Observational Gait Analysis: What people said... (CGA Sept. '97)

I am currently seeking information about observational gait analysis education as part of my PhD studies.

In particular I am interested in

While there is an abundance of literature pertaining to gait kinematics and specific pathologies, I am finding very little that relates to observational methods to help students systematically assess gait and arrive at a reliable and valid decision.

Any assistance would be much appreciated.

Your with thanks

Nerrolyn Ford

Nerrolyn Ford School of Human Biosciences La Trobe University Bundoora 3083 Victoria, Australia.

Phone: (03) 9479 5729 Fax: (03) 9479 5784

There is an Observational Gait Analysis link


on the CGA page to Dr. Christopher E. Bork Ph.D., Dean of Allied Health at the Medical College of Ohio, which you might check out. I put it on there a long time ago, and I notice that it's never been updated since, so I don't know whether the work is going on.

I think your question is a very good one - despite all the interest in gait analysis these days observational analysis is, as you say, very rarely mentioned. The Ranchos group used to put out a video, but I think that was mainly for prosthestics.

As far as I'm concerned, I tend to go the opposite way to the way you suggested, i.e. foot upwards. The reason for this is David Winter's work showing how important the ankle A2 push-off is. Many gait abnormalities at the knee and hip are compensations for a weak push-off. For this reason, the step length on each side is probably the first thing I go for, followed by the ankle angle at contact (forefoot or flatfoot contact?) and toe-off (apropulsive?).

Knee angle at contact is next, and amount of swing and stance phase flexion (taking into account the cadence). Finally, hip extension in terminal stance and trunk angle (forward flexed?).

In the frontal/transverse planes, stride width, foot angle (varus/valgus), pelvic obliquity & use of the arms (Trendelenberg).

Obviously, these are just the main pointers, and obvious abnormalities would be noted first just as in reporting a radiograph.

Hope this helps - I'll be very intereested to see what others say.

Chris -- Dr. Chris Kirtley (Kwok Kei Chi) MD PhD Assistant Professor Department of Rehabilitation Sciences The Hong Kong Polytechnic University

I would tend to agree with Chris although in the transvese plane I would also include the alignment of the patella (although we know from 3D kinematics that pelvic/hip rotation is essential for sensible interpretation of the transverse plane). I have recently been involved with a working party for ISPO trying to establish a standardised protocol for observational gait analysis for use in orthotics research (the task is not completed yet so please don't anyone ask for details). One thing which is apparent is that producing a tool with low inter-tester repeatibility is very difficult. If you want something which is clinically useful then I suggest you follow the route adopted in the CAMARC project, as in the protocol for Cerebral Palsy, where observation is based on identifying gait phenomenon related to known physical limitations associated with the particular clinical condition. This method does not allow for the expansion of knowledge, but this would require a repeatable method of observation anyway.

Jeremy Linskell Manager, Gait Analysis Laboratory Co-Ordinator, Electronic Assistive Tehcnology Service Dundee Limb Fitting Centre Dundee, DD5 1AG, Scotland tel +1382-730104, fax +1382-480194 email: j.r.linskell@dth.scot.nhs.uk

I've attend workshop on "observational gait analysis" by tutor from Rancho Los Amingo Hosptial so I might be able to explain a bit on what they emphazied.

Video tape of subject is taken and slow played preferably frame by frame. Gait cycle is divided into 8 phases. The hip, knee and ankle joint are observed for abnormal kinematic and tick against the check box. They don't specific which joint to start with but I agree with Chris that starting from the ankle is a good idea. The abnormality observed could be due to several factors: pain, muscle weakness, spasticity, contracutre. If you suspect the ankle had decrease ROM due to contracture, check out if the problem exist in other part of the gait cycle. Together with physical examination, you could have a better "guess" on the cause of the abnormalities observed.......

I've played the video tape of amputees and some orthopaedic case to students in class. Amazingly most of them can pick up the major abnormalities of the subjects! Of course, let along we could not derive the cause of abnormalities based on video alone. I belived observational gait anlaysis has definite role in clinical situation. I think Chris will agree with me that video has a lot to tell us! If you have quantitative gait data, that is perfrect (provided they provide accuate 2-D or 3-D data or else they are useless). Well, the major drawback on observational gait analysis, I believed is subjectivness. Thought there is paper saying good repeatability of it, I doubt it is possible. Anyone have comments about the repeatability?

Dora Assistant Professor Physiotherapy Section Department of Rehabilitation Sciences The Hong Kong Polytechnic University Hung Hom, Hong Kong Tel: (852) 2766-6746 Fax: (852) 2330-8656 E-mail: rsdpoon@polyu.edu.hk

As a clinician, one of the "big picture" ways I look at gait and decide then how to intervene is whether the phase appears to be active or passive, and then at the range of motion used versus the available range of motion. Although gait overall is an active process, patient strategies often reflect weakness, or an attempt to put one foot in front of the other without the strength necessary to have a "normal" gait. Gait is all about efficency with all of the eccentric contractions, and momentum generated events such as the first 40 degrees of preswing that is mostly passive on the swing limb. When an injury occurs (traumatic e.g. stroke or fracture, or atraumatic e.g. loss of ROM), the strategy is to minimize the use of the involved part, similar to the "learned non use" theory (Taub). I think this occurs in both orthopedic and neurologically involved patients. Here are a few examples:

1. If in stance a patient's pelvis/hip drops out posterio-laterally and remains there while the swing limb advances, I see the stance phase as relatively passive as far as the lower extremity is involved, particularly the gluteus medius. I see the patient hanging on their hip joint capsule and ligaments, or the passive elastic components of the muscle, or whatever they can possibly use to avoid firing the gluteus medius in a way that would require it to support their body weight.

