The ‘Cosmography’ written by an unknown author in Ravenna during the early eighth century has never been fully explored as a source of Roman toponymy. This is especially true in Britain, which is less well-served than many other parts of the empire by Roman-period geographers. Full of corruptions, with little evident logic in its ordering of names, the Cosmography has always been regarded as less useful than Ptolemy’s Geography, the Antonine Itinerary or the Peutinger Table in providing the names of places, rivers and islands for Britain in the Roman period. However, of all the ancient geographical sources relating to Britain, it contains more names than any other and this fact alone should encourage us to examine it in greater detail.

Previous studies of the text

The text was first published by Porcheron in 1688 from a manuscript in the Vatican Library, that now known as Urb. Lat. 961 (Schnetz 1940 A; Richmond & Crawford 1949 V). Its first use as a source of Romano-British placenames seems to have been by the antiquary Gale (1709, as quoted in Horsley 1732, 490). However, neither he nor his eighteenth- and nineteenth-century successors seem to have realised that there were more than the vaguest of ordering principles in the list of names it gives. The resemblance of a modern name to one recorded in the ancient sources was generally considered sufficient proof of identity. By the middle of the nineteenth century, though, it had been recognised that the antiquarian method of assigning Roman placenames was unscientific (Giles 1906, 474) and a more historical approach to the linguistics was developed (Hooppell 1877, 290). Furthermore, as archaeology became more systematic in its ability to identify places of Romano-British habitation and Roman military remains, so it became accepted that evidence for Roman period occupation in a place was needed before identifications could be accepted. Increasing use was made of the Antonine Itinerary and the Itinerary of Richard of Cirencester which, by giving the distances between named places, allowed more reasonable guesses to be made. After the exposure of Richard of Cirencester’s Itinerary (or ‘Diaphragmata’ as it is called in the text) as an eighteenth-century hoax (Mayor 1869, xvii), the Antonine Itinerary became the single most important source of Romano-British toponymy. This was followed by Ptolemy’s Geography which, from the late nineteenth century onwards, was explored in scientific detail (Bradley 1885; Rylands 1893). However, the impenetrable nature of the Ravenna Cosmography made it less favoured as a source of names in its own right until the twentieth century.

During the twentieth century, three major attempts were made to unravel the tangle in the chapter of the Cosmography that deals with Britain (V.31); a pioneering attempt by Sir Ian Richmond and O G S Crawford, published in 1949, with a contribution from Sir Ifor Williams on the Celtic philological aspects of the work (Richmond & Crawford 1949); an important reassessment by Louis Dillemann published in 1979, for the first time presenting in English the theories of J Schnetz, the most recent editor and commentator on the work as a whole (Dillemann 1979); and a systematic analysis by A L F Rivet and Colin Smith in their monograph, The Place-Names of Roman Britain, also published in 1979, with philological advice from Professor D Ellis Evans (Rivet & Smith 1979).

The first of these studies took the Cosmography at face value, trusting many of the bizarre names for which Sir Ifor Williams struggled valiantly to find good Celtic derivations. For instance, the grotesque form Corsula (Cosmography 1099), which is almost without doubt an error for the Latin insula, was analysed by Williams as possibly containing an element *cors-, meaning ‘reed’ (Richmond & Crawford 1949, 30). There are dozens of similar examples in the study, reflecting the authors’ lack of access to the contemporary German work on the Cosmography, which was highlighting for the first time these textual propblems,0 owing to the troubled times in which they were writing. However, their attempt broke new ground in our understanding of the British section, particularly regarding the methods of the Cosmographer, whom they showed to have been using an itinerary-like map in a vaguely ordered way. Their careful study of each name has been the model followed by all subsequent detailed analyses of Romano-British toponymy (eg Rivet 1970; Rivet & Smith 1979), and the contribution of the leading Celticist of his day was of inestimable value in correcting many of the errors in the text.

