The ancient name of the town of Cirencester, Gloucestershire, has caused considerable problems of interpretation. It is recorded in the second century ad in Ptolemy's Geography (ii.3.13) as Korínion (variants Korínnion, Korónion).Footnote ¹Korínion, taken as authentic, has been generally understood to represent a Romano-British *Corinium.Footnote ² The most recent commentators, A L F Rivet and Colin Smith, do not state in which of the many surviving manuscripts of Ptolemy the variants occur. I have not been able to locate a variorum edition of the Geography, and I cannot therefore form an independent view on their relative authority. Ptolemy expert Dr Florian Mittenhuber has kindly provided me with the information set out in the Appendix (below), but I am responsible for the use made of it. Of the three forms, Korínion has always carried most weight so far, no doubt because it is overwhelmingly the most frequent form in the record. Korínnion is what we find in the manuscript considered to show the most ancient features, whilst Korónion is a ghost-form (it is, in other words, an error or invention that has entered the academic literature). However, I shall presume that the standard opinion is essentially correct, and that the form with <nn>, despite its venerability, has been influenced by knowledge in the Greek world of such personal names as Korínna and Korínnos. Note that if Korínnion rather than Korínion were taken as the authentic name of the town, none of the philological essentials below would be affected except in a single place (noted below).

There is no other known mention of Cirencester in antiquity, for example on coins of the local people, the Dobunni, or in inscriptions in stone. The place is notoriously omitted from its expected position in Iter XIII of the Antonine Itinerary,Footnote ³ which therefore provides no corroboration of Ptolemy. Corinium is the form which appears in all modern historical and touristic writing about the ancient place (and, for example, in the name of the local museum and football stadium). It is a regular latinisation of the Greek. However, this form does not actually appear in the record; hence the linguistic asterisk for a form reconstructed or inferred from other evidence. Linguists wrestling with this name have had to confront some difficulties. All have assumed that *Corinium is what lies behind the first element, Ciren-, of the current place name, and all have therefore encountered philological problems, which we must rehearse here.

Before we embark on the discussion, we need to attend to some terminology for the languages of Britain in antiquity and later. The main native language of Britain in the Roman period was British Celtic, recorded almost exclusively in a latinised form called Romano-British (RB), for which the main evidence is personal and place names. In the period after the collapse of Roman rule, this language evolved through a stage called, in this paper, Brittonic (roughly ad 450–550), for which there is little direct evidence, and from this into Welsh, Cornish and Breton (each being a distinct modern language, collectively called the neo-Brittonic languages; ad 550 onwards). The earliest stages of these, before written texts appear (ad 550–750/800), are called Proto-Welsh, Proto-Cornish and Proto-Breton, and the earliest texts are described as being in Old Welsh, Old Cornish and Old Breton.Footnote ⁴ The incoming Anglo-Saxons spoke a range of West Germanic dialects that we can label Prehistoric or Pre-Old English.Footnote ⁵ For the Old English (OE) of the settled Anglo-Saxons, there is some textual evidence from the period which coincided with Proto-Welsh and Proto-Cornish. This later evolved into Middle English, which competed in its early centuries with Anglo-Norman French.