2. Similarly, if the knee accelerates into extension as with the "extensor pattern" seen in patients post CVA, I see that as avoiding the demands of loading response on the hip extensors and knee extensors, and hanging on the posterior capsule of the knee joint. The patient may also be accelerating their tibia posteriorly to compensate for a rapid acceleration of their trunk anteriorly.

3. Decreased weight shift to the affected side when using a unilateral assistive device--if your left leg is affected (CVA or otherwise), putting all of your weight on your right leg, right arm, and cane, does not make your left side stronger. I often shift the side that the cane is in and see a dramatic improvement in the quality of the gait and activity in the involved extremity. The difference can be very dramatic in ambulatory stroke patients.

4. Decreased step length can suggest gastroc/soleus weakness. The demand on the gastroc/soleus in terminal stance is significant, and the patient will minimize single limb support by taking a shorter step.

To test the hypotheses I make from observation, I test muscle strength relative to body weight, and range of motion. The Upright Control Test from Rancho is one way to test strength that works well with neuro patients. Closed chain activites (hip abduction, ankle plantarflexion) work well too.

The advantages I find with looking at gait in this way are 1) you start to see the forest through the trees--I keep from getting caught up at one joint that may or may not be the primary problem and look at the overall quality of the movement 2) patients that want to improve their gait need their strength to be able to deal with their body weight, not with a theraband or a cuff weight. This way I can give the patient a couple of exercises that give them the most "bang for their buck." 3) If you can get the patient to see the benefits of moving more efficiently (closer to normal gait values) they will be making themselves stronger by walking in that manner, and (I think) less likely to have problems in the future.

Somewhat rambling, but my thoughts nonetheless.

Melanie Weller MPT San Diego, CA

Thanks to everyone who contributed to my queries regarding observational gait analysis methodology.

The most popular response concerning visual search methods during gait analysis was to begin at the feet and work upward, although it was recognised that obvious anomolies would be detected initially.

The Rancho Los Amigos observational gait analysis workbook was by far the most popular teaching text. This text was reprinted in 1996 and is available from the Los Amigos Research & Education Institute.

I also had an enquiry about the repeatability/reliability of observational gait analysis. Below is a quick review of reliability studies to date, although the sample populations and data analysis techniques differ, I suggest that clinically acceptable levels of reliability are yet to be demonstrated.

Goodkin and Diller (1973) - 17 physiotherapists observing hemiplegic gait in a live setting. When required to document the two major deviations, average agreement was 30%. When asked to document the two major treatment approaches, agreement was 28%.

Saleh and Murdoch (1985) - Experienced observers rating gait of transtibial amputees in a live setting. Prosthetic limbs were intentionally misaligned in the sagittal plane. Agreement of observers with a biomechanical model was 22%.

Krebs et al (1985) - 3 physiotherapists rating video footage of a paediatric sample fitted with bilateral KAFO's. Pearson's correlation coefficient was 0.6 within observers and less between observers.

Eastlack et al (1991) - 54 physiotherapists rating video footage of rheumatoid arthritic patients. Cohen's kappa reached a maximum of 0.51 and a minimum of 0.11.

Keenan (1994) - 5 podiatric clinicians rating video footage of rearfoot motion. Kappa between raters was 0.19, kappa within raters varied from a negative value to k=0.59.

Ford et al (1995) - Prosthetic clinicians rating video footage of transtibial amputees with intentional misalignments in the sagittal and coronal plane. Between rater reliability over coronal plane misalignments was k=0.36. Between rater reliability over sagittal plane misalignments was k=0.08.

Nerrolyn Ford School of Human Biosciences La Trobe University Bundoora 3083 Victoria, Australia.

Phone: (03) 9479 5729 Fax: (03) 9479 5784

Here are some references...

CARR E.K.,1991,"Observational methods in rehabilitation research", Clin. Rehabil., 5, 89-94, E : COSTELLO A.J.,1973,"The reliability of direct observations", Bull. Br. Psychol. Soc., 26, 105-108,

E : FUGL-MEYER A.R., L. JAASKO, I. EYMAN, S. OLSSON & S. STEGLIND,1975,"The post stroke hemiplegic patient. A method for evaluation of physical performance", Scand. J. Rehabil. Med., 7, 13-31,

E : HUGHES K.A. & BELL F.,1994,"Visual assessment of hemiplegic gait following stroke: pilot study.", Arch. Phys. Med. Rehabil., 75(10): 1100-7, E : Hemiplegia-diagnosis; Observer-Variation; Pilot-Projects; Reproducibility-of-Results *Cerebrovascular-Disorders-physiopathology; *Disability-Evaluation; *Gait-physiology; *Hemiplegia-physiopathology

KREBS D.E., J.E. EDELSTEIN & S. FISHMAN,1985,"Reliability of observational kinematic gait analysis", Phys. Ther., 65, 1027-1033,

E : PERRY J., GARRETT M., GROLEY J.K. & MULROY S.J.,1995,"Classification of walking handicap in the stroke population.", Stroke., 26(6), 982-9, E : Activities-of-Daily-Living; Cerebrovascular-Disorders-rehabilitation; Gait-; Knee-Joint-physiopathology; Locomotion-; Middle-Age; Motor-Activity; Proprioception- Cerebrovascular-Disorders-physiopathology; *Disabled-classification; *Walking- Female; Human; Male; Support,-Non-U.S.-Gov't; Support,-U.S.-Gov't,-Non-P.H.S.

