How old is English?



Saxon Shore Fort Names Look Germanic

Roman-era names for the nine forts of the Saxon Shore can all be explained with the aid of Germanic languages plus Latin in ways that accurately fit their topographical situations. This pattern also extends inland to other ancient names. Very few, if any, ancient place names in south-east England were derived from Celtic languages.

The forts of the Saxon Shore have been much discussed. Were they built t o defend against seaborne Saxon attackers? Or was the coastal belt of south and east England already settled by Germanic peoples in Roman times?

This issue goes to the heart of a mystery about the English language. Why does it contain almost no early Celtic loanwords? Is it really true that something like 3 million people shifted from speaking language(s) like early Welsh in AD 400 to speaking Anglo-Saxon by AD 700? Could such a social transformation really be wrought by the few tens of thousands of Germanic immigrants who crossed the North Sea after the Roman army left?

M ost archaeologists no longer believe in an Anglo-Saxon “invasion” because they generally observe no great interruption in the settlement patterns of rural areas during the Dark Ages. Therefore archaeologists tend to be sympathetic to an idea, expressed most prominently by Oppenheimer (2006), that the Iron-Age boundary between Celtic and Germanic languages lay well inside Britannia rather than at the North Sea.

On the other hand, many historical linguists are very protective of established chronologies, especially that of Jackson (1953), for changes in the pronunciation and meanings of words. Therefore linguists often express views like this: “ when the first Anglo-Saxon fleets arrived ... Celtic dialects were spoken throughout most parts of Britain” (Laker, 2002).

Everyone agrees that south-east of a line across England from the Wash to the Solent there are v ery few places with a Celtic-looking name. But is their number artificially low? Is Coates (2002) correct in claiming that “the level of Celtic survival in England, as shown by place-names, is somewhat greater than has been admitted until now”? Or are many supposedly Celtic etymologies in south-east England a mistaken hangover from 19th-century nationalism? Will better information and more careful study make their number tend towards zero?

Goormachtigh and Durham (2009) examined place names in and near Kent that are widely cited as evidence for Celtic speech in Roman times, and found alternative (often better) Germanic explanations for all of them. That first paper needed more sophisticated following up, which this paper (and several others) try to supply.

Unexpectedly, this paper develops a focus on the Saxon Shore forts, all of whose names can be explained (in the sense of fitting their topographical situations and purpose) better with the aid of Germanic languages (plus Latin) than with Celtic. It also presents evidence that this pattern holds for inland place names as well as the forts on the coast. and extends from Kent almost to the Wash and the Solent.

Forts of the Saxon Shore

The definitive list of Saxon Shore forts is in the Notitia Dignitatum, a sort of Roman Army order of battle dating from about AD 400, which exists now only in late medieval copies of copies. From the available manuscripts Seeck (1876) compiled a master version, which is widely reprinted on the Internet. Under Comes Litoris Saxonici per Britanniam these fort names are listed:



Portum Adurni

Portchester Castle


Pevensey Castle












Great Yarmouth



Archaeological work on Saxon Shore forts was summarized by Johnston (1977) and more recently by Pearson (2002). For a reasoned argument that the forts were not so much military bases as guarded warehouses for taxes paid in kind, see Pryor (2004). For place name information, prime sources are Rivet and Smith (1979) for Roman Britain, Gelling and Cole (2003) and Mills (2003) for England, and Sims-Williams (2006) for Celtic Europe.

Abbreviations used here are: ND = Notitia Dignitatum, AI = Antonine Itinerary, ME = modern English, OE = Old English or Anglo-Saxon, PIE = proto-Indo-European, PG = proto-Germanic, RC = Ravenna Cosmography. All the main ancient documents may be consulted online, as well as dictionaries of OE (Bosworth-Toller), PIE (Pokorny, 1959), and PG (Orel, 2003). In what follows, the ND forms of names are used, with no attempt to shift them into a Latin nominative form, or to replace them with a consensus spelling based on multiple documents.


Pevensey Castle was built on top of a Saxon Shore fort, which in Roman times lay on a peninsula of land rising above coastal marshes. The –rid- part of its name has hitherto been attributed to Celtic ritu 'ford', based largely on Jackson (1948), who wrote that “to the eye of the Celticist the correct form is obviously Romano-Britain Anderitum ”. He was apparently influenced by the common phonological evolution of T into D, a variant manuscript reading Anderitos , and an Anderitum in Aquitania.

OE rið 'small stream' (from PG * riþaz , PIE * rei 'to flow') is just as good. Its archetypal descendant place name is Ryde (the childhood home of one author), where one can still see a topographical situation that perfectly parallels ancient Anderidos, where a stream used to open into a marsh and then across a dover, i.e. a 'two sided or split beach'. The word was used in Anglo-Saxon charters in phrases like ' on ð riþi' , and as a dialect term it is still applied in Hampshire to the small channels through mud at low tide in Portsmouth and neighbouring harbours (Grundy, 1922; Coates, 1991).

The Ande- part has prompted much discussion, with a majority view holding that there was a Celtic intensive prefix * ande- (seen in some Gaulish names), so that the whole name meant 'great ford'. However, Germanic languages have better ways to explain ande : PG * anda- 'along, etc' has numerous modern descendants in words that begin with ont- , ent- , on- , and and- , while the Latin/Greek prefixes anti- 'against' and ante- 'before' are related. Everything converges on a sense of 'close to'. So a fluent ME translation of Anderidos would be ' brookside '.