The first element of Cirencester is believed to be etymologically connected with those of the nearby villages North and South Cerney and with that of the River Churn, which flows through all three places.Footnote ⁶ The exact development of the relation between the three place names is not particularly important in the present context. Suffice it to say the consensus, at least since Edmund McClure,Footnote ⁷ accepts that Ciren- and Cern- are forms of Churn that have been affected by the pronunciation of Anglo-Norman, and/or by the spelling systems used by scribes accustomed to writing administrative and legal documents in that language; or perhaps the scribes simply represented early Middle English as best they could using their current spelling conventions for vernacular languages (ie, everything except Latin), which had originally been devised for (Anglo-)Norman. I accept that the three are indeed related, and that the river name is probably the basis for the others. All four names are recorded from Anglo-Saxon times in spellings that indicate a pronunciation with initial [ $$$$ ],Footnote ⁸ ie <ci-> and <cy-> (but the latter is ambiguous, and for more thorough discussion see below). It is interesting that only three spellings for the inhabited places in the entire post-Conquest record have been noted that could suggest a pronunciation in [ $$$$ ]: Chiringecestre, in Gervase of Canterbury's Mappa Mundi (c 1270 in its present form), but this record is problematic in other ways too; Chirenchestre in Layamon's poem Brut (c 1200), and North Cherney, once in a foot of a fine written in 1646.Footnote ⁹ On the other hand the river name is not amply recorded, but all spellings which have come down to us from the post-Conquest period have <ch>. The form of 1646 is clearly an aberration influenced by the name of the river.

Irrespective of the detail of the relations between these names, their collective relation to *Corinium is problematic, and problematic to an extent that has been glossed over somewhat, both in the academic and in the public mind. Commentators have struggled, using philological special pleadings, to connect them with the RB name. Margaret Gelling remarked: ‘The transformation of Corinium into the first part of Cirencester has never been fully explained.’Footnote ¹⁰ The main difficulty begins with the fact that, on the ancient evidence, the British Celtic and RB name which represents it must have begun with the sound [k], as must its descendants in Brittonic and then Welsh. Therefore the change to OE [ $$$$ ], spelt <c>, needs to be explained. It is not straightforward for a British Celtic form beginning */ko-/ to yield an OE form spelt <ce-> or <ci->. We need a way of showing how the British Celtic rounded vowel *[o] represented by the <o> of Ptolemy's form can become an unrounded front vowel in Old English, ie [e] or [i], because these are the only vowels that effect the change of a preceding [k] to [ $$$$ ] in all dialects of Old English.Footnote ¹¹ A H Smith shows that W H Stevenson was wrong to believe that *Corinium could directly yield an ancestor, *Kerin-, for the OE name by way of the British Celtic phonetic change called internal i-affection, because that change is one which operated in seventh- or eighth-century Proto-Welsh, that is after the period of Saxon expansion that engulfed the territory around Cirencester.Footnote ¹² If the Saxon takeover of the area was linguistically effective from their capture of the town in ad 577, as the Anglo-Saxon Chronicle allows but does not require us to believe, internal i-affection would have operated too late to have produced a Brittonic form which could give rise to an OE form of the type *Cerin. Brittonic speech may well have survived in the Cotswolds after 577, but it is unlikely that changes in their language after that date would have been copied into name-forms already borrowed into English. Given Cirencester's importance, it may have been known to the Anglo-Saxons well before the date of this battle and therefore have been even less likely to be subject to a later Brittonic sound-change.Footnote ¹³

