When you walk, your center of mass is highest at mid-stride — at the moment when one foot is directly beneath you and the other is in the air. This is the peak of an arc. You spend each step falling off it.
The physics model that captures walking is called an inverted pendulum: a ball on the end of a rigid rod, pivoting over the ground contact point. As you vault over your planted foot, gravity pulls you down and forward. Your kinetic energy and potential energy trade off in opposite phase — as you descend toward the next step, you speed up, and as you climb to the next peak, you slow down. This exchange is nearly free. Your muscles don't do much work during mid-stride. They do their real work at the moment of transition — when one foot hits the ground and you have to redirect your center of mass from downward to upward again. That push is where the metabolic cost of walking lives.
Not in the motion. In the interruption.
Running is a different thing entirely. Your center of mass is lowest at mid-stance — the opposite geometry. Your leg isn't a rigid rod you vault over; it's a spring that compresses under load and then releases. The Achilles tendon stores something like 35% of the energy of each running stride elastically. Running is not the same mechanism as walking, just faster. It's a different physical solution to the same problem.
So there has to be a transition between them. What determines when it happens?
The answer turns out to be a ratio. Speed squared, divided by the product of gravitational acceleration and leg length: v²/gL. This is the Froude number, borrowed from hydrodynamics, where it describes the ratio of inertial force to gravitational force in a fluid. Applied to walking, it describes the ratio of centripetal force to gravitational pull as the body moves through the arc. When Froude equals 1.0, the math says the foot can't stay on the ground — the centripetal acceleration required to maintain the arc exceeds gravity. You go airborne. Walking becomes geometrically impossible.
In practice, people transition to running around Froude 0.5 — well before the strict mechanical limit. This turns out to be about energy efficiency: the metabolic cost of the two gaits crosses around that value, and the body switches to whichever is cheaper. The transition isn't purely physics; it has a biological layer. But the biological layer sits on top of a physical constraint that doesn't bend.
What's strange is that this dimensionless number predicts gait transitions across animals that differ by orders of magnitude in size. A child with legs half as long as an adult's transitions at a lower absolute speed, but the same Froude number. A horse with much longer legs transitions at a higher absolute speed, but the same Froude number. Froude 0.5 appears in humans, dogs, horses, ostriches. When paleontologists find dinosaur trackways in sediment — preserved fossilized footprints that record stride length and estimated hip height — they can calculate Froude numbers from stone and infer which animals were walking and which were running. The number travels across a hundred million years.
And then it fails.
Elephants transition at Froude 0.24, roughly half the predicted value. And they never develop a full running gait — no aerial phase, never all four feet off the ground at once. At high speed, they shift to a different movement pattern: still grounded, but their hindlimbs begin behaving like springs even while the front legs maintain the pendulum mechanics. A hybrid. Not the canonical switch but something improvised under constraints the formula doesn't account for.
The constraint is impact force. When a body goes airborne and returns to ground, the skeleton absorbs the collision. That force scales with mass in ways the Froude number ignores. A running elephant would require a skeletal architecture that doesn't exist and probably couldn't be evolved quickly enough to matter. So elephants walk fast instead — they push the walking gait as far as it goes, achieve speeds that look like running to a casual observer, and never make the switch. They're doing something the formula said they should have done at a lower speed, but the formula didn't carry the mass term.
I find the failure point more interesting than the universality. The Froude number is elegant — a clean dimensionless ratio that extracts something real about gait mechanics and holds across radically different bodies. But it's a model of the center-of-mass dynamics, and center-of-mass dynamics are not the only thing a body has to manage. The universality tells you what's shared. The exception tells you what's hidden in the shared part: an assumption about what size costs, implicit and invisible until you get large enough to violate it.
Most laws are like this. The domain of validity is part of what the law means. The elephant is not an error in the theory. It's a probe into where the theory's abstraction lives.