This Saxon Shore fort, plus an apparent precursor smaller fort and an adjacent settlement, are visible from the air as crop marks between Brancaster and Brancaster Staithe, on the north coast of Norfolk, near the Wash (Edwards & Green, 1977). A Celtic word bran- 'crow, raven' has been suggested to explain the first part of the name Branoduno (Rivet and Smith 1979). The second part, dunum , was a common component of Latin place names, originally a Celtic loan word, whose sense developed from hill to fort and ultimately to town. In fact Branoduno does sit on a slight elevation, the nearest approach to a proper hill that exists in that part of East Anglia.

The element Bran- in English place names is variously explained as 'broom', 'brain', 'burnt' and a personal name, but much the most likely root here is PG *brunnon 'spring, fountain'. The modern river Burn runs just too far away to provide fresh water to the site of Branoduno, though the configuration of the coast has changed since Roman times because of silting and agriculture. It seems likely that there were other stream(s) running off chalky higher ground to the south, of the type often called bournes. The A in bran- is unremarkable because so many vowels appear in words that embody an idea of fire or water springing up, such as modern brand, brew, broth and burn. In summary, rather than 'crow fort' Branoduno makes better sense as ' stream fort '.


Dover is probably the place name most commonly cited as evidence for Celtic speech in south-east Britannia, because it was called Dubris in Latin. The standard argument was set out by Rivet and Smith (1979) in these words: “t he British name was *Dubras 'waters, stream' (perhaps 'streams'), plural of *dubro- 'water' (Welsh dwfr, dwr , Cornish dofer, dour , Breton dour ; Old Irish dobur ) … all records of the name, even those of the Antonine Itinerary set in a grammatical structure, show it as a locative plural in -is ”.

There are three separate difficulties with this logic. The first is a matter of Latin grammar: dubris was almost certainly not plural, and probably not locative. The noun cases in Latin of place names are often misinterpreted (Arias, 1987; Williams, 2007). The ancient wording ad Portum Dubris occurs twice in AI, a document from about AD 300 but probably based on Roman army marching orders from before AD 100. Ptolemy, writing in about AD 140, did not mention Dover. Then, as a single word, Dubris occurs once in ND, once in Tabula Peutingeriana, and twice in RC though once mis-spelled Durbis .

So dubris was most likely a genitive singular. After ad portum , mediaeval Latin writers unhesitatingly used the genitive not the locative. In classical Latin, the Itinerarium Maritimum uses an apparent genitive plural in ad portum ritupium . And Julius Caesar wrote ad portum itium about his departure point for the invasion of 55 BC, where itium has been interpreted as an adjective but looks suspiciously close to genitive plural ituum 'departures'.

The second difficulty is also linguistic. Most English place names are duplexes (qualifier plus generic) among which a simplex 'waters' would be a rarity, while 'port of the waters' looks daft. However, if a watery meaning is considered acceptable, Celtic is not the best source. The PIE root * d h eub- has many descendants, including the English words dub (northern dialect for a dark or muddy pool) plus deep, dimple, and dip. According to Pokorny (1959), the exact word dubris existed in Illyrian. This sounds remote, but Illyrian was the first language of many of Rome's best soldiers and sailors, including officers up to the level of emperor and part of the Britannia garrison between at least AD 105 and 400. Julius Caesar spent the winter of 55 BC in Illyria, near modern Dubrovnik, between his two trips to Dover. In fact, Illyrian was just one of a band of ancient languages that got squeezed out when Latin expanded towards Germanic or Celtic.

The third difficulty is topographic. More than 170 place names around England contain an element ofer , or its variants ufer , yfre , and ora (Gelling and Cole, 2003). Why should Dover in Kent (and a few other places ending in –dover) be different from all the other –over places for which an Anglo-Saxon origin is not disputed?

The exact meaning of OE ofer and its relatives is complex, and there may have been two distinct forms of the word . The one with a long vowel O, like its modern Dutch or German cognates, could mean river bank or shore. Ekwall (see Mills, 2003) thought it primarily applied to a firm beach or gravelly shore, which suggests that Caesar's landing on the beaches could have been remembered like D-day in 1944. The ofer with a short O may have evolved into ME over, with links to upper, hyper, super, etc. Whatever the precise etymology, ofer place names are associated with a distinctive topographical feature – a flat-topped ridge with a convex shoulder, like the end of an upturned canoe, usable as a landmark by ancient travellers on land or water.

Near the south coast, ora was the commonest name element for this type of landmark. In Latin ora meant 'sea shore' and Latin would have been the language of command in the Classis Britannica and perhaps merchant ships too, so Cole (1990) wondered if ora came to mean 'land ahoy', then passed as a loan-word into Saxon speech to partially supplant a native Germanic ofer that was retained in Anglian areas further north.