The Swedish toponymist Eilert Ekwall had originally accepted Stevenson's view, but later recanted,Footnote ¹⁴ and countered this difficulty by proposing that Ptolemy's Korínion was actually a scribal error for *Kornion, a reduced form of a hypothetical *Cornovion [sic] related to the well-known tribal name Cornovii. He says that ‘[s]hortened … forms of tribal names and of place names occur occasionally in Celtic languages’.Footnote ¹⁵ A name of this alternative shape could be subject to the somewhat earlier Brittonic phonetic change of final (as opposed to internal) i-affection, which could indeed yield a Brittonic form *Kern-. This is a form that would be of the right shape when borrowed into Old English, and could, for technical reasons within Old English, yield the range of spellings found in the Anglo-Saxon period, ie <ci-> and <cy-> (and the <ce-> found only in Welsh or Welsh-mediated sources). The Brittonic sound [k], traditionally spelt <c> in the Celtic languages, would become [ $$$$ ] in Old English before an unrounded front vowel, ie [e] or [i], as noted; <y>, in the West Saxon Old English of the records in which Cirencester and the Cerneys appear, may in some words also represent [i], ie a vowel compatible with a preceding [ $$$$ ], namely those which in earlier Old English had [i] or [i:] or the sound of more than one origin spelt <ie>,Footnote ¹⁶ and such spellings in the record (of which some are given in note 6) are best taken as being equivalent to <ci->. The consonant cluster [rn] might be broken up by a so-called epenthetic vowel, represented by the <e> of later spellings. The borrowed name might also have been influenced by the form of the OE word cerin or cyrin, which means ‘churn’ and is the ancestor of this modern word. Ekwall suggests, as an alternative, that the Brittonic name could after all have been borrowed if the relevant sub-case of internal i-affection had operated earlier than the others, as *Kerin from Corinium, and that this would, because of the structure of OE phonology at the time, naturally have taken on the shape Cirin- or Cyrin-. But that is so only in early West Saxon, ie the West Saxon dialect attested from about ad 890 onwards, because Cirin-/Cyrin- could arise from earlier *Kerin only through well-attested pronunciation changes in West Saxon,Footnote ¹⁷ and there is no textual evidence from the intervening years that might substantiate Ekwall's special pleading. Yet Ekwall is dependent on this possibility, because there are no truly OE spellings in <ce->.

Cirencester is mentioned in Asser's Life of Alfred in the Old Welsh form Cairceri, and by Nennius as Cair Ceri, which seems to indicate the reality of the Brittonic form used to explain the Old English.Footnote ¹⁸ Ekwall suggests, however, that this form has been influenced by the name of the village of Kerry in Asser's native Wales (modern Montgomeryshire; Ceri in WelshFootnote ¹⁹) and may not in fact represent an authentic name for Cirencester. It is also true that the disappearance of the final <n> from the presumed *Kerin has no ready explanation within Welsh.Footnote ²⁰ Moreover, Asser was writing in the late ninth century, so it is possible that Cairceri is an invention by an earlier scholar, welshifying the already 300-year-old English name on the basis of an existing Welsh word, most likely the ancestor of ceri, ‘berrying or fruiting trees; eg service-trees or medlars’;Footnote ²¹ that is, it is not truly an inherited Welsh name for Cirencester at all. Something comparable has certainly happened with other, later, Welsh inventions for English place names, such as Rhydychen for Oxford, of which it is a literal translation, and indeed the kaer vudei, an alternative for Cirencester itself, in the story of Arthur in the Red Book of Hergest, which means ‘churn town’ (not ‘(River) Churn town’).Footnote ²²

This is all problematical in several ways. Ekwall's first solution requires (1) the presence of the first <i> in Ptolemy's Korínion to be a scribal blunder; (2) the name of this place belonging to the territory of the presumably Anglian-speakingFootnote ²³ Hwicce people, as recorded in Anglo-Saxon times, to be always in a West Saxon form rather than Anglian; and (3) a vowel to appear precisely in place of the ‘lost’ <i> thanks to a wholly English process in later centuries. His second solution falls foul of his own objection based on the date of internal i-affection, though his formulation allows him the space to believe that that objection might just be unfounded; I think his original position was right. A further problem is that *korin-, the alleged base of the name, has no known British Celtic etymology, though that is not an insuperable obstacle, since ancient and obscured river names often outlast language shifts. David Parsons classifies Korínion among the ‘very difficult’ names in Ptolemy and, according to Patrick Sims-Williams, ‘The Celticity of [the name in Britain] is uncertain.’Footnote ²⁴ Smith associates *korin- with the stem of the RB tribal name Cornovii and with that of two forts in Britain mentioned in Roman-period documents and called Durocornovio.Footnote ²⁵ It was originally Ekwall who suggested the connection, and speculated that Cirencester, identified with one of the instances of Durocornovio,Footnote ²⁶ was a colony of the Cornovii, or that the local Dobunni were a branch of that people. However, since Cornovii cannot account for the shape of the OE name without the special pleading of ‘shortened forms of tribal names’, the phonological difficulties set out in the last paragraph but one need to be addressed to explain the name on the basis of Ptolemy's authority.

