Caersws, a village in mid-Wales, was described long ago as a hamlet in the ‘county of Montgomery, North Wales, five miles (west) from Newtown; containing 342 inhabitants. This is said to have been a Roman city of considerable extent, though unnoticed as such by early writers: its Roman name is not even known’.Footnote 1 Unnoticed, too, is the possibility that the lost Roman name may exist in the eighth-century Ravenna Cosmography, an ample source for the place-names of early Britain, if (thanks to profound textual corruption) the most puzzling. Even experienced scholars might feel an urge to let it alone, since the location of the toponyms that it lists is often of the vaguest. Nevertheless, some of its tougher nuts may be cracked by consecutive reasoning. This paper takes on one such nut, Argistillum.

Standard opinion is that the name is related to Welsh gwystl (‘hostage’), early Irish gíall (also ‘hostage’) and Arwystli (the south-west region of the former county of Montgomeryshire). The meaning is taken as ‘at the hostage’. The significance of that being opaque, it is explained somewhat weakly as perhaps from ‘a folklore reference now lost’. Difficult though the sense is, it is offered on the grounds that no one has thought of anything better.

The location of Argistillum has been as obscure as its meaning, but the name occurs in the Ravenna Cosmography between Glebon (Gloucester) and Vertis (Worcester), which is followed by Salinis (Droitwich). For this reason, it has been suggested (despite the admitted anomalies in Ravenna's lists of toponyms, which mean that they are not always listed in neat geographical order) that Ariconium, the Roman settlement at Weston under Penyard, near Hereford, could be a corrupt form of Argistillum, and that the name might contain the Celtic root *arganto (‘silver’, as in Welsh arian (‘money’)).Footnote 2 Of this suggestion, one can say merely that it would be strange indeed if Ariconium were so grievously corrupted as to leave only the first two and last two letters intact, while at the same time producing a form that makes tolerable sense in British Latin.

So let us survey instead what has been said about the Welsh word arwystl (‘pledge’) and the remote district of Arwystli, in Powys. Arwystli was described in stately terms by Sir John Lloyd as ‘the land of the headwaters of the Severn’. He declared that it was surely part of ancient Powys, but at an early date became connected with Gwynedd. It thereby ‘became permanently attached to the bishopric of Bangor’, forming an enclave that looks so odd on maps of Welsh dioceses. It had royal manors at Talgarth, in its lower part, and Caersws, in its upper. It main churches were at Llandinam (near Llanidloes) and, in its highlands, Llangurig, whose patron, St Curig, was (according to Gerald of Wales) held ‘in high esteem as a worker of miracles’, aided by his crozier, still preserved there in the twelfth century.Footnote 3

Lloyd made no comment on Arwystli's name. For that we turn to others. Lloyd-Jones's glossary of Welsh bardic language defines arwystl as ‘pledge, warranty, bond’ and the simplex gwystl as ‘pledge; hostage’. It also lists Gwystl as a personal name or name-element in early Welsh (as with Arwystl and Nant Catguistil in the twelfth-century Book of Llandaff) and Cornish (a manumitted slave at Bodmin was called Tancwoystel).Footnote 4 Arwystli has been called the territory of a man called Arwystl or of his descendants, just as Cydweli, or Kidwelly (now a town in south-west Wales), was once the territory of a chieftain called Cadwal or of his heirs, while the -i in Arwystli would derive from an original ending in reconstructed British -ion, where i is long and bears the stress.Footnote 5 The early Irish cognate of Welsh gwystl is gíall (‘hostage’), the oldest word for a personal surety, and familiar from legal texts, which state that some hostages have a place of honour near the king, but those whose givers have revolted or turned traitor must be kept in chains with armed guards beside them, their lives being forfeit.Footnote 6 Gwystl likewise figures in Welsh law tracts, though in milder terms.Footnote 7

If Welsh gwystl has (with a slightly anomalous development) a cognate in early Irish gíall, Welsh Arwystli has a near Irish equivalent in Airgíalla. This began as the name of a people, but was later used of their territory in what are now the counties of Fermanagh, Monaghan and Armagh (it is thus possible that Arwystli began as an ethnic form rather than a territorial one). Airgíalla is anglicized as ‘Oriel’. Because the Airgíalla were subordinate to their neighbours, their name has been understood as ‘the submitted, the givers of hostages’.Footnote 8 Yet this is unlikely. They would hardly call themselves anything so unflattering, and Welsh evidence points to another meaning. Arwystl, which is both a common noun and a personal name, is explained from the intensive prefix ar- plus gwystl (‘pledge’); the second has cognates in the Gaulish personal name Congeistlus, as well as in Cornish and Breton.Footnote 9 So an arwystl was a pledge, security, surety, earnest or bond of special weight or seriousness.

Having cleared up some philological misunderstandings, we turn to the archaeology of the Romans in Arwystli. Commentary here is dominated by accounts of its main site, at the village of Caersws (Powys: so 0391), which was long ago taken, somewhat doubtfully, as the Ravenna Cosmography's Mediomano.Footnote 10 But Mediomano is now accepted by standard authorities as a corrupt variant of Mediolanum, or Whitchurch, in Shropshire. Studies underline the importance of the location of Caersws for Roman Wales, when it was a natural centre of north to south and east to west communications. The earliest fort, discovered from crop-marks in 1958, lay a mile east (at so 042925), and was used probably only during the initial assault on Wales in the first century ad. It was replaced in ad 78 or so by a smaller fort, north west of the modern village. Caersws railway station now occupies much of the site, with the former goods shed standing on top of the Roman bath house.

