Introduction

1     In a sense the study of Roman place-names in Britain got off on the wrong foot. This was mainly because the Antonine Itinerary was so much more accessible than the other main sources, the Geography of Ptolemy, the Ravenna Cosmography and the Notitia Dignitatum. The Antonine Itinerary provided fifteen itineraries within Britain, each listing a number of place-names and the distance between each place-name and the place-name previous to it in that particular list. And because some of the place-names could be identified with  certainty, it was in most cases relatively easy to trace a particular itinerary along known Roman roads and, using the distances given in that itinerary, identify the other place-names in that particular list. And because the Antonine Itinerary was more accessible it was discussed and written about more often than the other sources and so people became familiar with the forms of the names given in it. Those forms came to be regarded more often than not as the correct forms of the names and other forms, given in other sources, tended to be regarded as erroneous. This is a great pity, because the Antonine Itinerary is believed to have been produced no earlier than AD200 and perhaps even 50 years later, so at best some 150 years after the Roman invasion of Britain, and that is a sufficiently long time for many names to be altered, to be latinised, to make them sound better, sound more right, to people who spoke Latin. The Geography of Ptolemy is believed to have been produced around AD140, some 50 years at least earlier than the Antonine Itinerary, and so where the forms of a particular name in the Antonine Itinerary and the Geography of Ptolemy differ, the form in the Geography of Ptolemy is inherently likely to be closer to the original Celtic name. And whereas the Ravenna Cosmography is believed to have been produced around AD700, the maps used by the compiler of the Cosmography appear to date mostly from the first century. Indeed some of the Ravenna names, for example those in central southern England, and possibly also those in southwest England, appear to have been taken from early Claudian maps. It follows that many of the names in Ravenna are around 100 years earlier, others some 150 years earlier, than those in the Antonine Itinerary, so where the same name appears in both the Ravenna Cosmography and in the Antonine Itinerary, but the form of the name differs in the two sources, then the form in the Ravenna Cosmography is inherently likely to be closer to the original Celtic name. There has been considerable debate as to the date of the Notitia Dignitatum, but the group of ND names relating to forts along the line of Hadrian’s Wall is at the earliest Hadrianic, since it includes Pons Aelius at Newcastle and Aelius was Hadrian’s family name (in the interests of brevity the Notitia Dignitatum will sometimes be referred to as ND later in this study, and the Antonine Itinerary as AI).

