One of the more startling claims made by Steve Blake and Scott Lloyd in their dreadful book The Keys to Avalon: the true location of Arthur’s Kingdom revealed (Shaftesbury: Element, 2000) is that what has been known for the past thousand years and more as ‘Offa’s Dyke’ was in fact built by Roman armies during the reign of Septimius Severus (Emperor 197-211). They use literary texts and, as a ‘stop press’, a radiocarbon date for Wat’s Dyke, whose relevance is not immediately apparent. Interestingly, they hardly use archaeological evidence.

Redesigning geography and recasting history

On what do they base their claim? The argument begins in their redesigned geography of Britannia/Ynys Prydein. They locate the Saxon invasion in Gwent, which they derive from the Ceint of the Historia Brittonum in contrast to the accepted derivation Venta>Guent>Gwent. They then take the Historia’s statement (Chapter 38) that illis regiones quae sunt in aquilone iuxta murum qui uocatur guaul ‘those regions which are in the north, next to the wall called Gwawl’ with the earlier statement (Chapter 23) seuerus… murum et aggerem a mari usque ad mare per latitudinem brittanniae, id est per cxxxii milia passuum deduxit et uocatur brittannico sermone guaul (“Severus… built a wall and embankment from sea to sea across the width of Britain, that is for 132 miles, and it is called Gwawl in the British language”.

Turning then to the Historia Augusta, a late fourth-century compilation of Lives of various emperors from Hadrian to the later third century, they make the astounding claim that “this one text alone forms the basis of our knowledge of the names of the Romans who built the Hadrian and Antonine Walls”. This does immense disservice to many years of archaeological investigation of the Walls, but no matter. They are more impressed with the statement in the Historia Augusta’s Life of Severus (xviii.2) that brittanniam, quod maximum eius imperii decus est, muro per transuersam insulam ducto utrimque ad finem oceani muniuit “he fortified Britain, which is a great ornament of his reign, with a wall led across the breath of the island to the edge of the Ocean on both sides.” Again, at xxii.1, the Life mentions murum aut uallum… in Brittannia “a wall or rampart… in Britain.”

Eccentric translations

They attempt to translate Latin insula as ‘realm’, which is a completely unattested use of the word in a text written in the Mediterranean world, let alone Britain, but once again, let us pass over that. They criticise the consensus view that the documentary references to a Wall of Severus merely refer to a refurbishment of Hadrian’s Wall on the grounds that “the texts clearly state that there was a third Roman wall in the British Isles” (their emphasis). Once again, there is a problem that they gloss over: there is not one single text that states that there were three walls in Britain.

However, they draw a number of conclusions from their reading of the texts:

• Septimius Severus built a defensive earthwork
• It went from sea to sea
• It was 132 milia passuum long
• It separated Deifr and Alban

Trusting dubious and late sources

Having established this to their satisfaction, they then turn to the Brut for evidence about the location of the Wall. They take its statement that Oswald, King of Northumbria, fought Penda, King of Mercia, at Maes Nafawl or Heavenfield, near Severus’s wall. This derives from Geoffrey of Monmouth (Historia Regum Britannię xii.10), where the mistake of confusing the Battle of Heavenfield with the Battle of Maserfelth is first made. They do not spot this and instead castigate those who “have located this battle near Hexham… because of the reference to ‘the wall’, despite the fact that no evidence exists to back up this claim”. Evidently they have not read Bede (Historia Ecclesiastica iii.2), where hefenfelth… hagustaldensis ecclesiae, quae non longe abest, “Hefenfelth… the church of Hexham, which is not far away”, writing less than a century after the event. But again, never mind. How do they pursue their quest for the true geography of Britannia/Ynys Prydein? They accept Geoffrey’s confusion and identify the Battle of Heavenfield, as a result of which Oswald became King of Northumbria, with the Battle of Maserfelth, at which he lost his life. They then use the twelfth-century Vita Sancti Oswaldi to show that this battle was near Oswestry, Shropshire. A major earthwork frontier passes close to Oswestery: Offa’s Dyke. This, they argue, is Gwawl, the ‘Wall of Severus’.

They make a great deal of the length of the Wall of Severus given by Late Roman and Insular writers, where the figure cxxxii (132) milia passuum (Roman miles of 1,480 m) is generally given. This is a little over 195 km (a little over 121 statute miles); Offa’s Dyke is around 192 km long (a little over 119 statute miles). The coincidence of length is impressive, it has to be said, even though there are only 130 km (81 statute miles) of earthwork. But what about the name?

