The uniformitarian principle and pre-English river-names

Most reputable linguists nowadays take for granted in historical investigations what is called the uniformitarian principle, that though the frequency of linguistic phenomena was not always the same in the past as it is now, the causes which operated in human language were. Proper names tend to have a somewhat semi-detached relationship to the history of their languages at large. They have not much been grist to the mills of those scholars who have sought to define universals of language, and the relevance of the uniformitarian principle to particular questions is perhaps easier to overlook in them. One area where I think this has been done is that of pre-English river-names, that is, names such as Thames and Humber which are not meaningful in English but have been adopted from languages earlier spoken in what is now England (and neighbouring parts) in which when coined they were meaningful.

The language from which English directly adopted them was that of the Celtic Britons, Brittonic (and/or occasionally perhaps Latin as spoken by Britons, but forms of river-names will hardly have differed between these). The linguistic material of some is perspicuous, showing them to have been either coined as compounds in Brittonic (thus Candover and Douglas, ‘fair water’ and ‘black stream’ respectively) or created in the process of borrowing into English: thus Avon = Welsh afon, the common noun for ‘river’, whose etymon can no more have constituted a complete river-name in Brittonic than ēa ‘river’ did in Old English or river does in modern English. Some make sense as other kinds of formation in what would for various reasons be earlier stages of Celtic, thus Trent. Some are suffixal formations on bases not perspicuously Celtic: thus most of the names of major rivers, Thames, Humber, and the like. It is reasonable to regard them as borrowed from (an) earlier-spoken language(s) into Celtic, just as they later were from the Brittonic stage of Celtic into English.