Teach-in #21: Push-off & The Effect of Speed on Gait , What people said...

Perhaps the question should be 'why isn't the power generation shared between hip, knee and ankle at normal speeds as it is at high speeds?'.  It seems sensible to distribute the load in this way.  I suspect the real answer is that 0.2 J/kg is the maximum energy generation rate that can be sustained comfortably, and presuming A2 is the most efficient energy input, this will be favoured up to maximal generation, and then less efficient recruitment made at K and H.

Rob Strachan B.Eng MIPEM AMIMechE C.Eng SRCS
Clinical Engineer
Medical Engineering Section
Directorate of Medical Physics and Clinical Technology
Royal Hallamshire Hospital, SHEFFIELD  S10 2JF     Tel 0114 271 3138  Fax 3137
e-mail  Rob.Strachan@csuh.nhs.uk

Clinical Engineer
The Gait Laboratory
Ryegate Children's Centre, Tapton Crescent Road
SHEFFIELD  S10 5DD    Tel 0114 271 7629

Hello Readers,

Dr. Chris Kirtley <kirtleymd@yahoo.com> writes on Tue, 3 Apr 2001
>Finally, I am interested to know how many labs
>out there routinely adjust their normative data for speed.

The issue of normalising to velocity is not new. Krebs wrote a chapter
on the Interpretation Standards in Locomotor Studies (in Craik and
Oatis: Gait Analysis Theory and Application, 1995, ISBN0801669642) and
addresses the problem clearly. Stansfield et al. in 99 showed that "gait
is predominantly characterised by speed of progression, not age, in 5 to
12 year old normal children".

When collecting normative data for gait analysis I get kinematics,
kinetics and EMG at self selected walking speed, "slow" walk and "fast"
walk. The definitions of slow and fast may appear blurred but when
working with children it seems to be more feasible than controlling
walking speed in a more rigorous way.


Dr Gabor Barton MD                   CGA@gaitlab.demon.co.uk
Clinical Scientist
Gait Analysis Laboratory             tel: +44 (0)151 252 5949
Alder Hey Children's Hospital        fax1: +44 (0)870 052 1935
Eaton Road, Liverpool, L12 2AP, UK   fax2: +44 (0)151 252 5846

Hello All,

I believe that confusion remains because we are not rigorous in defining
propulsion and support.  We sort of treat support as maintaining the height of
the body CG and propulsion as maintaining the forward velocity of the CG.  In
that case very little propulsion and support work is being done.  The CG height
is nearly constant so there is very little displacement.  The CG AP forward
velocity is close enough to being constant that we have to be careful how we
process our data if we really want to measure the velocity variations.  Hence,
very little force is required to maintain forward velocity.  Thus the energy
production measured under the A2 wave is not being used to propel or support the
CG, it isn't needed.

The work is being done in the lower limbs.  The stance leg travels slower than
the body, the swing leg travels faster than the body.  The kinetic and potential
energy of the limbs is changing constantly.  This is where the work is being
done.  In a paper that should be required reading for all in gait analysis,
Meinders (Scan J Rehab Med, 30, 39-46) showed that energy flows into the leg in
push-off and very little flows from the leg to the upper body.  Some of that
energy is recovered, transferred to the upper body in late swing, but only
through the action of the hip extensors.  We looked at how this pattern of
energy flow varies with speed (Riley, J Biomech.  34, 197-202).  Our analysis
combined induced acceleration analysis with the energy flow analysis.  Since
accelerations are vectors you can distinguish propulsion from support.  In
summary, we found that, after the CoP moves to the forefoot, ankle plantar
flexion is responsible for support, i.e., maintaining the CG vertical height.
An independent kinematic sensitivity analysis verified the ankle's role as a
"determinant of gait" (Arch PM&R, 81 1077-1080).  The hip determines propulsion.
Apparently the knee keeps itself stable and cooperates with the ankle and hip to
allow toe-clearance.  (The knee is probably doing a lot during the loading
phase, but this period is the weak point of our analysis and I cannot say much
about that.)