Excavations of 1966–7 revealed the complicated history of the defences. The original turf rampart was reconstructed in the late first century; in the late second century it was rebuilt with a stone facing. The fort then acquired central buildings of stone, although its barracks remained of timber.Footnote 11 Occupation continued into the fifth century, both of the fort and the civil settlement outside, so that pottery is still often found in village gardens. The unimpressive remains of the fort's rampart can be viewed where it is crossed by the main road to Machynlleth.Footnote 12 Even if little is now visible, the settlement had four centuries of continuous habitation, until after ad 400, no doubt because of its strategic position at the junction of five Roman roads. Some of these are obscured by the modern highways that follow their course, but others can be detected (as at so 046956 and 052922) in the countryside near Caersws.Footnote 13

The Romans mined lead near Caersws – as indicated by the Roman road leading westwards to the ancient mines near Dylife – and mines needed soldiers to guard them.Footnote 14 A map of Wales makes obvious the strategic advantages of a fort at this spot. If there were trouble-makers anywhere in mid-Wales, troops might be dispatched on roads running in five directions to deal with them. So the long-term significance of this Roman establishment in Arwystli is evident. Its success as a community in ancient and medieval times is in contrast to its poverty during the early modern period, if we are to believe John Leland.Footnote 15 He remarked that there was no town worth the name in Arwystli nor any market except at Llanidloes, but conceded that ‘at poore Cairllews hathe bene a market and borow privilegid’. The abandoned streets of that ‘borow’, larger than the shrunken village succeeding it, can still be traced in hedgerows and lanes beyond the modern houses.

All this implies that Caersws, for centuries the nub of military communications in mid-Wales, and even in medieval times a community of some pretensions in Arwystli, could be the Ravenna Cosmography's mysterious Argistillum.Footnote 16 The form could be understood as ‘place of Argistillos’, or (in the light of Gaulish-Latin Congeistlus) ‘place of ^* Aregeistlos’, the scribes having perhaps been confused by Latin argilla (‘potter's earth’) or Argiletum (the booksellers’ area of ancient Rome). Argistillum might, on the same basis, be emended and reconstructed as British ^* Aregeistlion, with stress on the final ‘i’ (as argued by Jackson above). Ravenna's Argistillum and modern Arwystli would be related forms. The element -geistlo- was clearly potent in ancient times, being borrowed by Germanic to give Old English gisl (‘hostage’).Footnote 17 The power of the word is confirmed by its use to describe Irish hostages awaiting execution once their givers had proved treacherous.

Nevertheless, the element was rare in place-names, as is shown by its absence from one recent survey.Footnote 18 Yet later evidence from Wales indicates its high status there, since Argistillos or ^* Aregeistlos gave the medieval name Arwystl, borne by (amongst others) an eleventh-century Bishop of St Davids.Footnote 19 Gwystl itself occurs as a personal name in the tale of Culhwch and Olwen, also of the eleventh century.Footnote 20 As for the formation, this is paralleled in other British-Latin forms, as with the Roman settlement at Leintwardine in Herefordshire, which was called Branogenium (‘place of Branogenos’; that is, ‘place of the crow-descended one’).Footnote 21 Argistillos would thus surely be an ancient British chieftain or local ruler who gave his name to the cantref and (after Henry viii) the hundred of Arwystli, Powys. Caersws itself was the ‘fort of Swys’, Swys being a person otherwise unknown (despite the unlikely suggestion that she was a ‘beautiful queen’).Footnote 22

If Argistillum is correctly identified as Caersws, it has several implications. We can stop translating the form as ‘at the hostage’ and invoking obscure folklore. We should similarly refrain from emending it to ‘Ariconium’ or Weston under Penyard, or words for ‘silver’.Footnote 23 It can be seen instead as called after an ancient local ruler called Argistillos or, better, Aregeistlos, whose name meant ‘pledge, surety, bond’. He was given out as trustworthy, a man of his word. Ravenna's Argistillum would, on that basis, be a Latinization of reconstructed British Aregeistlion, place of Aregeistlos.

Second, we can see the formation of the territory of Arwystli as going back to pre-Roman times. It would be an ancient domain, like Dyfed (compare Ptolemy's Demetae), and did not take its name from some early medieval ruler, as is the case with (for example) Ceredigion or Glamorgan. Third, we may believe that the Airgíalla of Ulster, who gave their name to Oriel, were not known humiliatingly as ‘the submitted, the givers of hostages’, but more proudly as ‘those who give their pledge’, a people deserving trust.

Fourth, we can, with some confidence, fill a gap on maps of Roman Britain, and where the Ordnance Survey had Mediomanum at Caersws, we can put Argistillum.Footnote 24 Fifth, we are reminded of Arwystli's perennial role as a corridor through which travellers pass (even if its western parts are some of the loneliest places in south Britain). Formerly they were traders, soldiers and those on more nefarious errands, such as Gilfaethwy and the sorcerer Gwydion in the Mabinogi tale of Math, who hurry through Arwystli northwards with swine stolen from the Prince of Dyfed.Footnote 25 Sixth, it fosters the hope that further place-names will be recovered from our early sources by the application of local knowledge and coherent thought.

Armed with such knowledge, we may feel that the identity and location of Argistillum is not an unidentified spot near Gloucester, but an important mining centre and military base in mid-Wales.Footnote 26 Despite the pessimism of the topographer quoted at this paper's beginning, we can be reasonably certain that this ‘Roman city of considerable extent’ was actually noticed by early writers and that its ‘Roman name’ has come down to us. Such was the place's importance that it was known to the eighth-century Ravenna Cosmographer, writing far from the summit of Pen Pumlumon Arwystli (sn 8187), on the western fringe of Powys, whose name even now preserves this ancient form.