All round the coast, from Exeter in the west to Maidenhead high up the Thames, every port of any significance in Roman times seems to have been marked by at least one ofer/ora place. More than 20 ofer/ora place names cluster near the Isle of Wight, where to this day boats navigate mostly by reference to lighthouses, seamarks, forts, church steeples, etc. A sailor proceeding east from there along the coast and round into the Thames Estuary would pass by the following landmarks, all distinctively visible though often not particularly high:

The Owers

near Bognor, probably “Cymene's Ora”


i nland from Shoreham


near Hastings


west of Dover


i nland from Dover


s outh-east end of Wantsum Channel

Oar Farm

n orth end of Wantsum Channel


near Faversham

The Nore

near Sheerness


in the Medway

Within this list, Drellingore and River are actually the weakest candidates to be ofer/ora names, so an obvious suggestion is that the cliffs of ancient Dover were the landmark for its small port. Cole (1990) argued that ofer/ora landmarks needed to be green hills not white cliffs, but perhaps she was just intimidated by the traditional Celtic etymology. The sheer visibility of Dover's notch in white cliffs (plus its smaller neighbours at Shakespeare's Bay and St Margaret's) cannot have failed to impress every local seaman and the Classis Britannica.

If the name dubris (or a hypothetical nominative singular * duber ) described this seamark, the critical issue is how PG *obera- (Philippa, Debrabandere & Quak, 2007) could have picked up an initial D. Ancient scribes were not entirely consistent in how they wrote down phonemes that were not familiar in Latin, including D versus TH, U versus O, and B versus V or F, but early post-Roman spellings (Dofras, Dobrum, Doferum, Doferan, etc) seem to have stuck with an initial D.

In fact many modern ofer/ora place names have developed astonishingly far from their probable originals, often by gaining just a single letter prefix, presumably by transfer from a preceding word. They include several instances of The Nore (formerly atten ore ) and River (probably built from atter 'at there'), plus Hever. How then did Dover get its initial D? Was it from something like Dutch aan de oever 'on the shore' or OE æt ofer 'at the shore'? Or maybe Latin dua orae 'two shores'?

The answer can be found at two places well known to the present authors. In Bruges, Dijver or Dyver is the name of a historic boat-unloading basin flanked by two beaches. In the Isle of Wight, duver or dover is a generic local word for “a low-lying piece of land along the coast, subject to occasional inundation by the sea” (Pope, 1989). The common feature uniting all four duvers (Hamstead, Ryde, Seaview, St. Helens) is that they are (or were) more in the nature of a sandbar than a beach, because they had water or marsh on their landward side.

OE to- was a prefix denoting separation or division, common in OE words like tofær 'departure' or tobriting 'destruction', which have generally been superseded by ME words beginning with the equivalent Latin prefixes dis- or di- . So what at ancient Dover could justify a name with a sense of 'double bank' or 'split beach'? Maybe its clefts in the cliffs constituted double seamarks as discussed above?

However, there is another possibility to consider. North of Dover, the Goodwin Sands lie about 6 miles off the coast and can be partially exposed at low tides. South of Dover, deeper and further out, lies the Varne Bank. These sandbanks are notorious for swallowing ships, but they also created the historic sailing-ship anchorage off Deal known as the Downs. Sandbanks tend to move and the whole coastline of Kent has changed greatly since Roman times, especially since sea levels in the Channel have risen relative to the land by at least a metre since then. So it is entirely plausible that there was a substantial offshore sandbank at or near ancient Dover.

On balance , the original meaning of Dover/ Dubris is most likely to have been the same as dover/duver in the Isle of Wight: something like ' double beach '. The old Celtic 'waters' etymology is wrong and must stop being cited as evidence for Celtic speech in Kent.

As an interesting afterthought, ofer/ora places distinctively mark the routes of long-distance trade arteries into and inside England. Conventional thinking that the Saxons arrived in AD 450 has difficulty explaining why there are several inland ofer/ora places in Kent and Sussex that make sense only as landmarks for transport of iron smelted in the Weald at sites that ceased operation well before the Romans left.


The precise spelling of this name is open to discussion (Rivet & Smith, 1979), but clearly it refers to the river Yare (Ptolemy's Γαριεννου ποταμου, OE Gerne) near Great Yarmouth. The Saxon Shore fort itself was probably at Caister-on-Sea (described on an English Heritage web site) but it was backed up by other Roman-era forts some way inland at Burgh Castle, and possibly at Reedham.

For the start of the name, Celtic *gar- 'to shout' (cognate with garrulous) has been suggested, but this is unconvincing for such a gentle river. Garan, Welsh for 'crane' (Breeze, 2006) is better. However, Germanic gar 'spear' provides much the best fit, perhaps best known in the name Garibaldi. Among its many derivatives was gara 'promontory', or Gore in modern place names. This perfectly describes the spit of land on which Great Yarmouth sits. At the end of the name, -or (OE 'bank') makes one think of Spitbank Fort, off Portsmouth, built in AD 1859. However, more prosaically, -or was a common ending of verbal nouns in Latin.

The really interesting part of the name lies in the middle. In Latin, annona militaris was the annual tax or payment in kind levied on places such as Britannia to pay for the army. One of the major functions suggested for Saxon Shore forts was to collect, store, and export annonae . So the name Gariannonor can be confidently translated as something like ' tax collecting depot on the promontory' .