Dillemann (1979) examined the Cosmography as a whole, taking the British section as a usefully discrete geographical unit with which to illustrate the general problems of the source. He also offered a detailed critique of some of the identifications and etymologies proposed by Richmond & Crawford, although he occasionally fell into serious error, presumably through a relative unfamiliarity with Roman Britain (for instance his confident assertion (1979, 69) that the province of Valentia lay to the north of Hadrian’s Wall). Nevertheless, by introducing British archaeologists and toponymists to the analyses of Schnetz, which had previously been overlooked a they were published in Germany at the height of the Second World War, Dillemann demonstrated that a way through the problems presented by the text was at hand.

Working at the same time as Dillemann, Rivet & Smith (1979, 185-215) reached broadly similar conclusions about the reliability of the British section. The study was initially the work of Colin Smith (Rivet & Smith 1979, xlii); perhaps as a result of this the discussion of the linguistic background to the Cosmography is admirable, but the treatment of the names as a toponymic source is, in too many instances, summary. Their greater familiarity with Britain enabled them to avoid some of the traps into which Dillemann had fallen. For instance, whereas Dillemann (1979, 65) had suggested that the Auentio of the Cosmography (10828) is an error for the Λουεντινον of Ptolemy (Geography II.3,12), it is a perfectly acceptable Celtic form and the antecedent of the Modern Welsh Afon Ewenni. Rivet and Smith (1979, 260) rightly dismiss Dillemann’s argument.

However, there are clear indications that Rivet and Smith went too far in purging the Cosmography of cacographies and dittographies (Frere 1980; Jones & Mattingly 1990, 33). Firstly, they attempted to emend perfectly reasonable Celtic forms, such as Volitanio (10752) and Clauinio (10612), apparently to fit their theories regarding the compilation of the British section. They were also too eager to see duplications where it is not necessary to do so. For instance, where river-names seem to have been written in an inland position on the Cosmographer’s putative map source, such as *Tamesa, the Thames (10638), they do not recur in the river list; it is thus likely that where a supposed river-name occurs in both the list of the ciuitates and in the list of rivers, its first occurrence does not in fact refer to the river but to a settlement on it. Thus Tamaris (10548) and Tamaris (10826) are not both the River Tamar, and the former is probably a settlement or fort on the river and the place named Ταμαρη by Ptolemy (Geography II.3,3), probably near Launceston, where the only known road into Cornwall crosses the river (contra Strang 1997, 30). The conclusions of Rivet and Smith require considerable modification, and we should be especially wary of altering the Cosmography’s forms where they appear to be genuinely Celtic names. Unfortunately, I have been forced to write almost as if it is a detailed critique of Rivet and Smith, which was not my original intention; however, the sheer weight of their authority renders point-by-point discussion of each name necessary if we are to approach the text afresh.

The Cosmographer’s method

There is general agreement that the Cosmography is the most difficult of all the sources of Romano-British placenames to use (Jackson 1953, 33). This is largely a result of the corruption of the text, a corruption which in most instances undoubtedly goes back to the author himself. He clearly did not understand his sources properly, even to the extent of not appreciating their purpose, and misspells names that should have been familiar to him (Dillemann 1979, 61). He may also have been working with poor originals (Rivet & Smith 1979, 199), which could explain duplications of a single name in variant spellings as he struggled to make out the writing in his source, occasionally putting down two attempts. It is little wonder that when we turn to the section dealing with the remote (and no longer Roman) island of Britain we find an exceptionally large number of very curious-looking names and a lack of others that we might expect to find.

That he was working largely without a recognisable system is clear from his opening words to the chapter dealing with Britain: ‘In that Britain we read that there were many civitates and forts, of which we wish to name a few’. This sentence also makes it clear that he was not attempting to name all the civitates and forts but making a selection, an important consideration when we come to compare the Cosmography with other sources, especially Ptolemy. It is interesting that he is aware that these civitates and forts no longer exist, as he uses the perfect infinitive, although later in the text he consistently employs the present indicative. It is also obvious that he was working from at least one map, since the text is full of visual impressions such as ‘iuxta’ (‘next to’) or ‘ubi et ipsa britania plus angustissima de oceano in oceanum esse dinoscitur’ (‘where that same Britain is seen to be narrowest from Ocean to Ocean’). However, the ordering of names is not entirely haphazard, and it is often possible to distinguish groups that lie either along a single road or clustered around a central point, as Richmond & Crawford (1949, 5) were the first to recognise. There are, moreover, significant duplications which suggest that the Cosmographer’s eye returned periodically to especially prominent nodal points. The most startling example of this is the four-fold repetition of Moridunum, perhaps Sidford, which appears no less than four times (as <Melamoni>, <Milidunum>, Moriduno and <Morionio> at 1062, 1064, 1069 and 10613 respectively).