If the name is originally that of the River Churn, it has no analogues in the corpus of Welsh river names assembled by R J Thomas.Footnote ²⁷ Rivet and Smith reject O'Rahilly's suggestion of a connection with the Irish mythic personal name Cuirenn, which they relate instead to British Celtic *korjo-, ‘host’, plus a suffix *-no-.Footnote ²⁸Corio is in fact the name of a Dobunnic king,Footnote ²⁹ but there seems to be no pressing reason to associate it with the town name; if his name, apparently an -n-stem noun, had been suffixed with the *-jon seen in other town names, it would have regularly yielded RB *Corionium, which could not have given the recorded OE forms for the same technical reason involving internal i-affection already set out.

Ptolemy's witness, the only secure Roman-period attestation of the place, has been regarded as authoritative. But Cirencester is also mentioned in the Ravenna Cosmography, a text from very late antiquity, in the form in which it is known to us, as Cironium Dobunorum.Footnote ³⁰ This has evidently not been taken seriously by commentators because of the well-known and demonstrable unreliability of the anonymous Cosmographer's text as a whole, including, as it does, such letter transpositions as Condecor for the etymologically transparent British Celtic place name Conderco (Benwell, Northumberland).Footnote ³¹ The Cosmographer's editor Schnetz confidently emends to Corinium, presumably on Ptolemy's authority. Rivet and Smith regard Cironium as an error of the same type,Footnote ³² and there is a close and suggestive (though inverse) parallel for the vowel transposition in Termonin for Terminon.Footnote ³³ But the Cosmographer is not always wrong. His Brano Genium agrees with Ptolemy's Bran(n)ogénion (Leintwardine, Herefordshire), which is taken as authoritative, better than does the Antonine Itinerary's Bravonio,Footnote ³⁴ and it has been shown, for example, that the odd-looking two-line entry Rate Corion/Eltauori (Leicester) preserves a tribal name, Corieltauvi, found in recently reappraised inscriptions better than does the formerly accepted form Coritani (Koritanoí) of Ptolemy (Geography ii.3.11).Footnote ³⁵ Suppose Ptolemy's form is in error for the *Kirónion that the Cosmographer's form seems to attest. All the complex philology set out by Ekwall, and argued for ingeniously yet diffidently, is unnecessary. The Brittonic descendant of *Kirónion at the dawn of the Anglo-Saxon period would be *Kïren′, with [o] becoming [e] by final (not medial) i-affection, borrowed directly as OE Ciren-, and this can, as it stands, explain the recorded forms of Cirencester, Cerney and Churn.

There are more technical issues to explore, but for now note simply that this proposal resolves at a stroke the chronologically tricky phonological issues. It is true that the suggested original form of the name, *Kir/Kīr + -on-jon (see further below), with its transparent suffixation, makes the <-nn-> in ms X of Ptolemy (noted above), harder to account for, if not by the associations mentioned there. A simple explanation for the letter transposition in Ptolemy would be the influence of another place called Korínion, in Illyricum, modern Karin Donji, near Benkovac, Croatia (Ptolemy, Geography ii.16.2),Footnote ³⁶ or, even more likely, that of the familiar Greek place name Kórinthos (Corinth). In either case, this must have happened extremely early in the textual history of the Geography, in fact before the archetype of all known manuscripts.Footnote ³⁷

Patrick Sims-WilliamsFootnote ³⁸ has considered that the RB name of London, Londinium, might result from a transparent *Lindonium, ‘pool place’, by a vowel transposition identical to the one envisaged here, and has made the parallel explicit in personal correspondence (January 2013). However, I do not think Londinium can have the etymology he suggests, for reasons to do with the development of the name in Welsh,Footnote ³⁹ and the parallel cannot therefore be sustained.