Leman nis

The name of this fort, at modern Lympne is often explained with a Celtic word for elm. The Germanic interpretation based on similarity to PG *laimon , Latin limus , Greek λειμον, etc (Goormachtigh & Durham, 2009), seems better linguistically and topographically. Appropriately, Lemannis “has seen greater post-Roman changes in its environment than any other shore fort” (Burnham, 1989). Estuaries silted up during the Roman era because light loess soils of the Kentish Weald eroded rapidly once trees were cut down for the iron industry and agriculture. So Lemannis meant something like 'marshy' or 'silted-up' or ' muddy' .


This fort lay near Bradwell, on the seaward side of the Dengie peninsula in Essex, south of the estuary leading to Colchester and Chelmsford. Its name (which later evolved into Ythancaestir) has long been a puzzle, and TH is not a normal letter combination in Celtic, though Coomes (2002b) managed to construct a meaning 'ox-place' in Celtic. More possibilities are available in the Germanic languages (notably based around the root that led to OE æðel 'noble' and o ðel 'home'), but the best ones arise from the prefix oþ- , which evolved into modern out-. In particular, oþirnan 'to run away, escape', gave ic oþierne ' I depart' . As the closest Saxon Shore fort to London, to the economic heartland of Roman Britannia, and to the mouths of the Rhine, Othonae has a logical claim to be translated as something like ' departures '.

Portum Adurni

This fort is generally accepted to be Portchester Castle, at the north of Portsmouth Harbour. No satisfactory etymology for its name has been offere d, but a Germanic possibility lies in the archaic English word dern, meaning hidden, secret, etc, found in place names and surnames like Durnford, Dornford, and Darnford, plus possible Continental cognates. Recorded dictionary spellings include dierne, darn, durn, and dyrne, from PG *darnjaz .

In OE charters, dierne was applied to natural objects that one comes across at the last moment, because they are hidden from view till close at hand (Grundy, 1922). This fits the situation of Portchester Castle, because to see it a mariner needs to leave the open Solent, pass through the narrow entrance of Portsmouth Harbour, go to the north of a huge lagoon, and then into one of many side creeks.

Modern English uses A as a prefix in the sense of 'at' (especially in nautical examples like astern, aboard, ashore) or as an intensifier (in await, aplenty or a-hunting) though most would be considered grammatical adverbs rather than adjectives. ' Port a-hiding' might not be dictionary English but it would certainly be understandable.


Reculver is regularly described as coming from Celtic 'great headland', which is manifestly wrong for the coastline there, even allowing for two millennia of erosion. In fact Reculver preserves an ancient name of the Wantsum Channel, combining two elements like the English sailing terms reach and whelve, from PG *raikjanan and *welbanan , so Goormachtigh and Durham (2009) confidently translated it as ' curved reach'.


T he ancient port near modern Richborough may have been named something like ' red tops' because it had prominent salt-making operations like the Red Hills of ancient Essex (Durham and Goormachtigh, in press).

Other forts

Besides the nine Saxon Shore forts listed in ND, there were other coastal forts that may have formed part of the same Roman military system. Walton Castle , lost to the sea off Felixstowe had OE name Dommoc, but its Roman name is unknown. Other possibilities include Carisbrooke, in the centre of the Isle of Wight, and Brough-on-Humber. And on the other side of the Channel there were coastal sites that formed part of the same Roman military system, near modern Aardenburg, Boulogne, St. Malo, and Brest.

One such fort had an interesting name, Clausentum, suggestive of Latin clausus 'enclosed, secret' and reminiscent of adurni perhaps meaning 'hidden'. Clausentum is generally taken to have been at Bitterne, up the river Itchen, off Southampton Water, but that location is uncertain, because manuscripts of AI appear to miss out a line next to it.

Other Roman-Era Names

This paper can fill in some information missed by Goormachtigh and Durham (2009). Vennemann (2006) discussed Thanet at length and firmly rejected a Celtic etymology, but suggested an origin from the Punic goddess Tanit . We continue to prefer some precursor to Middle English ed ten ende 'at the end', making Thanet analogous with Land's End and Finisterre.

Noviomago , mentioned in AI, has long been a puzzle since it does not correspond with any obvious Roman remains. However, the mystery was in fact solved by Arias (1987), who noticed that Romany army marching routes often bypassed named places (just like modern main roads), sometimes at considerable distances, and that road junctions were indicated by a grammatical marker, the accusative case, which was not always correctly transmitted by medieval copyists. Many apparent anomalies in AI disappear once this fact is recognized and allowed for. Noviomago in AI actually indicates a junction where the road coming from Noviomagus (Chichester and/or Fishbourne) met the Canterbury-London road.

Rivet (1980) recognized that the element Duro- occurred distinctively in Belgic areas , but did not fully abandon the idea of a Celtic root like Latin durus 'hard' in favour of something like PIE *d h ur- 'door'. It is now possible to expand the list of place names beginning with Duro- in Britain (Goormachtigh & Durham, 2009) with these ones from the Continent:

Durocatalaunum Châlons-en-Champagne, north-east France

Durocortorum Reims, north-east France

Duroicoregum Domqueur, north France, near the Channel

Duronum Etroeungt, north France, near Belgium

Durocassium Dreux, France, west of Paris

Durotincum Villejoubert, south central France

Durostorum Sinistra, Bulgaria, on the Danube

The bias towards northern, Belgic France is clear, but the two outliers are interesting. Durostorum lay on the south side of the Danube where the Roman Empire bordered on Germanic (Gothic) tribes. There may have been more than one Durotincum, but the best candidate lay within territory that became the Germanic kingdom of Toulouse.


