River name survival

Why do a region's river names outlast the language of every people who have ruled it?

Across much of Europe, the towns have been renamed, the administrative districts have been invented and abolished, the field names have turned over — and the rivers are still called what they were called before any of the recorded languages arrived. Names like Thames, Severn, Trent and Avon are not explicable from English, or from Latin, or from Celtic in any straightforward way. They are older than the evidence.

The general fact that names outlive their languages is not itself surprising — names are inherited rather than reinvented. But rivers are a special case even among names: they are the most conservative class in the landscape, more durable than settlements, hills, or regions. That specificity is what needs explaining. Something about a river, rather than something about naming, is doing the work.

Start with what has to happen for any name to be replaced, because replacement is the event we are trying to explain the absence of.

A newcomer arrives and needs to refer to a thing. They can either learn the existing name or coin a new one. Coining requires that they have no need to communicate with anyone who already uses the old name, or enough power to make their version stick in every context where the thing comes up.

Now compare a settlement with a river against that requirement.

A settlement can be renamed because it can be possessed. It has an owner, a charter, a lord, a tax record. Renaming it is an administrative act by someone with the standing to perform it, and the new name propagates through the documents that matter. It can also be destroyed and refounded, which resets the name entirely.

A river has no proprietor. There is no document in which it is registered, no ceremony that renames it, and nobody with the standing to declare a new name binding. It also cannot be destroyed, moved, or refounded — it is exactly where it was, doing exactly what it did.

That asymmetry is the first half. The second half is about who has to talk about it, and when.

A newcomer to a region can go a long time without needing the name of a particular field or hamlet. They cannot go a long time without needing the name of the river, because a river is what you must cross, what you navigate, what marks the edge of a territory, what a settlement is sited on, and what floods. It comes up on the first day, in conversation with the people who are already there — traders, guides, the population being displaced or absorbed. And the only name available in that conversation is the existing one.

So the newcomer learns it, before they have the numbers or the authority to impose anything, and by the time they might have had either, their own people have been using the inherited name for a generation.

Follow this through several conquests and the pattern is self-reinforcing. Each incoming group learns the hydronym from the group before, who learned it from the group before them. The name is transmitted through a chain of speakers none of whom shared a language with the originators — which is exactly why it ends up meaning nothing in any of the languages spoken along the way.

And that explains the most-cited curiosity in the field: the river names that turn out, on analysis, simply to mean "river" or "water" in a lost language. When a newcomer asks what that is and gets an answer in a language they do not speak, the reply "the water" is heard as a proper name and passed on as one. Repeat across enough transitions and you get names that are, unwittingly, tautologies — the sort where a modern speaker says "river" in front of a word that already meant it.

This builds on the general account of place names rather than repeating it. That companion piece establishes the broad point: names survive because they are inherited rather than coined, and a landscape carries layers from successive languages. What is specific here is why hydronyms sit at the bottom of that stratigraphy — not how naming works, but what kind of referent a river is. Unownable, unmovable, indestructible, unavoidable in first contact: no other landscape feature has all four.

Size matters, and it is the cleanest confirmation available. The effect is strongest for large rivers and weakest for small streams, which are renamed freely — locally relevant, often described rather than named, and possible to ditch, culvert or divert out of existence. If the mechanism were about naming conventions, size would not predict conservatism; if it is about the referent, it should, and it does.

The evidence is thinner than the confident tellings suggest. Reconstructing pre-recorded hydronymy means working backwards from much later forms, and the field has a history of over-reading. The proposal of a single ancient European hydronymic layer, associated with Krahe, is contested — both its dating and whether the resemblances need a common source. Individual etymologies are argued over, which is why I have given none. The pattern is far better attested than any reconstruction within it.

Rivers do get renamed, and the exceptions are instructive. It happens where a state has reach and motive: colonial administrations naming rivers after officials, and later governments restoring or replacing those names. What those share is a centralised authority with map-making power and a documentary apparatus — precisely the machinery a river normally escapes.

The falsification test. If the mechanism is the referent's properties rather than naming practice, then within one region the age of a name should track the referent's size, permanence and unownability — major rivers oldest, then hills, then settlements, then fields — and river names should be renewed exactly where a centralised mapping authority existed. If settlement names in a region proved systematically older than its river names, this account would be wrong.