Teach in #26, Loading Response: What people said...

At a risk of sounding very naive about this, could it be an aim to prevent hyperextension of the knee at the point of heel strike, when the knee is near
 extension, and as seen, the vector is anterior to the knee which would produce a knee extending moment - so the hamstrings are simply contracting to
 counteract this further extension??

 As the hamstrings are active in terminal swing, could this either be voluntary or involuntary overflow into IC of the next cycle?

 Once the vector is no longer creating an extending moment, the hams then turn off and the quads kick in to eccentrically control the degree of knee
 flexion  in loading.

 Why has it been ignored? No idea!

 Anyhow, those are my simplified thoughts. I await the onslaught of responses!

 Helen Evans
 Research Physiotherapist
 Division of Physiotherapy Education
 Clinical Sciences Building
 City Hospital Campus

 tel: 0115 8404894/80
 e-mail: helen.evans@nottingham.ac.uk

Dear Helen and others,

Good point, but you would then expect the power to be NEGATIVE
(eccentric), whereas it is very definitely positive (concentric). So the
hamstrings are no longer preventing motion by this time, they are
generating it!


Dr. Chris Kirtley MD PhD
Associate Professor
Dept. of Biomedical Engineering
Catholic University of America
620 Michigan Ave NE, Washington, DC 20064
Tel. 202-319-6134,  fax 202-319-4287
Email: kirtleymd@yahoo.com

Hi Chris,

In my opinion the positive Ko burst is real but present only in those subjects who can achieve full knee extension at terminal swing.  The
eccentric contraction of the hamstrings to control end range knee flexion is not always effective, resulting in slightly excessive knee
extension past the 4 or 5 degrees of knee flexion that is desirable at initial contact.  In an effort to correct this excessive extension, the
hamstrings fire forcefully such that at initial contact the knee is flexing from a slightly extended position at late terminal swing, applying a
posterior shear force to the force platform at initial contact, leading to the brief anterior shear GRF.  This explains the brief concentric
contraction of the knee flexors at initial contact and early loading response, and the subsequent Ko power generation.  The co-contraction
of the knee flexors and extensors at early loading response is a good thing, since it increases the stability of the knee joint.

We’ve noticed that the sagittal plane knee kinematic curves of many of our normal subjects display a maximum extension valley just prior
to initial contact, and at first blamed it on an inaccurate foot/floor contact recording.  Closer review and a faster sampling rate have
indicated that this valley does indeed occur before initial contact in some subjects, and those are the same subjects who display the Ko
positive power generation.  In addition, many of our patients with CP lack full knee extension at terminal swing, and also have a reduced or
non-existent Ko peak.  Have others seen a similar correlation?

Best regards,


James J. Carollo, Ph.D., P.E.
Director, Center for Gait and Movement Analysis (CGMA)
    The Children's Hospital Denver
Assistant Professor, Rehabilitation Medicine and Orthopaedics Department
    University of Colorado Health Sciences Center
Assistant Research Professor, Engineering Division
    Colorado School of Mines
1056 E. 19th Ave., B476,  Denver, CO  80218
voice: 303-864-5806, fax: 303-864-5815, carollo.james@tchden.org

Thanks, James, but once again, I disagree. The K0 burst occurs AFTER initial contact, not before, so it cannot be part of any
"reach" phenomenon. I also think it's an almost constant finding - not confined to a few individuals. Note that it appears in every lab's
normative data.


To obtain maximal and rapid energy storage (eccentric) a muscle needs to
be both in a contracted state and ideally at a short length. So perhaps
its better to be well-prepared (a mis-placed step will place a high
overload on energy storage) than doing simply just enough to absorb the
average impact.

From Dr Christopher Smith
Head of Biomedical Science BSc
Centre for Applied Biomedical Research, King's College London
4.1 Shepherd's House, Guy's Campus, London Bridge SE1 1UK
Phone/fax 020 7848 6301, Biomed Office 020 7848 6400
If need be, phone me on 0797 0713507


Could the effect of concentric hamstring contraction impact the more proximal structures (ie, pelvis and sacrum) rather than the distal one
(knee)?  Vleeming has described that the biceps femoris actually originates from the sacrotuberous ligament rather than from the ischial
tuberosity.  So, would it make sense that by concentrically contacting, a counternutation of the sacrum would be created that would
"release" the sacrum from its maximum stable position as the body is transitioning from single to double support.  It would then nutate (as
the contraction of the hamstring ceases) as double support ends and single support again occurs, permitting sacroiliac joint stability at a
time when it would be required (single limb support).

Howard Dananberg, DPM
21 Eastman Avenue
Bedford, New Hampshire 03110
fax 603-625-9889

Here are some interesting articles to look at in regards to the HS activity
upon loading and its effects on anterior translation (especially in the ACL
deficient pt).  Clinically, addressing impairment in HS power and speed of
contraction are important in the prevention of initial and recurrent ACL
injury following grafting, especially in the athletic patient.