TURNBULL G.I. & J.C. WALL,1985,"The development of a system for the clinical assessment of gait following a stroke", Physiotherapy, 71(7), 294-298,

E : WADE D.T., V.A. WOOD, A. HELLER J. MAGGS, R.L. HEWER,1987,"Walking after stroke : measure and recovery over the first 3 months", Scand. J Rheab. Med., 19, 25-30,

E : WINTER D.A.,1985,"Concerning the scientific basis for the diagnosis of pathological gait and for rehabilitation protocols", Physiother. Canada, 37(4), 245-252, E :


With regard to the recent discussion on observational gait analysis, I think the following papers add an interesting dimension to this debate. It seems to me that the problem with "pathological" gait patterns is often the fact that the manifestation of a single problem is not always stereotypical, and also that several different disorders can produce quite a similar output, therefore the observed limitation is not, in itself, diagnostic. The point is, then, that we need to follow up the observed deviation with other assessment procedures to clarify the source of the gait disorder. I would like to see more development along the lines proposed in these two papers, if only to test the validity of the approach.

Jack Crosbie University of Sydney

Moseley et al (1993) Observation and analysis of hemiplegic gait: stance phase. Australian Journal of Physiotherapy 39 (4): 259-267.

Moore et al (1993) Observation and analysis of hemiplegic gait: swing phase. Australian Journal of Physiotherapy 39(4): 271-278.

In response to Chris' comments about the limited accessibility of the Australian Journal of Physiotherapy, I have taken the liberty of abstracting what I think is the core of the material presented in the papers I referred to. The idea is that the observer looks for the presence of key gait deviations, in this case with the stroke patient, then assesses the likelihood of the possible causes. More often than not, there will be several deviations noted, and where a problem, such as shortening of the ankle plantarflexors for example, occurs in all or several of them, this helps direct the clinician to pursue this through further examination. In some cases, mutually contradictory explanations can be ruled out, or at least rendered less likely. As I said in my previous posting, I think this approach would benefit from further discussion and validation. I would also like to see a similar approach, which I suspect might involve only a small amount of modification, taken to a wider range of gait disorders. Despite the inconvenience involved, I would encourage people who are interested in this to read the original papers. They should be obtainable through the usual inter-library loan services.

References: Moseley et al (1993) Observation and analysis of hemiplegic gait: stance phase. Australian Journal of Physiotherapy 39 (4): 259-267.

Moore et al (1993) Observation and analysis of hemiplegic gait: swing phase. Australian Journal of Physiotherapy 39(4): 271-278.

Apologies for the size of this message, I hope it provokes some interest.

Jack Crosbie University of Sydney

We have just begun a study of expert clinicians from a variety of vocations (Orthotics, physiotherapy, rehabilitation medicine) to investigate observational gait analysis. The study uses video assisted recall strategies to elicit knowledge from experts immediatly after a client consultation.

We hope to identify

Nerrolyn Ford School of Human Biosciences La Trobe University Bundoora 3083 Victoria, Australia.

Phone: (03) 9479 5729 Fax: (03) 9479 5784

I have seen only one paper where a neural net was trained to identify abnormal gait. Its training was based on clinician's diagnosis that the gaits it "saw" were "normal" or "abnormal".

Holzreiter SH, Köhle ME. Assessment of gait patterns using neural networks. J Biomech 1993;26(6):645-651.

I think the point is well made and central to the discussion. If it is not diagnostic then the main value of it would appear to be in monitoring. Even though there do appear to be publications which show some inter-observer reliability in recognising aspects of pathological gait, I do not believe that such approaches can give the resolution required to allow effective monitoring in a clinical setting. I can't remember how many times I have sat in on a CP clinic and watched a consultant look at his written description of a child's gait and then look quizzically at the child, trying to decide if the child has deteriorated as the parents claim.

There is nothing wrong with using observational gait analysis in the clinical setting. It is a subjective tool which has a place along side all the other subjective tools. The question is, 'is it science?'! (just stirring!) .

Jeremy Linskell Manager, Gait Analysis Laboratory Co-Ordinator, Electronic Assistive Tehcnology Service Dundee Limb Fitting Centre Dundee, DD5 1AG, Scotland tel +1382-730104, fax +1382-480194 email: j.r.linskell@dth.scot.nhs.uk

I get the feeling this discussion is starting to bring out a lot of skeletons from the cupboard...

I've wondered for a long time why clinicians (doctors & therapists alike) rarely measure stride length. It's simply done, using a wrist watch and a walkway of known length, and the formula SL=120*Velocity/Cadence yet hardly anyone seems to do it! I often use familiarity of the method as a test to see if someone is really interested in gait analysis before advising them which system to buy. I consider it's rather like a cardiologist ordering an electrocardiogram before taking a patient's pulse!

