You would expect that if you give people complete freedom, things would go in all directions. Give ten groups the same task, without examples, without consultation, and you expect variation. A lot of variation. And you will probably get that. But rarely is that variation completely random. If you look closely, patterns start to appear. Solutions that resemble each other, even though no one copied from anyone else.

A recent study did something similar, but on a much larger scale. Researchers looked at hundreds of languages worldwide. Languages that developed over thousands of years, often independently from one another. If there is any room for randomness, it should be here. But that is not what they found.

Languages differ enormously. In sounds, in words, in grammar. But they do not differ randomly. There are patterns. Preferred directions. As if languages, despite all their freedom, keep ending up in similar kinds of solutions.

That may not sound surprising. Linguists have long talked about so-called “universals”, rules that would hold across all languages. But in recent years, that idea has come under pressure. Many of those supposed universals turned out to be less robust than once thought. And that is exactly what makes this study by Annemarie Verkerk and colleagues interesting.

When you properly control for language relationships and geographic influences, a large share of these universals disappears. What first looked like a near-universal pattern often turns out to be the result of languages being related, or of neighbouring languages influencing each other. In other words, much of what “occurs everywhere” simply occurs often in the same places.

What remains is a much smaller set of patterns. Roughly a third of the proposed universals still hold. And that is precisely the point. What remains cannot be dismissed as a coincidence. These patterns are mainly about how languages organise and structure things. About consistency. About how elements relate to one another. And perhaps more importantly, these patterns do not emerge once. They emerge again and again, in different parts of the world.

It is hard to call that a coincidence. Languages, despite their diversity, seem to be shaped by shared pressures in how people speak, listen and understand. They converge on solutions that are workable for communication.

There is no single correct way. But some ways seem to work better than others. You can compare it to water flowing downhill. It can take different paths, but it often ends up in the same valleys. Or to cities that develop independently, yet end up with similar structures.

There is variation. A lot of variation. But within limits.