The above comments presume that the <i> in the first syllable represents British Celtic, and therefore Romano-British, short */i/. The essentials of this analysis are also valid if the first vowel was British Celtic long */i:/. The OE form will probably, and the Middle English form certainly, have had short /i/ even if the British source had a long vowel, because long vowels are regularly shortened in the first syllable of a late OE word or name having at least three syllables.Footnote ⁴⁰ It is true that neither *kiron- nor *kīron- can be explained as British Celtic any more than can *korin-, but a proposal about its etymology is explored below.

If the argument presented so far is accepted, then the RB name of Cirencester is more likely to have been Cironium than *Corinium,Footnote ⁴¹ with a structure like that of the attested Viroconium (Wroxeter) and Cataractonium (Catterick). As noted above, there is no ready British Celtic explanation of a name with either form. Pursuing an etymology, we need to explore both of the possibilities raised above, namely an original short or long vowel in the first syllable. Following the idea that the river name is basic, we might propose a form cognate with Old Irish cíar, ‘dark, etc’, which is semantically plausible in a river name, and indeed has already been proposed for the river Keer in Westmorland and Lancashire and as the base of the Norfolk place name Carbrooke.Footnote ⁴² This is from Common Celtic (CC) *keir-, which has no surviving Brittonic cognate. Such a connection would seem to entail an irregular change of *[ei] (via [e:]) to long [i:] before the stressed medial -on- of the hypothetical British *Keironjon. That said, however, it must be noted that Isca, the RB name of Caerleon, some forty miles to the west, has exactly the same problematic relation to Wysg, the Welsh name of the River Usk on which it stands, which must descend from CC *Eiskā, British *Ēska, and the initial RB <I> representing British *[e:] appears aberrant.Footnote ⁴³

The argument offered about Isca (Caerleon) cannot be used to account for the name of the other Isca (Exeter), which must have RB short /i/. The English name supports the idea that the initial vowel was short, since the [e] can descend straightforwardly from Brittonic *[e], representing British */i/ normally in a syllable before one containing the suffix *-ā. Cornish evidence supports the idea too, if the (unique) form used by Edward Lhuyd in Archaeologica Britannica Footnote ⁴⁴ is trustworthy. Retranscribing the Cornish folk tale ‘John of Chyanhor’, originally written down by John Boson,Footnote ⁴⁵ Lhuyd writes Karēsk, where the <ē>, despite representing a long vowel in late Cornish, must descend from original short */i/.Footnote ⁴⁶ Oliver Padel has suggestedFootnote ⁴⁷ that this name-form may be a creation modelled on the Welsh use of caer, rather than a traditional Cornish form, and that the entire form needs to be treated with caution. There is room to wonder whether the spelling originally used by Boson, Ka'rEsk, has <e>, representing short /e/, as a borrowed feature of general south-western English, since Middle English short [i] is recorded as sporadically lowered to [ε] in Cornwall, Devon and Somerset at the time of the Survey of English Dialects in the 1950s,Footnote ⁴⁸ or even whether his supposedly Cornish form reflects the English initial vowel of Exeter directly. Either way, there is no evidence for a long vowel in Isca (Exeter) as there is in Isca (Caerleon), and this discrepancy remains a puzzle.

We might therefore tentatively suggest that the CC */ei/ which later became */e:/ could appear dialectally in the Romano-British of the Severn area as /i:/, ie that our Cironium might stand for a RB *Cēronium, just as Isca appears to stand for RB *Ēska. If CC */ei/ can indeed sometimes appear as RB /i:/ rather than /e:/, we could recall O'Rahilly's argument,Footnote ⁴⁹ by no means universally accepted, that certain insular names with */i:/, such as the ancient name of Ireland traditionally written as Hibernia, are in fact alternatives for older forms in */e:/ from CC */ei/. That would allow us equally tentatively to add to our short list of affected names the Ivernium in the Ravenna Cosmography, generally taken to be (an unidentified place on) the River Iwerne in Dorset.Footnote ⁵⁰ However, this is doubtful, and the parallel to Cironium offered by Isca (Caerleon) remains the single most defensible one.Footnote ⁵¹