2     One can see the sort of error that has occurred in the past by considering the name which is Gabaglanda in Ravenna, Camboglans on the Rudge cup, Cambog…s on the Amiens patera and Amboglanna in the Notitia Dignitatum. The Ravenna name shows the fairly common omission of m during medieval copying and the confusion of c and g – the name was presumably originally Cambaglanda.  And the ND name has lost its initial C – this name will have been Camboglanna. The Rudge cup and Amiens patera are in fact souvenirs of the Trajanic and not the Hadrianic frontier (this is explained in Chapter 8), and so the form Camboglans is earlier than Camboglanna. But Cambaglanda appears in Ravenna within a group of Flavian fort-names (the others are the forts at Plumpton Wall, Carlisle, Carvoran and Chesterholm), so one can be fairly certain that Cambaglanda was also a Flavian fort. In other words the form Cambaglanda is the earliest we have. This had developed to Camboglans by the time the Rudge cup was produced and to Camboglanna by the time the Notitia Dignitatum was compiled. If one really wishes to work out the meaning of a place-name one should work with the earliest form available, in this case Cambaglanda, since it is the form which is inherently likely to be closest to the original Celtic name, but scholars have nonetheless chosen the latest form available, Camboglanna, as being ‘correct’, and so the understanding of this name most certainly got off on the wrong foot. But once scholars have selected the ‘correct’ form of a name they then seek to identify its component elements, and it will not always be clear what these are. But in the case of Camboglanna the o was seen as a linking vowel and the two elements were thus identified as camb and glanna. There can be no guarantee that this is the correct subdivision of the name, no matter how widely it is accepted as being correct, and in this example it is clear that glanna was most certainly not an element of the original Celtic name. But once scholars have identified what, for them, are the component elements of the name, they then look around for any similar words in a modern Celtic language and, if they find any, they then try to extrapolate backwards in time to invent hypothetical British roots which might have produced those name-elements. There are thus three steps in the process: (1) select the ‘correct’ form of a name, (2) decide what its elements are, and (3) look for similar modern Celtic words and then invent hypothetical roots which might possibly have led to those elements. It is of course possible to make an error at any or all of the three steps, and in the case of Cambaglanda an error was made in the very first step, so that the generally accepted meaning of the name clearly cannot be right. But the camb of Camboglanna is taken by scholars to be derived from a hypothetical British root cambo thought to mean ‘curved, crooked’, and because of the chance resemblance between glanna and modern Welsh glann meaning ‘bank, shore’, the glanna of Camboglanna is thought to be derived from a hypothetical British root glanno with the same meaning. The name Camboglanna is thus taken to mean ‘curved bank’ or ‘bank at the bend’. But of course since glanna was not an element of the original name, one can be quite sure that the original name had nothing to do with any hypothetical root glanno. And that is only one example - there are others one could mention, for example the name which is Glannibanta in the Notitia Dignitatum, Clanoventa in the Antonine Itinerary and Cantiventi in Ravenna. The Ravenna form is again the earliest available, but scholars accept a hybrid of the AI and ND forms – Glannoventa - as being ‘correct’, and they see its elements as being glanno and venta. And because of the unfounded assumption that Glannoventa was the fort at Ravenglass, glanno, which was taken to mean ‘bank’ in the case of Camboglanna, is now assumed to mean ‘shore’, and because venta has simply been assumed to mean ‘market’ or ‘field’, the name Glannoventa has been taken to mean ‘shore field’ or ‘shore market’. But of course if Glannoventa was not at Ravenglass (as indeed it was not) or anywhere else on the coast, and if venta really has nothing to do with a field or market (as indeed it does not), then the generally accepted meaning of Glannoventa cannot be correct. Two other examples spring to mind. Firstly, because of the chance resemblance between the lact of Lactodoro (Towcester, in the Antonine Itinerary) and the Latin word for ‘milk’, Roman Towcester is assumed to have been a walled town of dairymen. But no-one has ever explained satisfactorily why the dairymen of Roman Towcester should have required a wall round their town or, if dairymen in general required walls round their towns, why there are no other places called Lactodoro elsewhere in Roman Britain – surely Roman Towcester didn’t have a monopoly on the supply of dairy products! Secondly, because the mandu of Manduesedo (Mancetter, this name also appearing in the AI) happens to look like a hypothetical British root mandu taken to mean ‘small horse, pony’ and the esedo element looks like a hypothetical root essedo taken to mean ‘war-chariot’ or ‘cart’, Manduesedo is thought to mean ‘horse-chariot’, though no scholar has ever explained satisfactorily why there should have been such a significant number of chariots at Mancetter that the place was given a name which actually meant ‘horse-chariot’. One could quote other examples, but there should be no need. The above four examples should suffice to show that the long-established technique for explaining Roman place-names in Britain is wholly inadequate. It is simply guesswork, guesswork dressed up in an academic gown. It is hopelessly unscientific and has held up Roman place-name studies for decades.

3     But it is possible to adopt a scientific approach. If we are right in believing that the Celts coined place-names based on topographical features of the landscape, then the information as to those features will be concealed within the names themselves, and it ought to be possible to identify that information by analysing the names in conjunction with topographical maps. This analysis can be successful if we have enough data to work with, i.e. enough names with locations which are certainly correct and if we have the original forms of the Celtic names or, failing that, if we use the earliest available forms of those names, so that we can be reasonably certain that those forms are the same as, or close to, the original Celtic names. In this way we can find out what elements the Celts did actually use when coining topographical names, and this information is infinitely more important and relevant than anything  any academic might say as to what elements the Celts could have or might have used two thousand years ago.