Sometimes the Cosmographer seems to have been working along the line of one road but takes in the names of places situated on others nearby, evidently because there were written in such a way as to suggest that they referred to points on the road which interested him. A good example of this is his progression from Durouiguto (10653, probably Godmanchester) to Lindum Colonia (10655, Lincoln) which also takes in <Ventacenomum> (10654, Venta Icenorum, Caistor St Edmund), over one hundred kilometres east of the road being followed. In this instance we might suggest that the name was written to the west of the symbol to which it referred. As all the demonstrable instances of his misreading names in this way are of places farther east than the line being followed, it may be suggested that his map source usually placed names to the east of their symbols; where it placed the name to the west, the Cosmographer was misled into assigning it to the wrong road. Elsewhere he seems to work in a rough circle, as in the progression Nauione, Aquis arnemeza, Zerdotalia, Mantio, Alicuna, Camulodono (10656 to 10659, Brough-on-Noe, Buxton, Melandra Castle, Manchester, ?Elslack, Slack).

Rivet and Smith (1979, 204) criticised the presentation of the text by Richmond and Crawford, with its division into neat sections, as they give no indication of the essential continuity of the listing. After due consideration, I have also decided to divide the text into short sections in the main part of this study to break up the discussion into manageable units, although it is also (and first) presented in full elsewhere. It must be emphasised that the breaks adopted here are not indicated in the text itself, but are based upon what appear to be the major jumps made by the Cosmographer in his ordering of names, and it will be shown that there is a vague logic to this order. Although he was not strictly following the road system, he does seem to have worked in lines, perhaps following his finger or a pointer across his map source, occasionally jumping back to his point of departure. This is particularly clear in the first few sections, where he states that he has moved to a place next to one already listed (‘Iterum iuxta super scriptam ciuitatem <scadonamorum> est ciuitas ...’ 1066-1068).

That the Cosmographer did have rather more method than Rivet and Smith allow him is amply demonstrated by plotting the order in which he listed places, as defined by Rivet and Smith (see, for instance, Jones & Mattingly 1990, map 2.4). Where comparison with the Antonine Itinerary and Notitia Dignitatum have enabled firm identifications of names with known sites to be made, the Cosmographer can be seen to be working in a series of zigzags across the island; unsystematic, to be sure, but without making wild leaps, crossing or passing through areas already covered. When we turn to the otherwise poorly-documented south-west and Scotland, however, the identifications proposed by Rivet and Smith appear to imply that the Cosmographer was working entirely at random, which is surely wrong. It is to attempt to discover what logic the Cosmographer was using in these more obscure areas that the present study was initially undertaken.

As a general principle in the identification of names, it is assumed that where they can be assigned to sites of Roman date with reasonable certainty, the progression of unknown names between them follows a generally logical order unless there is good evidence for a jump to another area. In particular, no emendations of names will be attempted which point in an otherwise illogical direction. For instance, whereas Rivet and Smith (1979, 345) follow Horsley (1732, 490) and emend Duriarno (10552) to Durnouaria, Dorchester (Dorset) (on the grounds that while the Cosmographer appears generally to have been at pains to list Civitas capitals, this one is not otherwise listed in the Cosmography) I have preferred to accept the form (which is a perfectly good British Celtic form) and believe that the Cosmographer merely overlooked Dorchester. In this way the general logic of the progression of this part of the list westwards from Exeter can be preserved.

There are, however, obvious jumps in the text; these are never uncontrolled and the general system employed seems to have been a series of broad sweeps across the island, starting in the south-west and progressing to the north in lines roughly parallel with Watling Street. This is a more trusting approach to the text than Rivet & Smith’s, but it is one which makes the British section of the Cosmography less mysterious and impenetrable and, I believe, more accessible to interpretation.

The text