Languages change. This is not controversial. What is controversial is why, because from a biological perspective, the change looks wasteful. Communication is adaptive. Communication requires shared conventions. Shared conventions are undermined by change. So why hasn't evolution stabilized language — locked in a grammar, frozen the vowels, fixed the word order?
Balthasar Bickel and colleagues argue, in a 2024 paper in Physics of Life Reviews, that the question is backwards. Evolution did not fail to stabilize language. It stabilized the instability. What biological evolution selected for was not any particular language configuration but the capacity to keep changing — a system whose characteristic property is stationary dynamics, where the system never reaches equilibrium but maintains stable probability distributions of change.
The mechanism that makes this adaptive is social differentiation. Linguistic variation — dialect, register, slang, accent — is how populations mark group boundaries. Who sounds like whom determines who trusts whom, who cooperates with whom, who mates with whom. This requires variation. But variation that converges to a fixed pattern ceases to differentiate, because all groups eventually adopt the same signals. The differentiation engine requires continuous production of new variation, which means the system must keep moving. A language that stopped changing would lose its social function.
This is not drift. Drift is change without function — the neutral theory applied to culture. Bickel's claim is stronger: the change itself is the function. The restlessness is not noise that biological evolution failed to filter out. It is the signal that biological evolution selected for.
The empirical prediction is specific. If language change is functionally adaptive, then the rate and pattern of change should show statistical regularity across unrelated languages — not because they share ancestry, but because they share the same adaptive pressure. Companion work using the Grambank database of 1,700 languages confirms exactly this: grammatical features converge repeatedly and independently across language families, not toward a single configuration, but toward a distribution. Languages don't settle. They orbit.
The standard evolutionary question about language asks: why does this valuable system keep breaking? The stationary dynamics answer inverts it: a valuable system that stopped changing would lose the property that makes it valuable. The instability is not a failure to optimize. It is the optimization.