The element ver in Durovernum Cantiacorum (Canterbury, the crossroads of Kent) may also be present in two other ancient names, Verulamium and Durnovaria, which were topographically similar. So it is worth exploring the idea that ver had a specific meaning that one might translate 'causeway' or what could become the common English place name Longford. Gelling and Cole (2003) recognised OE f ær 'passage, perhaps difficult passage' in modern Denver (on the Fen Causeway) and Laver (also in East Anglia). It descends from PG * faranan 'to travel, especially over water', had one obsolete meaning 'street', and survives in ME seafarer, wayfarer, ferry, etc.

Modern Canterbury still has a road called The Causeway, lying a few hundred metres north of the most likely location where causeway(s) over the river Stour would have n eeded to exist in Roman times. St Albans museum confidently shows one ancient causeway approaching Verulamium, one was archaeologically excavated at Strood (Jessup, 1932), and many more can be inferred from the landscape, for example approaching Rutupiae. In fact, causeways were common in the ancient landscape, wherever traffic needed to cross a broad, marshy river valley, and Pryor (2004) discusses the ritual symbolism that prehistoric societies used to organise their construction.

Durnovaria became modern Dorchester, where two channels of the river Frome run through what would once have been a marshy area. The natural Celtic translation of Durn (something like 'fist') is much harder to accommodate than the natural Germanic 'hidden', 'to spot at the last moment', discussed above. At Great Durnford, in Wiltshire, a long ford across the river Avon must be very ancient because it is overlooked by Ogbury hill-fort and marked with several tumuli. So a plausible translation of Durnovaria is ' hard-to-find causeways' .

Verulamium (near St Albans) was one of the earliest written place names recorded in Britain, on coins bearing inscriptions such as VERLAMIO. Rivet and Smith (1979) discussed and rejected all proposed Celtic etymologies, though Coates (2005) constructed a Celtic explanation as 'place of a man named Crooked Hand'. In fact Celtic is almost the only language family in which the element lam is not trivially simple to explain. Anglo-Saxon lam 'c lay, mud, mire, earth' was recognized by Gelling and Cole (2003) in modern place names Lamas and Lamarsh, both in formerly marshy areas of East Anglia. It descends from PIE *lama then PG * laimon , and its relatives include Latin lama 'bog' and limus 'slime', plus ME lime and slime, plus Portus Lemannis discussed above. Therefore Verulamium can be confidently translated as ' causeway over marsh '.

O rthography on ancient coin inscriptions in general seems to have adapted to changes in spoken language sooner than manuscripts, and naturally tended to err on the side of easier letters to engrave. Nevertheless, o ne may ask why an initial letter V was chosen for the initial consonant of fær by the coin mint of Tasciovanus, when the Roman alphabet already had a letter F available. Pronouncing initial F as V is a known regional feature of speech in south-west England and parts of Germany, and has long been suspected of originating much earlier than mediaeval documents can prove (Voitl, 1988).

Glacial landscapes

Combs in Suffolk was explained by Gelling and Cole (2003) as derived from OE camb , PG * kambaz 'comb'. They knew that comb is a common place-name element in northern Britain, meaning a ridge a bit like a cock's comb, but for Suffolk they could only speculate that it meant a series of low ridges running up to a river. In fact kaims or kames are what modern geologists call the elongated mounds of gravel deposited by melting glaciers, which can impose a wiggly course upon a river. One famous example of such a glacial feature, the Blakeney Esker, was left in Norfolk by the last Ice Age, but the glacial deposits further south in Suffolk were left by the Ice Age before that.

Camborico and Combretovium were two ancient places mentioned by AI, just inland from the Saxon Shore in Suffolk. Camborico (presumably an oblique case of Camboritum in messy handwriting) lay where the Icknield Way crosses the river Lark, near modern Lackford and West Stowe, while Combretovium lay further south, near modern Baylham. Both are widely assumed to contain a Celtic root *kambo- 'crooked' and therefore to mean 'curved ford'. However, PG * kambaz plus * riþaz 'stream' provides a Germanic alternative as ' comb stream '.

Nowadays the landscape signature of glacial deposits is often a series of worked-out, water-filled gravel pits, so it is hard to know exactly what ancient people recognised there and called 'crooked' or a 'comb'. Was it up-and-down changes in ground level (drumlins), side-to-side wiggles of a river, poor farmland due to stones and sand, or a combination? Clearly it is wrong to assume (as many authors do) that Camb- in place names is diagnostically Celtic or can be safely translated as 'curved'.

However, Camb- may not be diagnostically Germanic either. Camboglanna fort on Hadrian's Wall lies in a very Celtic area, yet “Extensive accumulations of sands and gravels, formed in part by glacial meltwaters, form highly distinctive hummocky country, notably around Brampton” (UNESCO, 2008). Similar comments can be read about Morecambe (in Lancashire, Ptolemy's Μορικαμβε), Kempten (in Bavaria, Strabo's Cambodunum), and De Kempen (in Belgium).