A colleague and I just discussed this phenomenon, along with many other instances of seeming disinterest in low-level measurement technology, and we concluded that it really demonstrates a fervent desire for using high-tech equipment in an effort to avoid thinking in the clinical situation, perhaps because of a reluctance to "waste" precious time that could otherwise be used for "doing", i.e. treating hands-on. Of course, this is a paradox, since computerised analysis requires far more time than observational analysis.

Someone once defined a paradox as a "truth standing on it's head"!

What do people think? (waits for deluge of abuse...)

Chris -- Dr. Chris Kirtley (Kwok Kei Chi) MD PhD Assistant Professor Department of Rehabilitation Sciences The Hong Kong Polytechnic University

In response to Chris' posting about use of measures such as stride length as clinical indicators, I might take this opportunity of adding my support. It seems to me that Richard Brand's comment (in "Biolocomotion; a century of moving pictures" Human Kinetics Publishers, 1991 p.227) remains valid and expresses, for me, the essential problem with contemporary gait analysis as a "discipline". I quote:- "...gait analysis largely remains technology searching for application.....locomotion investigators have often confused research utility or cleverness with clinical usefulness."

The drive towards accreditation of gait analysis facilities on the basis of the technology housed within them is, in my opinion, highly regrettable. I understand that there is a wish to protect the public from unscrupulous 'cheap and cheerful' operators, but I cannot accept that there is always a need to conduct a full-on bells and whistles analysis, particularly when the clinical problem itself has not been clarified.

Chris wrote:

>I've wondered for a long time why clinicians (doctors & therapists alike) rarely measure stride length. It's simply done, using a wrist watch and a walkway of known length, and the formula SL=120*Velocity/Cadence yet hardly anyone seems to do it!

I would only comment that stride may conceal step assymetry, but certainly there is a great deal of useful information to be gained through relatively simple temporal-spatial analysis. As a number of authors have indicated (including Kirtley et al, 1985) there is a predictable and logical relationship between walking speed and the kinematics and kinetics of the lower limb. Hardly surprising, really!

>A colleague and I just discussed this phenomenon, along with many other instances of seeming disinterest in low-level measurement technology,

I would not use the term low-level, but rather domestically accessible. Much useful information can be obtained from a video camera, 4-head freeze-frame VCR, monitor and PC if we are prepared to make a little effort.

Dr. Jack Crosbie Associate Professor School of Physiotherapy University of Sydney PO Box 170, LIDCOMBE NSW 2141 AUSTRALIA

I strongly agree with using low-tech methods to assess aspects of gait, as they are useful in setting goals with patients who want to improve their gait. Patients often express to me the desire to walk faster, walk and carry something simultaneously, able to cross at a crosswalk on a crowded street, etc. For this reason, I sometimes incorporate gait velocity into the patients goals.

Stride length particularly intrigues me (it is not a measure I have used), because I often try to improve it in gait, particularly with hemiplegic patients to improve swing (foot clearance) of the involved side.

When analyzing gait (observationally), I look at

Whereas a lot of the technical equipment may be useful for diagnosing, it has been my experience (and that of many with whom I work) that physical therapy intervention for the motivated patient is most effectively spent first acquiring the necessary range of motion to perform the desired task and second with activities that force the use of the involved extremity/ies to promote increased strength and control for a normal loading response, stride length, single limb support, etc.

I see gait abnormalities are strategies for patients to get from point A to point B maximizing what they do have (e.g. bony stability) and minimizing what they don't have or what doesn't work as well (e.g. dynamic (muscular) stability/mobility, ROM).

One last note on gaining range of motion...I recently saw a dramatic before and after video of a patient who was scheduled for a rhizotomy to decrease his plantar flexor tone on his involved side. After one treatment of soft tissue mobilization to the gastroc/soleus the patient essentially had a normal gait, and his rhizotomy was cancelled. Interesting statement on the relationship of tone and function....

Melanie Weller MPT

I wish to ask the following questions of those who use observational gait analysis in clinical assessment. I ask them not to devalue the merits of it as an adjuct to other subjective methods used in the clincial assessment process, but to understand the significance which is attached to it.

I look forward to a stimulating discussion.

Jeremy Linskell Manager, Gait Analysis Laboratory Co-Ordinator, Electronic Assistive Tehcnology Service Dundee Limb Fitting Centre Dundee, DD5 1AG, Scotland tel +1382-730104, fax +1382-480194 email: j.r.linskell@dth.scot.nhs.uk

I shall take a stab at this....

I do not believe that a given disease process causes a certain pathological gait. I believe that biomechanical constraints (e.g. range of motion limitations, strength deficits, movement "choices") contribute to the disorder. For example: initiating swing from the pelvis when the pelvis is rotated posteriorly to the involved side (horizontal plane), can increase extensor synergy in both stroke and "normal" patients.

Beth Fisher is a PT who teaches courses on intervention with neurologically involved patients. At the beginning of her courses, she takes two, young, healthy individuals (class participants), and tapes them each in a different way. The first she tapes the pelvis into elevation and backward rotation. This young, healthy person dramatically ambulates as if she had a stroke, with extensor thrust, foot supination, etc., and even demonstrates a flexor synergy in her upper extremity on the taped side, even though there is no tape except at the pelvis. The second individual is taped anteriorly on the chest, into scapular protraction and thoracic flexion. This patient dramatically ambulates as if she had Parkinson's, with a shuffle, limited arm swing, etc.