What changes with speed?  Those CG oscillations get bigger.  The vertical
oscillations increase, but only a fraction of 1g.  Thus, the support function
varies only fractionally and ankle power looks more or less constant.  The
propulsion function, dealing with AP velocity changes, varies much more with
speed. It is not sitting on top of a large constant load, it is only dealing
with what varies.  Hence, variations in hip power varies with speed are
relatively more significant.

We (Steve Kautz, Tom Kepple, myself and some others) will be getting together
for an informal meeting at the GCMA Conference to talk about these analyses and
issues.  The focus will, I think, be on modeling methodology and how to report
our findings in a meaningful way.  Could be interesting if you're into the
modeling side of gait analysis.

Question: The "Determinants of Gait" concept says that CG vertical oscillations
are minimized to conserve energy.  Minimizing CG oscillations requires large and
powerful calf muscles.  This greatly increases the mass and inertia of the leg,
hence the energy required to move the legs.  So, does minimizing CG vertical
oscillations increase or decrease energy consumption?

Enough for today.

Patrick O. Riley, PhD
Spaulding Rehabilitation Hospital

I also learned a lot from reading Pat Riley's post.  It makes sense that if
ankle power is responsible for support rather than propulsion, then it does
not need to increase with speed.  I wonder, however, if the decrease in
ankle power mentioned by Chris in his original post might reflect
limitations on power due to muscle force-generating properties. In fast
walking, the ankle obviously plantarflexes more rapidly than in free or
slow walking, yet power (moment times velocity) is decreased.  This implies
that plantarflexion moment is decreased by a factor greater than that which
describes the decrease in power.  Because the moment arm of the Achilles
tendon is not likely to vary much with speed, there should be an
accompanying reduction in plantarflexor muscle force.

A study by Murray et al. (JOR 2:272-280, 1984) has shown that EMG activity
of the plantarflexors generally increases with walking speed, but that
plantarflexion velocity is higher (resulting in higher shortening velocity)
and the ankle is more plantarflexed (presumably resulting in shorter muscle
fibers) in fast walking.  These last two factors would be expected to
decrease plantarflexor force production via the force-velocity and
force-length relationships.  Whether force is reduced because the
plantarflexors are unable to meet the demand is debatable, but the
increased plantarflexor activity seen in fast walking would seem to support
this idea.  This relationship between speed and force-generating properties
(as well as the lever arm of the GRF at the ankle) has been explored using
a simple computer model created by Ahmet Erdemir, a graduate student
working with me.  We hope to publish some of our results soon in Gait &

Best regards,
Steve Piazza
Stephen Piazza, PhD
Assistant Professor
Departments of Kinesiology
   and Mechanical and Nuclear Engineering
Center for Locomotion Studies, 29 Recreation Building
The Pennsylvania State University
University Park, PA  16802

In partial reference to Ben and Gabor's comments on the control of walking

Isn't the range of walking speeds an indication of the function/flexibility
of the combined  motor control system and its load? It seems like a good
idea to test for the range of velocities (bandwidth) available to an
individual before and after intervention (or perhaps even the ability to
change velocity). You might be on to something with those LEDs, Ben! You
might be on to something asking your children to walk at their fastest and
slowest speeds, Gabor!!

Is there any role for metronomic walking in this? A plot of velocity against
walking frequency might be interesting.
I've tried metronomic walking on a few of my students with emg electrodes
over the tib ant and med gastroc. Interestingly, and quite contrary to my
expectations, across the range of walking cadences (60-160 steps/min)  there
was no change in levels of co-activation between the antagonist muscles.
This puzzles me because I anticipated that it was the level of coactivation
in high frequency walking that would make one change one's mode of transport
to running! Does anybody know why we have to change from walking to running
at a certain critical speed? If the answer is to do with inverted pendulums,
don't reply.

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

Hello CGAers All,

Many thanks to Adam and all the previous contributors to this very
thoughtful discussion.