There is evidence that Latin /i:/ was open enough (ie with the blade of the tongue lowered) to be rendered in some exceptional borrowed words as Brittonic */e:/, and thus ultimately and regularly as the Welsh diphthong spelt <wy>: note Welsh pabwyr, ‘wick’<papīrus, paradwys, ‘paradise’<*paratīsus, and synnwyr, ‘sense, skill’, if this is from Latin sentīre and not sentērium; also, the Berkshire place name Speen (from attested RB Spinis, ie Spīnīs) can only descend from a hypothetical Brittonic *Spēn-, which demonstrates the same change of [i:] to *[e:].Footnote ⁵² This fact clearly has a bearing on whether the inverse possibility floated above is plausible, namely that British Celtic */e:/ could be understood as British Latin /i:/ and appear as such in Romano-British place names.

If the vowel in the initial syllable of RB Cironium and in Isca (Caerleon) were short even though the vowel of the source were British long [i:], that would be a problem, because a short vowel deriving unconditionallyFootnote ⁵³ from a diphthong or long vowel would remain a puzzle from the phonetic perspective. But this is purely hypothetical, because, as we have seen, we do not need to postulate such a change in British or Romano-British since a surviving British long vowel is not ruled out in these names.Footnote ⁵⁴

What, on the other hand, if the CC and RB vowel in Cironium had been short */i/ after all? Whilst the proposed long vowel is clearly defensible, it might be possible to mount a defence of the short vowel starting from Old Irish cir, ‘grey’, a form of the same Proto-Indo-European (and hence Common Celtic) root meaning ‘dark’ that we have discussed, but representing what is technically called its zero grade, which appears as */i/ in Celtic.Footnote ⁵⁵ This would entail abandoning the partial parallel of Caerleon/Caer Wysg and all the issues of dialect variation and merger of the RB phonemes /i:/ and /e:/ that followed from it. The question of long vowel versus short vowel cannot be fully resolved, but it is important to realise that it has no bearing on the preference for Cironium rather than *Corinium.

In light of the possibility that RB <i> might represent British Celtic */e:/, the very obscure tribal name Silures, belonging to western neighbours of the Dobunni, might be understood as late British *Sēlo-ūr- (Welsh hwyl, ir), ‘(green) fresh, vigorous, for a journey, rush, or attack’,Footnote ⁵⁶ with the theme-vowel elided before another vowel (in Romano-British), and with the ancestor of ir displaying something of the meaning-potential apparent in the medieval poet Dafydd ap Gwilym's line ‘A'r enaid yn ir ynof’ (‘And the spirit fresh inside me’), though it is true that this stands in a poetic context where the literal meaning of ir, ‘fresh (of vegetation)’, is exploited.Footnote ⁵⁷ This suggestion is highly speculative and not offered confidently.

In summary, then, if the argument presented in this paper is correct, the RB name of Cirencester is more likely to have been Cironium than *Corinium. Cironium represents a long-vowelled RB *Cīronium, British *Kīronjon, or possibly a similar form with a short vowel RB *Cironium, British *Kironjon. This conclusion highlights the need for a more critical evaluation of our most important documentary sources, since in relation to this name the many manuscripts of Ptolemy are collectively found wanting, and the denigrated Ravenna Cosmography allows us simultaneously to clear up a difficult issue in English onomastics and to propose a satisfactory Celtic etymology for the place name. An additional consequence is to make it more unlikely that Asser's and Nennius’s already suspicious Caerceri/Caer Ceri is an authentic reference to Cirencester. It is possible that the town is intended, but the name-form is bogus, since it cannot descend either from a British */kir-/ or */ki:r-/ or from */ke:r-/.Footnote ⁵⁸