In the course of the taping neither of these individuals developed a stroke or Parkinson's, suggesting that the biomechanical contribution to these movement patterns is significant.

Taking the person with the pelvic elevation and backwards rotation: when this component was altered (tape removed), the patient walked normally again. Tone in the plantarflexors and supinators was not the primary or "control" component of her gait deviation, but the consequence of the positioning of her pelvis.

If optimum response means going from abnormal to normal gait in one treatment, I don't think most patients show the optimum response in one treatment. However, if I look at the phases of gait that the patient is avoiding (e.g. loading response), and begin to force the patient to load the limb in theraputic activities, and then in functional activities (e.g. gait), the patient begins to re-learn how to manage their body weight on the involved lower extremity. Scientifically, this is based on the forced-use studies by Taub and Wolf.

Taub et al Technique to improve chronic motor deficit after stroke. Arch Phys Med Rehabil 1993: 74:347-54; Wolf et al Forced use of hemiplegic upper extremities to reverse the effect of learned non use among chronic stroke and head-injured patients. Experimental Neurology 104, 1250132 (1989)

Patients are enormously creative and constantly surprise me with the ways in which they can accomplish a given task in a way that is most non-therapeutic (or compensatory v. recovery (LeVere TE, Recovery of function after brain damage: a theory of the behavioral deficit. Physiol Psych 8:297, 1980) ). I frequently determine if I am getting the response that I want through palpation of the muscles that are in the most demand at a given phase of gait (e.g. hip extensors in loading respolnse). I also must frequently change the contsraints of the task to get the response for which I am looking, by altering base of support, alignment, stability/mobility, and sequencing of the task.

Beth Fisher is finishing up her PhD and making this her life's work. For me, I constantly experiment with different activities to see what seems to "make it click" with the patient. Some NDT treatment courses teach forced use theory and intervention in order that this information may be spread.

Melanie Weller MPT

I think we have stumbled on a very fruitful topic for discussion here - perhaps something of an Emperor's clothes!

It seems to me that there are two separate phenomena going on:

I know that sounds contradictory, so let me give some examples:

1. I find these days that people seem to jump from one physical variable to another in an attempt to explain motion - angle, force, momentum, moment, power, body centre-of-gravity, gravity itself all seem to be recruited in succession. I often end up with my head spinning when listening to such explanations! I like to remind my students that motion can only be caused by muscles contracting - nothing else!

2. In our Teach-in on Joint Moment and EMG at /teach-in/emg we saw how just looking at raw EMG can be extremely misleading. At Hof nicely showed how EMG should always be normalised according to the length-tension (and probably also force-velocity) relationships. Yet, all the studies published seem to report raw filtered/rectified EMG.

These are just two examples of two parallel phenomena that seem to be happening, and I'm curious to know why.


-- Dr. Chris Kirtley (Kwok Kei Chi) MD PhD Assistant Professor Department of Rehabilitation Sciences The Hong Kong Polytechnic University

Beware of ignoring the dynamics of a system! Often, motion is the result of momentum transfer (e.g., knee extension in late swing during gait, the result of deceleration of the thigh segment). I would agree that a body at rest tends to stay at rest, but remember the other half of Newton's First Law.

Marcus P. Besser, PhD Assistant Director, Human Performance Laboratory Assistant Professor, Department of Physical Therapy Thomas Jefferson University Philadelphia, Pennsylvania, USA 19107-5233 Phone:215-503-1645 Fax:215-503-3499 email: marcus.besserm@mail.tju.edu


Just a cautionary note. I agree with most of what you said but I think your statement "I like to remind my students that motion can only be caused by muscles contracting - nothing else!" is an oversimplification and might be misleading. The movements we see or measure as clinicians and/or researchers is the result of a complex interplay of organismic, task, and environmental constraints. As humans usually operate in a gravito-inertial environment, holonomic constraints play a vital role in reducing the degrees of freedom problem facing the motor system. Muscle activation is used only when the system cannot find another less energy demanding way (e.g., gravity, momentum, elastic recoil) of producing or stopping a movement. Bernstein's notions emphasise how, with practice, the motor system learns to exploit the passive reactive forces and thus use less muscle activation. Recording EMGs thus only go part way in explaining the movement we see.

Ben Sidaway <bsidaway@husson.husson.edu>

I have been reading the comments of observational gait analysis with great interest.

1. I wish to illustrate the power of observation in the “normal”.

Imagine waiting for a close friend or family in a busy crowded place. At

a distance of 100 feet you can pick put out your friend or family member

long before you can see their face. How is this? Without any gait analysis training, we can observe at a great distance and unconsciously choose one or several characteristics that identify that this is as being your friend (even though we don’t consciously state these characteristics).

Obviously, even in normals without major gait deviations there are characteristics peculiar to each of us. Hence I would think that observation is a very valuable tool but the more subtle aspects difficult to measure.

Just one point without being an expert, I like to view people’s gait from the perspective of how well they manage to walk with their disability.

