that the whole of the language of early Britain can be recovered by the inferential methods that I have described. In all probability the ancient language contained many words that have been lost both in Welsh and Irish; and even of those that have survived, many, we may reasonably assume, had senses not precisely identical with those which they have come to express in historical times. Hence it is very likely that many of the British names found in Roman records will never be explained with certainty. Still, there are a considerable number of which the etymology can be clearly ascertained. It is of some importance to point out that the Roman spelling of British names, wherever we are able to control it, turns out to be remarkably accurate. There are no such strange deformations as are found in the English representations of Indian names in the early days of the East India Company. It is not true, as is popularly supposed, that the Romans euphonized the barbarous names by inserting vowels between the syllables; the phonetics of Latin and Old Celtic were so much alike that such a procedure was unnecessary. It is true that the mediaeval transcribers of Roman records have often blundered grievously, as we see, for instance, from the various readings in our copies of the Antonine Itinerary. But wherever the original reading of a Roman-British name is certain, we may rest secure that we have before us an exact reproduction of the native pronunciation.
One of the names which have been furnished with pretended explanations out of the Welsh dictionary is London, anciently Londinion—or, as Tacitus latinized it, Londinium. Most of the modern histories of London tell us that this name means either 'fort or town by the lake' (from the Welsh llyn, lake, and din, fort or town), or 'fort or town of ships', from the Welsh llong, a ship. Now these Welsh words are, unlike caint and gwent, genuine and ancient; but they have changed somewhat in form during eighteen hundred years.