4     This study adopts the scientific approach outlined in the previous paragraph and it works mainly with the place-names in Ravenna since these are in general the earliest forms known to us, no matter how strange some of the names may appear to those of us who are more familiar with the latinised names in the Antonine Itinerary and accept these as being correct. The Ravenna names used as a basis for the analysis were:

Bannio (Abergavenny), Bereda (Plumpton Wall), Bremenium (High Rochester), Bresnetenaci (Ribchester), Brocoliti (Carrawburgh), Cactabactonion (Catterick), Caleba (Silchester), Canubio (Caerhun), Celunno (Chesters on the North Tyne), Cesaromago (Chelmsford), Cironium (Cirencester), Condecor (Benwell), Cunetzione (Mildenhall), Deva (Chester), Dubris (Dover), Duroaverno (Canterbury), Durobrabis (Rochester), Isca (Exeter and Caerleon), Durobrisin (Water Newton), Eburacum (York), Esica (Great Chesters), Glebon (Gloucester), Lagentium (Castleford), Lagubalium (Carlisle), Lavaris (Bowes), Lemanis (Lympne), Lincovigla (Lanchester), Lindum (Lincoln), Londinium (London), Magnis (Kenchester), Manulodulo (Colchester), Onno (Halton Chesters), Rate (Leicester), Rutupis (Richborough), Seguntio (Caernarfon), Serduno (Wallsend), Velunia (Carriden), Velurcion (Housesteads), Venta (Winchester, Caerwent and Caistor St. Edmund), Vindolande (Chesterholm), Vindovala (Rudchester), Vinovia (Binchester), Viriconion (Wroxeter) and Virolanium (St. Albans).

Some Antonine Itinerary names were also taken into account, these being names which are generally considered to have been identified with certainty and which have the same ‘feel’ as the Ravenna names listed above, so that they might be assumed to date from the early period of the Roman occupation. The AI names are:

Bannaventa (Whilton Lodge), Brocavo (Brougham), Brovonacis (Kirkby Thore), Burrio (Usk), Crococalana (Brough, Lincolnshire), Dano (Doncaster), Durocobrivis (Dunstable), Isurium (Aldborough), Manduesedo (Mancetter), Margiduno (Castle Hill, Nottinghamshire), Pennocrucio (on Watling Street, near Penkridge), Uxacona (Red Hill), Venonis (High Cross), Vernemeto (Willoughby), Verteris (Brough under Stainmore) and Vindomora (Ebchester).

The above Romano-British names were analysed in conjunction with the topography of the landscape at the known locations of the places concerned, in order to identify, isolate, within the names, those letters or letter combinations which contained topographical information. The results of that analysis are presented in Chapter 1 as the building-blocks of Romano-British place-names. Note that those building-blocks are not in any sense hypothetical. Rather, because of the manner in which they were obtained, they are the building-blocks actually used by the Celts when coining place-names. Those building-blocks, in association with the elements identifying rivers - also introduced in Chapter 1 - form a powerful tool which enables us to identify most of the place-names and river-names in the Geography of Ptolemy and the Ravenna Cosmography. They also permit an explanation of some groups of names which have hitherto defied satisfactory explanation, for example the isca-type names (Chapter 3), names with an essa-type ending (Chapter 4) and the river-name Alavna (Chapter 5). They also permit one to see that the Ravenna names Serduno to Maia Fanococidi predate Hadrian’s Wall and do in fact define the Trajanic frontier between the Tyne and Solway (Chapter 8). And that information leads logically to the conclusion that the inscription on the Ilam pan is not authentic (Chapter 9). The study continues with various chapters dealing with the Roman place-names in different regions of Britain and Chapter 19 provides a detailed treatment of the river-names in Ptolemy and Ravenna.

[This page was last modified on 23 March 2021]