One of the most remarkable features of normal walking is how easy it is to
change walking speed.  Ralston (Ralston, H. J. Energy-speed relation and
optimal speed during level walking. Int. Z. angew. Physiol.17:277, 1958)
showed that, for each individual, there is a walking speed that requires
minimal expenditure of metabolic energy per unit distance walked.  Both
slower and faster speeds require more energy per unit distance.  Much was
made of the existence of this "comfy cadence," but little attention was
given to the more remarkable fact that an allowance of a mere 15 percent
increase in energy expenditure above this minimal value would allow almost a
three-to-one range of walking speeds for a normal subject.  That is, the
fastest walk achievable at this maximum allowable energy was almost three
times as fast as the slowest walk achievable at that same energy level.  For
disabled subjects--Ralston studied amputees--this range was greatly reduced.
Ralston elaborated on these ideas in Chapter three of the first edition of
the book "Human Walking" by Inman, Ralston and Todd, Williams & Wilkins,
Baltimore/London, 1981.  This book is out of print, but is available in many

In the 1960's, Frank Todd observed that the accumulating body of scientific
data on human walking did not seem to be leading to a better understanding
of the underlying principles that govern the walking process.  He concluded
that the reason for this lack of progress was that the effects of changes in
walking speed were not being taken into account in any quantitative way.  In
those days, and for decades thereafter, it was common for descriptions of
experimental locomotion research procedures to include the term "normal
walking speed."  Todd, on the other hand, observed that "normal" in walking
is not any particular speed, but the ability to change speed easily, so he
set about trying to find a practical, quantitative way to take this
important gait variable into account. His early studies centered around a
graph of stride length as a function of cadence.  His thoughts on the
subject up to 1980 were summarized in Chapter 2 of the above-referenced book
"Human Walking."

Todd's ideas eventually led to a graph of stride length as a function of
walking speed that has proven to be clinically useful to help assess the
gaits of children with a variety of walking disabilities at Shriners
Hospitals for Children in San Francisco and Sacramento, California.  This
graph, basically, is a specialized graph of temporal-distance parameters of
gait.  The origins and application of this graph (somewhat ambiguously known
locally as the "Shriners Gait Graph") were presented in the paper:

Todd, F.N., et. al.: Variations in the gait of normal children. J Bone &
Joint Surgery, 71-A: 196-204, Feb. 1989.

The current procedure at the Northern California Shriners Hospital (in
Sacramento) is to measure subjects at their freely chosen walking speed and
at their fastest comfortable speed.  Measuring the slowest speed has not
been found to be useful.  Values of stride length for each walking trial are
plotted against the walking speed for that trial, where they can be compared
on the graph with values expected for a person of the subject's height.  A
judgement can be made whether the subject's temporal-distance parameters are
as would be expected for a child of the subject's height, and whether the
subject changes speed in a normal manner.

With regard to the use of metronomes, a word of caution.  Most subjects can
follow a metronome without modifying their accustomed temporal-distance
parameters for the MEDIUM cadences, but at increasing cadences, the subject
will almost invariably begin to take abnormally (for that subject) short
steps, in order to keep up with the cadence.  This leads to a distortion of
the subject's accustomed gait at the higher cadences.  Todd identified this
error in some of the earliest gait research reported in the literature, when
he plotted temporal-distance parameters from a metronome experiment
conducted by the renowned French researcher Marey in the late 1800's.

The fundamentally important temporal-distance parameters of gait are easier
to visualize, and therefore become more useful, when they are plotted on a
graph.  Whether the graph is the "Shriners Gait Graph," the graph suggested
by Adam, or any of several other possibilities, the benefits of
systematically graphing the data should not be underestimated.

Best regards to you all.

Larry Lamoreux
Lafayette, California

Dear all,

This set me thinking that I have no intuitive grasp of the energy
consumption of walking, and since this relates to speed, I wonder if we
might move on to discuss that?

As far as I know, there are three methods for estimating energy consumption:

Would someone out there like to make a start by giving us some typical
values for normal gait for any or all of these three methods?
Presumably, the units would be J/kg or J/(kg.m) or J/(kg.m/s).