2. Kinematic features of locomotion can occur without precisely timed efferent signals, see the example from the cat.

Knee flexion in the cat’s step cycle is not accompanied by knee flexor activity, and the transition from knee flexion to extension can be entirely due to the whipping action of the pelvis (Hasan and Stuart 1988). The CNS exploits the physics of the system to produce an efficient and simplified pattern of locomotion. It is only when scientists try to produce computer models for human locomotion that they appreciate the effect of inertial interactions and segment coupling. In modelling ankle plantar flexion in gait it is necessary to also consider the resultant inertial forces through the knee, hip, pelvis and spine to produce an accurate outcome of ankle action in gait (Winter 1990).

Thanks Robert

Robert J Burgess <rburgess@gis.net>

I am glad that Jeremy Linskell has raised these points about observational and other forms of gait analysis. I think the difficulty about this philosophical discussion is that there is no obvious and generalisable answer, whether one espouses the observational or the quantifiable position. We accumulate knowledge and expertise through repetition and become expert through our ability to identify patterns, whether these present as what we see as we play a videotape of a patient walking or scrutinise a graph of hip joint powers and moments.

There are dangers in taking the line mooted by Melanie Weller in her example of the healthy individual whose pelvis is taped into an extended and rotated position. Melanie suggests that "tone in the plantarflexors … was not the primary component of the gait deviation but the consequences of the positioning of the pelvis". I could position a thumb tack in a healthy individual’s shoe and cause them to simulate a gait more typical of a patient with an acutely painful knee. Removal of the tack would restore their normal gait. That does not mean that in a patient with such a presentation I should conclude that the problem originated in their foot. >From my understanding of the literature, there is ample evidence of shortening of the plantarflexors associated with a number of neurological problems, and that shortening may not be associated with increased tone, but simply reflect adaptation to lack of stretch through normal use. I have not encountered any studies that show such convincing evidence of change in "tone" in the trunk in the absence of quantifiable change to the flexibility of the ankle.

As I said in an earlier posting, we have a limited repertoire of responses to disruption of our lower limbs (including pelvis and spine) which manifest in gait deviations. That being the case, we need to be quite circumspect in interpreting the outward manifestations of these deviations. If I did place a thumb tack in a shoe, I would see changes to the normal kinematic and kinetic patterns in all of the joints of the lower limb, with consequent alteration of the powers, moments, EMG. These would be consistent with a reluctance to take weight on the affected side, with short single support, reduced or absent hip extension in late stance, and so on. I could have made that observation without attaching a single marker to my patient.

Until I explored the problem further with the patient I would be unable to state with conviction what the problem was. If I had all the possible analytical data from the most sophisticated equipment, I would have had no more idea. I might end by assuring anyone who is interested that I am by no means a Luddite in this regard, and freely confess to having access to, and using, 3d motion analysis systems, force plates, EMG, etc. I may have become a little more circumspect in my old age regarding when I use them, but they do have a place in my arsenal of investigative techniques.

Jack Crosbie

I thought my little mention of the ill-posed problem would generate some feedback! I confess I was curious to see the reaction. I encountered the idea in Vienna in discussions with Prof. Hatze (Giannis will know him well from the recent BIOMCH-L spat on filtering!). He is of the opinion that much biomechanical analysis is flawed because of the large amount of imprecision in measurement of kinematics, and the subsequent amplification of this by inverse dyanamics will result in large errors in the derived kinetics. Note that observational gait analysis would be even more prone to such errors.

Now, I'm not sure this conclusion is correct - I don't think the situation is as bad as all that, especially in gait when, as you say Giannis, accelerations are minimal. But I thought I'd drop it in to see the response.

Having said that, the ground reaction vector, although certainly being the most important component in the joint moments, is only half the story. The other half is the location of the joint centre (instantaneous axis of rotation), which is particularly susceptible to measurement and modelling error.

Now, as regards to Melanie's question, foot position is important for two reasons:

1. In inverse dyanmics modelling we normally start at the foot and work up the limb. So any errors in the foot segment will tend to be magnified and distort the joint kinetics derived for the more proximal joints.

2. Co-incidentally, the foot/ankle complex seems to be especially important in generating forward power during gait. I say co-incidentally because this isn't an obvious fact - it was actually disputed (e.g. by Jackie Perry) until quite recently.

These two factors together mean that inaccuracies in the foot/ankle kinematics (especially centre of pressure and ankle joint centre) are generally the most deleterious to accuracy in gait analysis.

By the way, Nerrolyn Ford has kindly proposed a form for observational gait analysis, which can use in future Cases of the Week (one coming soon, incidentally - watch this space!). Ray Smith in Perth, Australia, has converted it to a CGI web script, and I'd be grateful for your feedback so we can modify it before we start using it:


Best wishes, and congratulations to all for an excellent discussion,

Chris -- Dr. Chris Kirtley (Kwok Kei Chi) MD PhD Assistant Professor Department of Rehabilitation Sciences The Hong Kong Polytechnic University


Yes the situation can be as bad as all that! Not so much in terms of the errors in acceleration calculation etc. (I'll keep out of the filtering controversy for now!) but in terms of the definition of the rotations about the long axes of the lower limb segments. Errors in these can lead to tremendous cross-talk between the different moment components, far worse than anything you will see in the kinematics. We are presenting a paper on this at this years ESMAC conference, where we have looked at the effects of rotational errors on 3-D data.

Its not so much foot position as the definition of the shank orientation (the relationship between the two will depend on your model).

