Watching someone walk, on purpose
By now you can grade a muscle, a tone, a reflex, and you have met the functional scales — FIM, Barthel, Berg — that ask whether a person can dress, balance, and move through a day. This last guide narrows onto a single act that those scales keep circling back to: walking. Walking is special because it is the one impairment a whole team, a family, and the patient can all watch at once, which makes it both the richest window and the most seductive trap. Everybody has an opinion about whether grandpa is 'walking better,' and almost nobody is watching the same thing. Observational gait analysis is the discipline of turning that crowd of impressions into one structured reading.
The trick that tames the chaos is the same one you used for sensation: structure. Instead of watching 'the walk' as a blur, you break it into the named phases of the gait cycle — heel strike, stance, push-off, swing — and then watch one body part through one phase at a time. Is the heel landing first, or is the toe slapping down? Does the knee lock or buckle in mid-stance? Does the foot clear the ground in swing, or does the hip hike and the leg swing out to the side to make room? Watched joint-by-joint, phase-by-phase, the same ten seconds of walking that produced five vague opinions produces one repeatable description that a colleague can read back and recognize.
Inside the gait laboratory
When the eye is not enough — when a child with cerebral palsy faces surgery, or a prosthetic limb needs precise tuning — the question moves to a gait laboratory, the home of instrumented gait analysis. Here, walking is decomposed into measurements no eye can make. The lab answers three different kinds of question, and it is worth keeping them apart, because they are not the same picture seen three ways — they are three genuinely different things.
WHAT THE GAIT LAB MEASURES
kinematics the MOTION itself cameras + skin markers track joint angles
(no forces) -> knee bends to 60 deg in swing
kinetics the FORCES behind motion force plates in the floor + math
(causes, not just shape) -> the hip pushed with X newton-metres
dynamic EMG WHICH muscle, WHEN surface/fine-wire electrodes on muscle
(the firing, not force) -> hamstring fires during stance (wrong time)
spatiotemporal the basic tallies step length, cadence, speed, symmetryWhy braid all three? Because a single channel can lie by omission. Cameras with reflective markers give you kinematics — the angles each joint sweeps through — and you can see that a knee fails to bend in swing. But kinematics alone never tells you why. Force plates buried in the floor add kinetics, the forces and torques driving the motion, so you can ask whether the knee won't bend because a muscle is yanking it straight or because nothing is pushing it. And dynamic EMG, electrodes reading the electrical chatter of muscles as they fire, answers the question neither motion nor force can: which muscle switched on, and at what instant in the cycle. This is a cousin of the needle electromyography you will meet on the electrodiagnosis rung, but aimed at timing during movement rather than at diagnosing a sick nerve at rest.
Honesty about the lab matters. It is gorgeous, expensive, and not magic. Skin markers shift over the muscle beneath them; a person walking down a marked runway in a lab is not the same as that person hurrying across a wet car park; and a beautiful curve still needs a human to interpret it. The lab earns its keep when a decision is costly and the eye is genuinely outmatched — planning multi-level surgery, settling a dispute about which muscle to weaken, fine-tuning a prosthesis. For most everyday rehab, careful observation plus a couple of simple, honest numbers carries the day. Which numbers? That is the next, and most important, part of this guide.
The three questions you ask of any measure
Step back from gait for a moment, because the deepest lesson of this whole rung is not about walking — it is about scales. The team's clipboard is crowded with measures, and a thoughtful clinician interrogates each one before trusting it. The formal name for what you are checking is the psychometric properties of outcome measures, and underneath the jargon sit three plain questions you can ask in everyday words.
- Reliability — does the ruler give the same reading twice? If two clinicians score the same patient, or one clinician scores the same unchanged patient on two days, do they land on the same number? A bathroom scale that shows a different weight each time you step on it is useless before you even ask if it is accurate.
- Validity — does it measure the thing you actually care about? A scale can be perfectly repeatable and still measure the wrong thing. Counting steps is reliable, but if your goal is independence at home, step count may be a poor stand-in for it. Validity asks whether the number tracks the real-world target it claims to.
- Responsiveness — can it detect change when change really happens? A measure may be reliable and valid yet so coarse that a patient genuinely improves and the score does not budge. A scale where most patients sit at the top or bottom (a ceiling or floor effect) cannot show movement, and so it is a poor choice for tracking progress even if it is excellent for one-time description.
These three are not interchangeable, and confusing them is one of the field's most common mistakes. A measure can be reliable but invalid, valid but unresponsive, responsive but unreliable. The choice of scale should follow the question: to describe a patient on admission you want validity and reliability; to prove that a six-week program helped, you above all need responsiveness. This is exactly why gait speed has become a favorite — it is dead simple to measure, strikingly reliable, validly predicts real outcomes like the ability to walk in the community, and it is responsive enough to register the gains rehab actually produces. It has even been nicknamed the 'sixth vital sign.'
Did it really change? The MCID
Suppose you re-measure a patient and the number has moved. The hardest question in all of measurement now arrives: is that change real, or is it noise? Every measure wobbles a little even when nothing has changed — the patient's effort, the time of day, the examiner's hand. The size of that wobble is the measure's noise floor (you will sometimes hear it called the minimal detectable change). A shift smaller than the noise floor tells you nothing; it could easily be the ruler breathing in and out. Only a change clearly larger than the noise is worth a second look.
But clearing the noise floor only proves a change is detectable, not that it matters. A patient might walk measurably faster and still not feel any different about their life. This is where the minimal clinically important difference, or MCID, comes in: it is the smallest change in a score that a patient or clinician would call meaningful — the difference that crosses from 'a real change' to 'a change worth caring about.' For gait speed, for instance, researchers have estimated a meaningful improvement to be on the order of a tenth of a metre per second; below that, even a real change may not show up in whether the person can cross a road before the light turns. The MCID is the bar a result must clear to count as a win, not just a measurement.
Putting it together: a walk worth trusting
Picture a man eight weeks after a stroke, halfway through his program. At admission the team timed his gait speed over a marked four metres and added a six-minute walk test to capture endurance, not just a burst. Today they repeat both under the same conditions — same hallway, same shoes, same brace, same time of day, no fresh pep talk. His speed has risen by a clear, above-noise margin that crosses the MCID, and his six-minute distance is up too. Because the measures are reliable, valid, and responsive, and because the change beat both the noise floor and the meaningfulness bar, the team can say something they could not say from impressions alone: he is genuinely walking better, and by roughly this much.
Notice what each tool contributed and where each stops. Observation told them how he walks — a knee that locks, a toe that drags. The simple timed tests told them, trustworthily, how far and how fast, and whether the change was real and meaningful. A full gait lab was never needed here, and a wise team does not order one to decorate a chart. And no number told them whether the gain was true recovery or smart compensation — that judgment still belongs to the functional assessment as a whole, read by a thinking clinician. This is the quiet creed that closes the whole assessment rung: measure carefully, choose the right tool for the question, respect what each number cannot say, and let the patient's life — not the prettiest curve — be the thing you are finally trying to move.