Dr. Chris Kirtley MD PhD
Associate Professor
HomeCare Technologies for the 21st Century (Whitaker Foundation)
NIDRR Rehabilitation Engineering Research Center on TeleRehabilitation
Dept. of Biomedical Engineering, Pangborn 105B
Catholic University of America

I know you want "real data", but couldn't resist sending the


60mins for 195lb person walking at 4.5 mph:
552 calories or the equivalent of:
     Pastrami beef, on rye
     Homemade Chicken potpie
     McDonald's Chicken
     McNuggets (12 pieces)
     5-oz. (2 links) Pork Smoked Sausage

Jason Harrison

Dear all,

I'm grateful to Jason Harrison for the figures below! Let me see if I
can convert:

195lb person walking at 4.5 mph for 60mins = 552 calories (I think this
should be Cal)
88.6 kg                 2 m/s       3600 s = 132 kJ (or 12 McNuggets!)

which works out as 132000/(88.6 * 3600) = 0.4 J/kg assuming a cadence of
120 steps/min
or 132/(88.6 * 2 * 3600) = 0.2 J/(kg.m/s)

Is that right?


Okay, so a real question -- slightly off topic -- how are the caloric
value of foods computed?  And do these values, and the "exercise"
consumption values take into account efficiency of the digestive

Jason Harrison

Chris, I have not completely followed all British to metric
conversions, glad that my homecountry once was invaded by
In my somewhat worn-out copy of Inman (1981) I find
Ew = 32 + 0.0050 v2
This is in calories and meters per minute (!), but it can be
recalculated into:

        Ew = 2.24 + 1.26 v2

with Ew in watts per kg  and v in meters per second.
In walking energy consumption thus increases with v2. It is thus
not proportional to v, thus not a constant energy per meter.
This relationship leads to the finding that there is an optimal
walking speed, with least energy per meter. But you better consult
Inman, chapter 3.
In running things are different: The rule of thumb is: 1 calorie per kg
per meter = 4.2 J/kg.m
 (from Margaria, 1976, Biomechanics and energetics of muscular

At Hof
Department of Medical Physiology &
Laboratory of Human Movement Analysis AZG
University of Groningen
A. Deusinglaan 1,  room 769

PO Box 196
Tel:   (31) 50 3632645
Fax:   (31) 50 3632751

Hello All,

Pat Riley ends his very thoughtful comments of April 3, 2001 with a
(rhetorical) question:

Question: The "Determinants of Gait" concept says that CG vertical
oscillations are minimized to conserve energy.  Minimizing CG oscillations requires large
and powerful calf muscles.  This greatly increases the mass and inertia of the
leg, hence the energy required to move the legs.  So, does minimizing CG vertical
oscillations increase or decrease energy consumption?

It is clear from his comments that the answer is no.

Another way to argue the case is that in alternating bipedal walking, the
center of mass MUST slow down and speed up again during each single-support
phase.  This is a necessity, not an option, because the single-limb support
phase starts out with the supporting foot in front of the center of mass,
causing an inclined ground reaction force vector with a posteriorly directed
component that decelerates the body center of mass, and ends up with the
supporting foot behind, causing an acceleration of the body center of mass.
That's a long sentence, but the concept is simple, I think.  We can't walk
on two legs with a useful step length without speeding up and slowing down
with every step.  And if we speed up and slow down, our progressional
kinetic energy must increase and decrease accordingly.

If the kinetic energy of the walking body must vary with each step, then it
is desirable for the potential energy to change in a complementary manner,
to minimize changes in the total energy of the body.  This doesn't require a
lot of control by the central nervous system.  It happens quite naturally as
a result of the mechanics of the system, I think.  A pendulum doesn't need a
central nervous system to swing.

Moving the body center of mass up and down is an option, not a necessity.
There are plenty of degrees of freedom in the lower limbs to allow us to
walk without moving the body center of mass up and down with each cycle.
But we don't, because the right amount of moving up and down, at the right
time, minimizes the total energy changes of the body center of mass.

Inman's Determinants of Gait concepts put the cart before the horse, and
it's taking us a very long time to get the old vehicle turned around again.

Best regards to you all.

Larry Lamoreux
Lafayette, California


Back to Teach-in