The "ill-posed" problem seems to have relevance when you look at the ankle power that some individual's wearing solid AFOs can generate!

Jeremy Linskell Manager, Gait Analysis Laboratory Co-Ordinator, Electronic Assistive Tehcnology Service Dundee Limb Fitting Centre Dundee, DD5 1AG, Scotland

I partially gree with the comment of Prof Hatze. Yes, the ill posed problem is a big problem, but not for all movements. It is a big problem in javelin throwing for example when we want to calculate joint dynamics of the higher limbs. The external force (weight of javelin), in that case, though does not contribute as much as the GRF during walking. Again the foot placement, that was mentioned by Melanie, it will obviously influence GRF vector, but the overall joint dynamic will be contributed mostly by this GRF, and not the kinematics.

It is a coincidence, very quickly, we have (you can delete the message here because it might sound a bit boring) examined the effects of noise and filtering on 3-D joint dynamics of the lower limbs during walking. The associated paper is going to be submitted

The results showed that the effects of filtering were insubstantial for the calculation of the ankle moment and forces in all directions. The medial-lateral force in the knee joint were affected only at the high levels of noise. Knee abduction-adduction moment was affected by filtering with high levels of noise. Hip medial-lateral force was highly affected by filtering especially in the higher levels of noise. Hip anterior-posterior force was affected mainly in the higher two levels of noise. All hip moments were substantially affected by filtering.

If you still read this, again we have the other problem as Crhis mentioned of calculating the centre of each joint. Well, this is another BIG problem.



-- Giannis Giakas Division of SHE Staffordshire University Stoke-on-Trent ST4 2DF Email: g.giakas@staffs.ac.uk


I find these two statements contradictory:

(1) ... the overall joint dynamic will be contributed mostly by this GRF, and not the kinematics.

(2) ... we have the other problem ... mentioned of calculating the centre of each joint. Well, this is > another BIG problem.

Surely the joint centre is a kinematic measurement? The GRV on its own is meaningless - it needs to be combined with kinematics (joint centre data, if you like) to calculate joint moments.

You can never escape from the dreaded kinematics! But thanks for the interesting data on effects of filtering. I'm looking forward to the paper.


> ... tremendous cross-talk between the different moment components, far worse than anything you will see in the kinematics.

> The "ill-posed" problem seems to have relevance when you look at the ankle power that some individual's wearing solid AFOs can generate!

Maybe I'll stick to my good old 2D analyses for the moment (no pun intended!).


Chris -- Dr. Chris Kirtley (Kwok Kei Chi) MD PhD Assistant Professor Department of Rehabilitation Sciences The Hong Kong Polytechnic University

Although the following papers are not closely related to *Observational* gait analysis, some might be interested:

Holzreiter, S. H. and Kohle, M. E. (1993) Assessment of gait patterns using neural networks. Journal of Biomechanics, 26: 645-651.

Barton, J. G. and Lees, A (1995) Development of a connectionist expert system to identify foot problems based on under foot pressure patterns. Clinical Biomechanics, 10: 385-391.

Barton, G. and Lees, A (1995) An investigation of selected methods for the assessment of insole materials used within the shoe. In G. Atkinson and T. Reilly (eds) Sport, Leisure and Ergonomics, London, E. & F. N. Spon., 125-129.

Barton, J.G. and Lees, A. (1996) An application of neural networks for distinguishing gait patterns on the basis of hip-knee angle diagrams. Gait and Posture, 5/1: 28-33.

Barton, J.G. and Lees, A. (1996) A comparison of shoe insole materials by neural network analysis. Medical and Biological Engineering & Computing, 34/6: 453-459.

I recently graduated from the University of Oregon, where I performed my research on gait with the use of an AFO, with the ankle joint locked in various anatomical positions. In response to Melanie Wellers question regarding the placement of the ankle joint at foot strike, here are my two cents worth. I found that with the ankle joint locked at 5 degrees plantarflexion, or even 15 degrees dorsiflexion, in an apparently healthy person, that the GRF's were obviously very different from "normal" GRF's. Additionally, however, the kinematics showed significant deviations from normal also. Normal being defined as the individuals gait kinetics and kinematics without a bracing device. I don't know if this helps answer any of the questions anyone had, but I feel that both the kinetics and kinematics play an important role in understanding a clinical gait question. Without either one can you really have a full picture of what is happening? Thanks to all for a very informative discussion!

Sharna Clark M.S. Biomechanics smclark@oregon.uoregon.edu

Hello everyone! I would like to thank the contributors to this list, especially to Chris Kirtley, who must spend a lot time with this. I have been enjoying the discussion, although I have difficulties in finding the time to participate actively. In the discussion about ill posed problem I think there is something that could be worth a comment, since I feel that the possibilities for kinetic analysis is not so bad as it might seem.

Chris Kirtley (Kwok Kei Chi) wrote:

Errors in the location the centre of pressure (which is calculated from the force plate data, and not from a kinematic devices) will directly influence the calculated values of the ankle joint moment and this error will be transferred to the more proximal joints.

If we have a large error in the ankle joint position, this will give a large error in the ankle joint moment, However, this error will not be transferred to the knee and hip joint! Therefore I do not agree that the foot is more important than more proximal segments!