Some Personal Names

Ancient personal names from south-eastern Britannia also seem more Germanic than Celtic, though the argument cannot yet be taken as far as with place names. Boudicca is discussed at length by Durham and Goormachtigh (2011), while south-eastern Britain's last pre-Roman king and his three sons had names that can now be analysed as Latin endings on Germanic cores.

Cunobelinus (Shakespeare's Cymbeline) lived until about AD 40 and (appropriately for a contemporary of Jesus) had a name that could mean ' son of the god Belenus '. The first part, cuno- is usually described as Celtic for hound, making it analogous with the wolf in many Germanic names. An alternative interpretation is PG *kuniz , OE cyning , ME king, related to kin and to Latin genus , and liable to be confused with several other roots that led to ME cunning, keen, ken, can, con, etc. Plenty of related Germanic names show up in the earliest records, from about AD 600, such as the Gepid leader Cunimundus or Anglo-Saxon Cynegils, Cynefrid, Cynerburga and Cynwise.

Adminius (or Amminius) may mean something like ' great '. His name is potentially confusable with multiple Roman names, including the historian Ammianus and Ammius Flaccus, mentioned by Pliny, but the most likely parallel is his contemporary Arminius, who defeated the Romans in AD 9 at the Teutoberg Forest. Arminius is generally translated as modern Herman, from PG *erminaz 'strong', which also gave OE eormen 'great, universal'. Related names include Eormenric and Eormenred in Kent around AD 550, Ermanaric king of the Goths around AD 370, plus modern Ermintrude and Emma.

Caratacus probably meant ' bold counsel '. His name is usually likened to later Welsh Caradog, with a first syllable meaning 'love', but there are better parallels in modern Kurt and Conrad, or in the Anglo-Saxon kings called Cenred (Wessex, about AD 600, Mercia, about AD 700). The first element, PG *koniz 'bold' survives in ME keen, while its second element is like OE ræd 'counsel' perhaps best known from King Rædwald (of Sutton Hoo) or modern German Rathaus 'town hall'. The ending  acus is sometimes claimed to be distinctively Celtic but in fact also exists in Latin, Greek and other languages.

Togodumnus is essentially the same word as modern 'dukedom' but can be translated better as ' leader of the nation ', the type of name favoured by many 20th-century dictators. Its tog part is in OE togian 'tow', distantly related to tug, duke and educate. Its second part dum is like OE and ME –dom 'jurisdiction (etc)', used as a suffix in modern words like kingdom and freedom, or German Eigentum 'property', and which also survives in ME 'doom'.

Some Tribal Names

Much effort has been expended finding etymologies for ancient tribal names (De Bernardo Stempel, 2008). Some proposed Celtic roots are highly debatable, so it seems only fair to draw attention to a possible OE tragan 'drag/draw' in the Durotrages , who controlled the ancient trade routes, trans-isthmus Avon to Avon, and across the sea to the Durocasses in Normandy.

Trinovantes is pure Latin for 'three renewings'. Rivet and Smith (1979) realised that the Trinovantes possessed three of something that the Novantae tribe in Dumfries and Galloway had only one of, but no one seems to have pointed out the obvious answer – three river estuaries with substantial ports. In the north, the rivers Orwell and Stour merge to reach the sea past the Roman fort at Felixstowe. In the centre, the Colne past Colchester (ancient Camulodunum) and the Chelmer past Chelmsford (Caesaromagnus) merge into one big gulf. In the south, lies the river Crouch, with plentiful prehistoric remains. 'Renewing' obviously fits a river, and RC refers to Novitia , which is generally taken to mean the modern river Nith, but it also refers to Anderelio Nuba and to Novia , which are taken, like Ptolemy's mention of Νοουίου ποταμοΰ εχβολαί as referring to the mouth of the Sussex river Ouse near Pevensey. So it seems reasonably certain that Trinovantes is Latin for ' three estuaries '.

Some River Names

River names can be older than place names, so they are often used to illuminate ethnic origins. However, much published logic needs to be questioned because early linguistic guesses conflict with modern topographical or archaeological information. For example, we discuss the rivers Cray (in Kent and elsewhere) at length in a forthcoming paper.

The common name Avon is routinely described as derived from Celtic abona 'river', but all the English Avons are in positions to be trans-isthmus trade routes, as defined by Sherratt (1996), important in the Neolithic and Bronze Ages, but not later. It is therefore arguable that avon meant a particular type of river, based on PIE *ab- 'water, river' and, because words most often evolve semantically from particular to general, it was a loanword into Celtic, not from Celtic.

Latin amnis 'river' is related to avon but also obviously to the river Eems/Ems (Latin Amisius ), around whose mouth the Anglo-Saxons, and hypothetically the Belgae, originally lived. Thames (Latin Tamesis ) has been a hard name to explain, but it looks very like the Ems plus an initial T transferred from a previous word, as often happed to mediaeval names with OE æt 'at the'. Such a process has not hitherto been accepted as happening in Roman times, but it is implied by the discussions of Dover and Thanet above, and could explain other puzzling river names such as Severn ( Sabrina ) and Humber.