        Hollman JH, Deusinger RH, Van Dillen LR, Matava MJ. Knee joint
movements in subjects without knee pathology and subjects with injured
anterior cruciate ligaments.Phys Ther. 2002 Oct;82(10):960-72

        Ciccotti MG, Kerlan RK, Perry J, Pink M. An electromyographic
analysis of the knee during functional activities, II. The anterior cruciate
ligament-deficient and -reconstructed profiles. Am J Sports Med.
1994;22:651-658. Medline

        Kaalund S, Sinkjaer T, Arendt-Nielsen L, Simonsen O. Altered timing
of hamstring muscle action in anterior cruciate ligament deficient patients.
Am J Sports Med. 1990;18:245-248.

        Lass P, Kaalund S, LeFevre S, et al. Muscle coordination following
rupture of the anterior cruciate ligament: electromyographic studies of 14
patients. Acta Orthop Scand. 1991;62:9-14.

        Limbird TJ, Shiavi R, Frazer M, Borra H. EMG profiles of knee
musculature during walking: changes induced by anterior cruciate ligament
deficiency. J Orthop Res. 1988;6:630-638.

        Shiavi R, Zhang L, Limbird T, Edmonstone M. Pattern analysis of
electyomyographic linear envelopes exhibited by subjects with uninjured and
injured knees during free and fast speed walking. J Orthop Res.

        Steele JR, Brown JM. Effects of chronic anterior cruciate ligament
deficiency on muscle activation patterns during an abrupt deceleration
task.Clin Biomech (Bristol, Avon). 1999 May;14(4):247-57.

John J Fraser, PT, MS

Dear Chris
Well, another challenging teach-in!
I doubt that this is an artifact. The task facing the subject that heel
strikes is to transfer load to the musculature that is designed to
accomodate it ie the extensors of the lower limb, the triceps surae, the
vasti and the glutei. Perhaps, this is best acheived by moving the centre of
pressure forward with some speed after initial contact. The concentric power
burst at the knee may be present to accelerate the first rocker so that the
knee extensors and the triceps surae support the greater loads experienced
by the body later in loading and in midstance.


Adam Shortland PhD, MIPEM, SRCS
One Small Step Gait Laboratory,
Guy's Hospital

Dear CGA members,

Just to begin with: don't bother about impossible muscle actions, this K0 burst is an artifact.

My PhD student Rob Bisseling is just at this time studying this phenomenon. He does not only do the usual inverse
dynamics, be it with an Optotrak operating at a sample rate of 200 Hz, but has also accelerometers affixed to foot and

At heel strike, the foot can have an acceleration of several hundreds m/s2 over periods of some 25 ms. In fact it is a
deceleration, the foot is moved down and backward at 2-3 m/s and is stopped by the floor very abruptly, in 10 - 15 ms.
From the impulse law : F.delta t = mass.velocity this gives forces in the order of 200 N, directed forward/upward (foot
mass is around 1 kg). This is exactly what the force plate detects and what you see in the animation.

In a good inverse dynamics, this force peak should be neatly compensated by the mass x acceleration of the foot, and
no peak should appear in the moment or the power. The problem is that usually the accelerations are calculated from the
position data. Due to the measurement noise, this calculation can only be done when the position data are low-pass
filtered at some 5-10 Hz. In this way the acceleration at heel strike, with frequencies around 40 - 60 Hz, is grossly
underestimated. As a result the sharp impact peak turns up in the knee and ankle moment, while is should not be there.

In fact, when you filter acceleration data at 5 Hz, you should filter your forceplate data with the same filter! Nobody
does that, of course, after having bought an expensive Kistler, recording up to 1000 Hz.

I would have liked to show you some figures, but Rob is just busy with analyzing and writing up this matter, so we better
wait for his paper.

This problem is a nice illustration of the fact that the inverse dynamics method, valuable as it is, has several major
methodological problems, that are far from solved.

Chris, thanks for an interesting problem again.

At Hof
Institute of Human Movement Science
University of Groningen
PO Box 196 9700 AD
The Netherlands
Tel: (31) 50 363 2645 

Dear At and others,

I thought someone might suggest this. I recall that errors from disparity in force plate and kinematic filtering was
discussed some time ago on BIOMCH-L - although I was somewhat shocked to find that it was all of 8 years ago!...


  "...if kinetics are to be derived then it is prefered to have
  the signals temporally synchronized and with the same number of data
  points. Normally, for inverse dynamics derivations of kinetics, we tend to
  'up' the joint marker data (increase the number of data points) using
  various techniques to the GRF sampling rate, so that all the
  useful information remains intact (i.e. in particular, the fairly fast
  force development during foot strike).

  If one were to decrease the GRF samples to that of the video sampling
  rate, one runs the risk that the derived resultant joint moments and forces
  are inaccurate. One can easily imagine this situation when, say  the
  vertical GRF signal has a transient which would be missed (under-estimated)
  if the sampling rate was too low.

