2007.[2] Gillen G. Stroke rehabilitation: a functional-based approach, 3rd ed., Mosbyr;

2010.[3] Umphred D, Carlson C. Neurorehabilitation for the Physical Therapist Assis-

tant. Slack Inc.; 2006.[4] Horak FB, Shupert CL, Mirka A. Components of postural dyscontrol in the

elderly: a review. Neurobiology of Aging 1989;10:727–38.[5] Nashner L, McCollum G. The organization of human postural movements: a

formal basis and experimental synthesis. Behavioral and Brain Sciences1985;8:135–72.

[6] Kurayama T, Kagehara A, Murakoshi T, Yamaguchi T, Kondo K, Otaka Y.Immediate effect of kneeling exercise on standing balance in stroke patients.Sogo Rihabiriteshon 2011;39:791–4.

[7] de Oliveira CB, de Medeiros IR, Frota NA, Greters ME, Conforto AB. Balancecontrol in hemiparetic stroke patients: main tools for evaluation. Journal ofRehabilitation Research and Development 2008;45:1215–26.

[8] Gallagher S. Trunk extension strength and muscle activity in standing andkneeling postures. Spine 1997;22:1864–72.

[9] Mezzarane RA, Kohn AF. Postural control during kneeling. Experimental BrainResearch 2008;187:395–405.

[10] Perotto A, Delagi E, Iazzetti J, Morrison D. Guide for the electromyographer: thelimbs and trunk, 4th ed., Charles C. Thomas; 2005.

[11] Meichtry A, Romkes J, Gobelet C, Brunner R, Muller R. Criterion validity of 3Dtrunk accelerations to assess external work and power in able-bodied gait. Gaitand Posture 2007;25:25–32.

[12] Peyrot N, Thivel D, Isacco L, Morin JB, Duche P, Belli A. Do mechanical gaitparameters explain the higher metabolic cost of walking in obese adolescents?Journal of Applied Physiology 2009;106:1763–70.

[13] Cavagna GA, Thys H, Zamboni A. The sources of external work in level walkingand running. Journal of Physiology 1976;262:639–57.

[14] Schepens B, Bastien GJ, Heglund NC, Willems PA. Mechanical work andmuscular efficiency in walking children. The Journal of Experimental Biology2004;207:587–96.

[15] Gordon KE, Ferris DP, Kuo AD. Metabolic and mechanical energy costs ofreducing vertical center of mass movement during gait. Archives of PhysicalMedicine and Rehabilitation 2009;90:136–44.

[16] Hansen AH, Childress DS, Miff SC. Roll-over characteristics of human walkingon inclined surfaces. Human Movement Science 2004;23:807–21.

[17] Kerrigan DC, Della Croce U, Marciello M, Riley PO. A refined view of thedeterminants of gait: significance of heel rise. Archives of Physical Medicineand Rehabilitation 2000;81:1077–80.

[18] Kuo AD. The six determinants of gait and the inverted pendulum analogy:A dynamic walking perspective. Human Movement Science 2007;26:617–56.

[19] Jensen LK, Rytter S, Bonde JP. Exposure assessment of kneeling work activitiesamong floor layers. Applied Ergonomics 2010;41:319–25.

[20] Harkness EF, Macfarlane GJ, Nahit ES, Silman AJ, McBeth J. Risk factors for new-onset low back pain amongst cohorts of newly employed workers. Rheuma-tology 2003;42:959–68.

[21] Alton F, Baldey L, Caplan S, Morrissey MC. A kinematic comparison of over-ground and treadmill walking. Clinical Biomechanics (Bristol Avon)1998;13:434–40.

[22] Riley PO, Paolini G, Della Croce U, Paylo KW, Kerrigan DC. A kinematic andkinetic comparison of overground and treadmill walking in healthy subjects.Gait and Posture 2007;26:17–24.

[23] Watt JR, Franz JR, Jackson K, Dicharry J, Riley PO, Kerrigan DC. A three-dimensional kinematic and kinetic comparison of overground and treadmillwalking in healthy elderly subjects. Clinical Biomechanics (Bristol Avon)2010;25:444–9.