King's Lynn is widely described as having a name like modern Welsh llyn 'pool'. In fact it almost certainly came from Anglo-Saxon hlinn 'torrent'. (OE initial HL regularly evolved into ME plain L.) Nowadays the Great Ouse is generally very placid, but the configuration of land, rivers, and sea around the Wash has changed greatly since Lynn was founded in about AD 1200 (Clarke, 1973). The phrase “torrential rain” trips off the tongue in English, and according to the Environment Agency: “The catchment has a history of flooding, generally due to high rainfall which has led to watercourses and drains being overwhelmed”. In the other direction, coming from the sea, King's Lynn harbour has some of the strongest tidal currents known to yachtsmen, while wind and tide can combine to create a tidal bore even in the much-modified modern river system. The great east coast floods of 1953 had a precursor in 1216 when, in the words of Charles Dickens, King John lost his treasure in the Wash as he “saw the roaring water sweep down in a torrent .”

Beult was suggested as possibly Celtic by Breeze (2010). In fact, the river Beult is notorious for flooding, because its catchment area, the Kentish Weald, acts like a giant clay-lined pond, with only a limited exit through the North Downs. PIE *b h el- 'swell' led to modern German es beult 'it bulges' and many other descendants. So Beult can be confidently translated as something like ' overflower '.

Brenchley is like a local opposite of Beult, lying on higher ground where several springs act as sources of streams, so one may suspect a link with PG *brunnon 'spring, fountain'. Breeze (1998) suggested that it might contain a Cornish personal name, which is a reminder that ancient Britain was probably multi-ethnic enough for minorities to give rise to some place names.

General Discussion

The central conclusion of this paper is that almost 100% of place names in Britain south-east of a line from the Wash to the Solent can be explained better with the aid of English than Welsh. This fact holds true for names that can be traced back before the traditional date of first arrival of Anglo-Saxons. It may also hold true for ancient personal names in that area.

Four possible explanations seem worth considering. First, maybe this whole paper is mistaken and the classic all-Celtic Britannia idea need not be declared dead. Many historical linguists doggedly defend the idea that Germanic languages did not much differentiate until after AD 400. Therefore (they argue) early Germanic-seeming names must all be blamed on linguistically naive investigators being fooled by chance coincidences and the differential survival of words in different language branches.

In fact, all writing about place names needs to come with a "hazard warning" that it contains mistakes. There is only a limited amount of data available, with no easy way to create more by experiments, so all scholars are forced to speculate and sometimes to be wrong. The fact that our papers sometimes contradict Eilert Ekwall, Kenneth Jackson, and Margaret Gelling in no way lessens our admiration for their brilliance. They just had less information available than nowadays from archaeology and the Internet.

The second possible explanation is that ancient place names were differentially created by a Germanic minority , such as Roman troops and administrators, traders from across the North Sea, or laeti or foederati settled in Britain by the empire. This explanation, like the first, is hard to reconcile with the sheer thoroughness with which Celtic names are absent from the south-east. Why would any minority, no matter how militarily or economically dominant, bother to change the name of every tiny village or hillfort that had been occupied for centuries if not millennia?

The third possibility is that there was a whole band of languages that died out between Latin and Insular Celtic or Germanic. There is some epigraphic evidence for Lusitanian (Prosper Perez, 2008), Illyrian, Ligurian, and Raetian, and one might wish to add hypothetical Belgic, Icenic, and proto-Kentish dialects to that list. In effect, this would half-accept the viewpoint of Schrijver (2007) that some sort of common dialects prevailed around the North Sea, except that he sees them as closer to Celtic than to Germanic.

The fourth possibility, the one we favour, is that all across south-eastern Britannia local peasants spoke a Germanic language (“proto-English”) that would have been more readily understood across the North Sea in Frisia or Flanders than in Wales or Ireland. In effect, this would accept the viewpoint of Alinei (2008) that Indo-European languages in general, and Germanic dialects in particular (Ballester, 2005), diversified many centuries earlier than textbooks now state. If correct, this means that much historical writing about Roman Britannia needs to be revised.


Alinei , M (2008) A Teoria da Continuidade Paleolítica das Origens Indo-Europeias: Uma Introdução, Apenas Livros, Lisbon. ISBN 9789896181925.

Arias, G (1987) Grammar in the Antonine Itinerary: a challenge to British archaeologists. ( may be consulted in Cambridge University Library)

Ballester, X (2005) The First Germanic Origin of the English Language, Quaderni di Semantica 26 29-41

Breeze, A (2003) Not Durotriges but Durotrages, Notes & Queries for Somerset and Dorset 35 , 213-215.

Breeze, A (2006) Ptolemy's Gariennus, Burgh Castle, and the Yare. Proc. Suffolk Inst. Arch. History 41 (2), 227-230.

Breeze, A (1998) Brenchley, Kent. Nomina 21 , 154-156.

Breeze, A (2010) The Celts and the river Beult. Arch. Cant. 130 , 385-387.

Clarke, H (1973) The changing riverline of King's Lynn, Norfolk, in the Middle Ages. Int. J. Nautical Arch. 2 , 95-106.

Coates, R (2000a) Thanet and its Alternative Name in the Historia Britonum. pp 32-39 in Coates, Breeze & Horovitz (2000).

Coates, R (2000b) Othona and Ythancaestir: Reconciling the Evidence. pp 167-171 in Coates, Breeze & Horovitz (2000).

