Inside the Signal 01Technical ArticleGaitAI Research8 min read
What Does Gait Symmetry Actually Mean?
Left against right, in time and in space — and why one number is never the whole answer
Symmetry is one of the most quoted movement measures and one of the least examined. It compares the left side of a walk with the right — but a walk has several things that can be compared, they can disagree, and nobody is perfectly symmetric to begin with. This is what the measure is, what moves it, and what it can and cannot say.
Before you scroll: two gait cycles, left and right, drawn as time. Drag one side away from the other and watch what the word asymmetric starts to mean.
Drag the track, tap a step, or use ← →
Text description
Two gait cycles drawn as horizontal bars of time, one for the left leg and one for the right, on a shared clock. Each bar is filled while the foot is on the ground (stance) and open while it is in the air (swing); a tick marks each heel strike. The right cycle starts half a cycle after the left. A control chooses which comparison to move — stance duration, swing duration or step timing — and a seven-step track shifts the right side away from the left, or the left away from the right. Stance: one foot stays on the ground longer and the other's stance shrinks. Swing: one leg spends longer in the air. Step timing: the right heel strikes early or late, so steps alternate short and long. The left cycle's boundaries stay drawn over the right bar as dashed guides so the misalignment is visible. The only reading is qualitative: symmetrical, mildly asymmetrical, asymmetrical or clearly asymmetrical, with a sentence saying what moved. Illustrative: the proportions are drawn, not measured, and no ratio or index value is shown.
Short on time? Read the 2-minute version
- A gait cycle is one stride: stance, when the foot is on the ground, and swing, when it is not. Symmetry compares the left cycle with the right.
- There is not one symmetry. Stance duration, swing duration, step timing and step length can each be compared, and they can move independently.
- Nobody is perfectly symmetric. Healthy walking carries a small, stable asymmetry, so the reference is a person's own pattern, not a mirror.
- Surface, footwear, a carried bag, walking speed, fatigue, camera viewpoint and pose error all move a symmetry measure without the person changing.
- The useful reading is a change against the same person's baseline, with capture quality attached — not a single value against a population cut-off.
- Symmetry is decision support: a reason to look closer, never a finding on its own.
The full article follows, with the diagrams and the caveats.
Ask what a movement system measures and symmetry is one of the first answers. It sounds simple: does the left side move like the right? It is also one of the least examined measures in the field, because the simple question hides three harder ones — symmetric in what, compared with whom, and measured how well.
This story takes the measure apart. Not to argue against it — symmetry is genuinely informative — but so that when a system reports an asymmetry, a reader knows what has actually been compared and what has not.
The figures are illustrative: the gait cycles are drawn to make each kind of asymmetry visible and carry no measured values. Nothing here is a clinical threshold, and nothing here diagnoses anything.
Two legs, one cycle
A is one stride of one leg: from the moment its heel strikes the ground to the moment the same heel strikes again. Inside that cycle the leg is either on the ground — stance — or moving through the air — swing. In ordinary walking stance takes the larger share, and for a short period both feet are down at once.
The other leg runs the same cycle, offset by roughly half a cycle. Symmetry is a comparison between the two: how alike are the left cycle and the right? That framing already contains the first surprise. The two cycles are drawn on the same clock, so they can differ in how long each phase lasts, in when each begins, and in how far the body travels during it — and those are three different comparisons.
What a symmetry measure actually compares
The hero figure above lets you move each comparison on its own. Choose stance duration and one foot stays on the ground longer than the other. Choose swing duration and one leg spends longer in the air. Choose step timing and the right heel strikes early or late, so steps are no longer evenly spaced even if every phase has the same length.
Temporal
How long, and when
- Stance time, swing time, double-support time — left against right
- Step time: the interval between one heel strike and the next
- Readable from a side view and from a wearable's timing signal
Spatial
How far
- Step length and stride length, left against right
- Foot placement relative to the body's path
- Sensitive to camera geometry, which the next story in Engineering GaitAI takes up
Because these are different comparisons, they can disagree. A person who protects one leg may spend less time on it — a temporal asymmetry — while keeping step lengths even. Another may take a shorter step with one leg at a normal rhythm. A single reported "" hides which of these happened, and the interpretation is different for each.
Why nobody is perfectly symmetric
If the reference were a mirror, everyone would fail it. Healthy walking carries a small asymmetry — most people have a preferred leg for propulsion and another for support, the way they have a preferred hand — and that asymmetry is remarkably stable from day to day for the same person.
The informative reference is not symmetry. It is this person's own asymmetry, and whether it has moved.
This is the same argument the fourth Foundation makes for fall risk, arriving from a different direction. A symmetry value compared against a population cut-off tells you where a person sits among strangers. The same value compared against that person's own earlier walks tells you whether something changed — which is usually the question that was being asked.
What can make it move
A change in symmetry is a change in the comparison between two cycles. The person is one thing that can move it. Here are the others, and every one of them has produced a convincing asymmetry in a recording of someone who was walking exactly as they always do.
None of these is exotic. They are what a corridor, a care home or a sports hall looks like on a normal day. A system that reports symmetry without recording the conditions it was measured under is reporting a number it cannot interpret.
Reading a change, not a number
Put the pieces together and a defensible reading of symmetry has four parts. Which comparison moved — stance, swing, timing, length — because they mean different things. How far it moved from this person's own baseline, because the population is the wrong reference. Whether the capture supports the comparison at all, because a foreshortened view or a filled-in joint can manufacture the whole effect. And whether it persists across walks, because a single asymmetric recording is one observation.
This is why, in a movement-intelligence system, symmetry does not arrive alone. It arrives with the cadence and of the same walk, the capture-quality judgement made before any number was computed, and the history it is being compared against. The signal is only as informative as the context that travels with it.
What symmetry can and cannot say
- It can say that one side of a walk is behaving differently from the other, in a named way, compared with the same person's earlier walks.
- It can say that the difference appeared after a change in conditions — or did not, which is the more interesting case.
- It cannot say why. A protective pattern, a stiff shoe, a slope and a pose error can all produce the same number.
- It cannot say that anything is wrong. It says that something is worth a closer look by someone qualified to take one.
That last line is not a disclaimer bolted on at the end; it is the design. A symmetry measure that knows what it compared, what it was compared against and how well it was captured is a good instrument. Read as a verdict, the same measure is a source of confident mistakes. The word is the difference between the two.
Symmetry is the most quoted movement measure because the idea is so clear: left should look like right. The measure earns its place when the idea is made precise — which comparison, whose baseline, what capture, how many walks — and loses it the moment a single number is asked to stand for all four.
Inside the Signal continues with the reference that makes a change readable at all: what a personal movement baseline is, and how it differs from a population threshold.
Explore symmetry in the Movement Intelligence Lab →Written by GaitAI Research. GaitAI produces movement measurements and decision-support outputs for research, clinical and operational teams — not diagnoses. Published research is listed in the publications library.
Inside the Signal · 1 of 2
You now know that symmetry is a family of comparisons — stance, swing, timing, length — and that a change is readable only against the same person's baseline.
Next
What does that baseline actually consist of?
Inside the Signal 02What Is a Personal Movement Baseline?The reference that knows the personContinue