  A far more pressing problem with inverse dynamics caluculations is
  the derivation of acceleration from joint marker data, where the
  selection and implementation of smoothing technique may (does) have
   a significant impact on the derived moments and forces.
  Normally we do not have an accelerometer on our subjects to
  validate our smoothing technique, as such we typically over-smooth
  the data  to minimize noise resulting in, I would say, under-estimates
  of the moments and forces."

  Dean Kriellaars, Ph.D.
  Human Perfromance Laboratory
  Assoc. Professor, School of Medical Rehabilitation
  University of Manitoba

Although this seems a plausible explanation for K0, I remain to be convinced until I see some data. Seems to me that it
should be quite easy to do and I will have a go.

A more important question arises, though. Do we know how the commercial software out there (Vicon
Plug-in-Gait/BodyBuilder, Orthotrak, C-Motion Visual3D) cope with this problem? I have checked BodyBuilder, for
example, and I also see an K0. The way BodyBuilder works I don't think there is any filtering unless I explicitly ask for
it - can anyone confirm this?

Dr. Chris Kirtley MD PhD
Associate Professor
Dept. of Biomedical Engineering
Catholic University of America 

                      Hi Chris,

                      I think that At is saying that it is caused by our inability to record
                      instantaneous changes in velocity caused by impact (not so instantaneous
                      because of soft tissue but still pretty quick). The concentric power
                      last for quite a long time so I am not convinced that it is only
                      artifact (but I could be wrong there). Modeling the foot floor
                      constraint, including impact is the bane of forward dynamics models of
                      walking. At's conclusion may be more compelling if heel strike in
                      simulated walking does not show this concentric power. Ton van den
                      Bogert may want to chime in with his experience modeling impact.

                     Filtering force platform data at 5 hz is an interesting idea provided
                      you maintain the sharp edge at impact and only filter the rest of the


                      W. Scott Selbie
                      C-Motion, Inc.
                      15821A Crabbs Branch Way
                      Rockville, MD 20855
                      (301) 840-1919
                      (613) 532-1766 (cell)
                      (301) 840-0271 (fax)

                      I agree 100% with At Hof.  In inverse dynamics you really must filter
                      kinematics and GRF with the same cutoff frequency.  Otherwise you
                      can introduce artifacts.  I once did a conference presentation on this,
                      but have never written this up properly.  In gait these artifacts may
                      be small, but in running they can be quite big.

                      See http://www.isbweb.org/data/invdyn

                      I understand Scott's concern about low-pass filtering of GRF.
                      If you use a zero-lag filter, the filtered GRF will rise even before
                      the foot hits the ground!  But your low-pass filtered acceleration
                      of the limb segments, which are related to the GRF, will also
                      rise before the foot hits the ground.  So both signals really
                      have to be treated the same way.

                      In simulated movements you never see impact peaks in joint moments.
                      This is because muscles are soft tissue, they do not have enough
                      stiffness to generate or transmit impact forces.  So in simulated
                      joint moments, there are no impact peaks.  Of course a model is
                      just a model, but I am pretty confident that real muscle forces
                      do not have impact peaks either.  Data from instrumented animals
                      (horse, cat) does not show any.

                      But of course, a model is just a model and real joints are not
                      perfect ball-and-socket mechanisms where muscles are the only
                      contributors to joint moments.   The passive varus-valgus (add-abd)
                      stiffness of the knee joint may be high enough to generate impact-like
                      peaks in that particular moment.  The ankle joint does not really
                      have 3 unrestrained degrees of freedom, so there is some
                      non-muscular stiffness there also.

                      In inverse dynamic studies on impact movements, you often see that
                      people low-ass filter the joint moments at the end.  I think this
                      is because they (rightly) do not trust the large impact peaks that
                      they see.  But it would be much better to make sure that at the input
                      the signals already have been filtered consistently.

                     A.J. (Ton) van den Bogert, PhD
                      Department of Biomedical Engineering
                      Cleveland Clinic Foundation
                      9500 Euclid Avenue (ND-20)
                      Cleveland, OH 44195, USA
                      Phone/Fax: (216) 444-5566/9198

Dear All,

Spurred on by At Hof's concluson that the Ko power birst is indeed an artifact, I thought I'd check back into the archives to see when it came about.

I have a report from the post-WW2 Berkeley study by Bresler and Berry (Energy and Power in the leg during normal level walking, 1951) kindly given to me by Larry Lamoraux, in which they show curves lacking the power artifact:

Having said that, there does seem to be a flexor knee moment immediately after contact, and so (since the knee angular velocity is always flexor) I'm not sure why there's no power.

Since digital filtering didn't come in until the 1960s, I presume no form of smoothing was performed on the kinematic data. Can anyone confirm this?


Dr. Chris Kirtley MD PhD
Associate Professor
Dept. of Biomedical Engineering
Catholic University of America

Back to Teach-in