Coates, T (2002) The significance of Celtic place-names in England. pp 47-85 in Filppula, M, Klemola, J & Pitkänen, H, eds. The Celtic Roots of En glish. Joensuu, ISBN 9524581647.

Coates, R (1991) The place-names of Hayling Island, Hampshire. Manuscript, revised version accessible electronically at /

Coates, R (2005) Verulamium, the Romano-British name of St Albans. Studia Celtica 39 , 169-176.

Coates, R, Breeze, A & Horovitz, D (2000) Celtic voices English places. Shaun Tyas, Stamford, ISBN 1900289415.

Cole, A (1990) The origin, distribution and use of the place-name element ora and its relationship to the element ofer . EPNS J. 23 , 26-48.

De Bernardo Stempel, P (2008 ) Linguistically Celtic ethnonyms, towards a classification. in Celtic and other languages in Ancient Europe . pp 101- 118 in García Alonso (2008).

Durham, A & Goormachtigh, M (2011) Boudicca does not mean Victoria. Submitted for publication.

Garcia Alonso, JL Ed. (2008) Celtic and Other Languages in Ancient Europe. Ediciones Universidad de Salamanca, ISBN 9788478003358.

Gelling, M and Cole A (2003) The Landscape of Place-Names, Shaun Tyas, Spalding, ISBN 1900289261.

Goormachtigh, M & Durham A, 2009. Kentish place names – were they ever Celtic? Arch. Cantiana 129 , 279-293.

Grundy, GB (1922) On the meanings of certain terms in the Anglo-Saxon charters. Essays & studies by members of the English Association, 8, 37-59.

Jackson, K (1948) On some Romano-British place-names. J. Roman Studies 38 , 54-58.

Jackson, KH (1953) Language and history in early Britain: a chronological survey of the Brittonic languages first to twelfth centuries AD . Edinburgh University Press.

Jessup, RF (1932) Romano-British Kent – Roads. Victoria County History of Kent, 3 , 134-142.

Laker, S (2002) An explanation for the changes kw-, hw- > χw- in the English dialects. pp 182-198 in Filppula, M, Klemola, J & Pitkänen, H, eds. The Celtic Roots of English. Joensuu, ISBN 9524581647

Mills, A D (2003) Dictionary of English Place Names. Oxford University Press, ISBN 0198527586.

Nicolaisen, WFH (2001) Scottish Place-Names. John Donald, Edinburgh. ISBN 0859765563.

Oppenheimer, A (2006) The origins of the Britis h. Constable & Robinson. ISBN 9781845294823.

Orel, V (2003) A handbook of Germanic etymology, Brill, Leiden, ISBN 9004128751.

Pearson, AE (2002) The Roman Shore forts: coastal defences of southern Britain. Tempus Books . ISBN 0752419498.

Philippa, M , Debrabandere F, & Quak, A (2007) Etymologisch Woorden van het Nederlands (Ke-R) Amsterdam University Press , ISBN 9789053567470

Pope, CR (1989) The historic flora of Ryde Dover. Proc. Isle of Wight Nat. Hist. Archaeol. Soc. 9 , 33-40.

Prosper Perez, B (2008) Lusitanian, a non-Celtic Indo-European language of western Hispania. pp 53-64 in Garcia Alonso (2008)

Pryor, F (2004) Britain AD. Harper Collins. ISBN 0007181876.

Pokorny, J (1959) Indogermanisches Etymologisches Woerterbuch. Francke, Bern. See for a revised edition.

Pujol, LP (2008) The annona militaris in the Tingitana: Observations on the organization and provisioning of Roman troops, British Archaeological Reports International Series 1782, ISBN 9781407302706, .

Rivet , ALF & Smith, C (1979) The Place-Names of Roman Britain . Batsford, London. 1979. Sections relevant to the Saxon Shore are reproduced online by .

Schrijver, PCH (2007) What Britons spoke around 400 AD. pp 165-171 in Highham, NJ, Ed., Britons in Anglo-Saxon England. The Boydell Press, Woodbridge, ISBN 9781843833123

Seeck, O (1876) Notitia Dignitatum: accedunt Notitia urbis Constantinopolitanae et laterculi provinciarum. Weidmann, Berlin.

Sherratt, A (1996) Why Wessex? The Avon route and river transport in later British prehistory. Oxford J. Arch. 15, 211-234.

Sims-Williams, P (2006) Ancient Celtic Place-Names in Europe and Asia Minor. Blackwell , Oxford. ISBN 9781405145701.

UNESCO (2008) Frontiers of the Roman Empire World Heritage Site, Hadrian's Wall Management Plan.'s Wall Management Plan 2008-2014 - Appendices.pdf

Vennemann, T (2006) The name of the Isle of Thanet, pp 345-374 in Language and text: Current perspectives on English and Germanic historical linguistics and philology, Eds Johnston AJ, von Mengden, F & Thim, S. Carl Winter, Heidelberg, ISBN 9783825351786.

Voitl, H (1988) The history of voicing of initial fricatives in Southern England: a case of conflict between regional and social dialect. pp 565-600 in Historical Dialectology, Regional and Social, Ed. Fisiak, J, Linguistics Studies and Monographs 37 , Mouton de Gruyter, ISBN 3110115506.

Williams , J (2007) New light on Latin in pre-conquest Britain. Britannia 38 , 1-12.