Instead, if we have an erroneous position of the ankle joint, this error will be totally cancelled in the calculation of the moments at the knee joint. This can be seem by a simple example. For simplicity let us assume that the foot and shank are weightless (this is almost perfectly true for kinetic analysis of normal and slow walking, since the ground reaction force (GRF) is so much larger than the gravitational and inertial forces created by the foot and shank mass). Further assume for simplicity that the ground reaction force is vertical (this is not at all essential but makes the argumentation simpler to follow).

Let Xf be the horisontal position of the centre of pressure (origin of the GRF vector), Xa be the horisontal position of the ankle joint centre (note that the centre of pressure at the joint surface should be used, and not the instantaneous centre of rotation), and Fv be the magnitude of the vertical GRF.

Then the moment at the ankle joint, Ma is

Ma = (Xf-Xa)*Fv

(This is the external moment at the ankle joint, which is counteracted by the internal moment, of equal amplitude but opposite in sign. If we)

Now let us move up to the knee. Here the moment, Mk, can be calculated by using the shank as a free body. In this case the shank is affected from the foot by a force equal to Fv, and a moment equal to -Ma. Therefore, the moment at the knee joint is:

Mk = (Xa-Xk)*Fv + Ma

By inserting the above expression for Mz we get:

Mk = (Xa-Xk)*Fv + (Xf-Xa)*Fv = (Xf-Xk)*Fv

This last expression is exactly what we would get by using the foot and shank as a free body in the first place, and directly calculate the knee joint moment without passing the ankle. Thus we see that we do not need to bother about errors in the ankle joint position if we are only interested in the situation at the knee! The position of the ankle joint can be dislocated by 10 kilometers without affecting the calculated knee joint moment!

From the above formula we can see the essential variables for kinetic assessment are: The position of the centre of pressure on the ground (Xf), the position of the investigated joint (Xk), and the ground reaction force vector. Since we assumed a vertical force vector, we only have the vertical force component in the formula. In general, the joint moment at the knee is mainly determined by the distance from the joint center to the line of action of the GRF vector, multiplied by the GRF magnitude. Therefore the origin and direction of the GRF vector is of paramount importance for the estimation of joint moments. A force plate with high accuracy is therefore very important!

Of course the above reasoning must be interpreted with care. As soon as the inertial or gravitational forces are relatively large, then the motion of the corresponding segment must be carefully measured. In this case it is important to keep track of the location of the segment centre of mass, but distal joint locations are still irrelevant! (Although they are often used indirectly in the determination of the location of the centre of mass and therefore they do play a role).

There is also a discussion about filtering going on that I will not go into. I only want to mention, that for general dynamic analysis this can be really important, but for normal and slow walking, filtering is not an important issue for the ankle and knee, since the inertial forces are usually so small. (they become more important the more proximal we get, as Giannis has pointed out)

Althouh I tried to keep this short, I realise that I failed completely! Yours - Håkan --

Håkan Lanshammar Systems and Control Group, Uppsala University hl@SysCon.uu.se P.O. Box 27, S-751 03 Uppsala, SWEDEN Tel:+46-18-471 30 33 Fax: +46-18-50 36 11 http://www.syscon.uu.se

Hello! I really agree with you that it is hard to believe. However, if you compare two methods for estimating the external forces and moments at the knee: 1) The foot and shank are treated as two separate free bodies. Then if you start at the ankle joint end then going to the knee joint, it is possible to believe that the errors in the ankle joint position could effect the result. 2) The shank and foot are treated as one single free body. Then the positino of the ankle joint is obiously not relevant.

Methods 1 and 2 are equivalent if the segment are massless, and therefore ankle joint position is not relevant in this case even if method 1 is used! (If you find a flaw I will be most interesed to hear it!)

Good luck with your calculations! - Håkan

On 17.Sept. Jack Crosbie sent a list of "observed deviation - possible causes" couples. In some of them I couldn't find the logical link, maybe the list can help me.

Thanks, Gabor

Perhaps I might respond, since I was responsible for the original posting. There are three aspects to a potential source of a problem; one could be termed the consequence of a limitation in a structure the second the attempt to minimise the consequence of a limitation. The third is the "no need" scenario, in which the overall gait pattern obviates the requirement in some way. Thus the apparent paradox of a weakness (or "inability to develop adequate tension") in the hip abductors can lead to an increased lateral pelvic shift, due to the loss of lateral control, or may be anticipated by the patient, leading to a deliberate avoidance of the distorted pattern. Of course, the avoidance strategy may in itself lead to a distortion, as in this case. Decreased peak lateral displacement might also be associated with a gait pattern which has such short steps on the unaffected side that little lateral weight displacement is needed, the third option (although mot very likely).

Taking the last cause first, if the hip is not extended, there will be a reduced requirement for the knee to flex to clear the ground during swing (this is equally true in healthy subjects - the "no need" scenario). The plantarflexors may have quite a profound effect on all aspects of the lower limb movement. Inability to dorsiflex the ankle during the late stance (whether caused by adaptive shortening or "spasticity") prevents the leg from moving forward over the foot. If knee flexion is not initiated in late stance, then the only way to attain normal peak knee flexion during early swing will be the production of very large knee flexor muscle moments in early swing. Possible, but not perhaps very likely.

I hope this may be of some help. It is my interpretation and I would welcome other comments.

Jack Crosbie

Back to FAQ