Dismissing accepted wisdom

Offa’s Dyke is first attributed to the King of Mercia (757-796) by Asser (Vita Regis Ælfredi Chapter 14), where it is said that fuit in mercia moderno tempore quidam strenuus atque uniuersis circa se regibus et regionibus finitimis formidolosus rex, nomine offa, qui uallum magnum inter britanniam atque merciam de mari usque ad mare fieri imperauit “In recent times, there was a certain king in Mercia, vigorous and terrifying to all the kings and regions around him, Offa by name, who ordered a great wall to be built between Wales and Mercia, from sea to sea”. However, Blake and Lloyd are worried by “how it was that the Saxons managed to build such a major structure whilst under constant attack from the Welsh and how a people with no history of building huge earthen banks… had the engineering skill to undertake such a vast project”. We can pass over in silence the Wansdyke and numerous other Early Saxon defensive earthworks and ignore the evidence (such as the Annales Cambrię) that suggests that Offa was the aggressor, who constantly harried the Welsh.

But what of the “Roman artefacts… found within the Dyke”? This is an archaeological question, and one that they do not pursue. Typical Bad Archaeologigists, they evidently do not understand the basic archaeological concepts of the terminus post quem – the principle that states that no deposit can be older than the date of the most recent object found within it – and residuality – the idea that objects can turn up in deposits of much more recent date. These are important points, as they are crucial to understanding the radiocarbon evidence.

Having established the coincidence of length between Offa’s Dyke and the ‘Wall of Severus’/Gwawl, Blake and Lloyd start looking for documentary confirmation that the two are one and the same. They find it in an eleventh-century Irish translation of the Historia Brittonum known as Lebor Bretnach (not “an Irish manuscript compiled in the eleventh century”, as five manuscripts are known, ranging from an eleventh-century fragment down to full texts of the fifteenth and sixteenth centuries). In the Irish translation of Chapter 23 of the Historia, Chapter 17 of Lebor Bretnach has Seuerus in treas ri tainic i mBretnaib. Is leis doronad clad Saxan a n-agaig na mBarbarda .i. Cruithneach. Da mile xxx ar c cemann ana fad, ₇ is e ainm an claide sin la Bretnu Guaul “The third king who came to Britain was Severus. He made the Saxon trench against the barbarians, i.e. Picts. It is 130 miles long and the name of the trench in British is Guaul”. Where the translator of Lebor Bretnach (an Irish scholar called Gilla Coemghin) got the description of Gwawl as being Saxon is unclear, but Blake and Lloyd see this as an important confirmation of their revised toponymy. They have no problem with Picts bordering Offa’s Dyke, as they have already relocated them in North Wales (as the Gwyddyl Ffichti of Powys, mentioned only in one of the genealogies collected in the fourteenth-century Jesus College, Oxford, MS 20).

However, Gilla Coemghin continues ₇ roforcongair clod aili do denum a n-agaid Gaedeal ₇ Cruithnech .i. clad na muice, “and he ordered another trench to be made against the Scots and the Picts, i.e. the Trench of the Swine”. This is interesting, and Blake and Lloyd do not quote it; Gilla Coemghin understood there to be two walls in Britain, so he corrects the Historia Brittonum and attributes the second wall to Severus, too. Whereas the first was called Guaul, the second is known as the Trench of the Swine. A variant of this name (Swine’s Dike) was recorded by the early eighteenth-century antiquary Horsley as a local name near Falkirk for the Antonine Wall. It is important to note that here Cruithneach, ‘The Picts’, are associated not with North Wales, but with Swine’s Dike in central Scotland; we may assume that Gilla Coemghin believed that the same Picts were to be associated with Guaul.

Notwithstanding this, Blake and Lloyd consider the case more-or-less proven at this point. They believe that Gilla Coemghin specified that the ‘Wall of Severus’ was clad Saxan, “the Saxon trench” because he knew of Asser’s attribution of it to Offa over a century earlier. This does not seem very likely, as Asser’s Vita Regis Ælfredi does not seem to have been widely known. Still, there is “a stone inscribed to him that was found in Caernarfon, and it has also been suggested that he rebuilt the walls of Chester”, which they take as evidence that the emperor was in Wales at some point.

Finally, how did Asser come to misattribute the ‘Wall of Severus’ to Offa? They note that there are several placenames associated with both Offa’s and Wat’s Dykes that contain the name Offa. This is an error, they suggest, for Hengist’s cousin Ossa, ”whose name is wiritten in many manuscripts as Offa because of the similar appearance of the letters s and f in old manuscripts”, although they do not suggest how palaeography might explain changes in placenames. There is a problem, predictably. In Geoffrey of Monmouth, following the Historia Brittonum, the names of the two relatives of Hengist given land near Guaul are Octha and Ebissa; the Brut changes Ebissa to Ossa Gyllefawr, identified with the grandfather of Ida of Northumbria in the Historia Brittonum (Chapter 57). This is one of numerous changes made by the Brut to Geoffrey’s text, which derives much of its narrative from the Historia Brittonum; these changes are generally to bring the Brut into line with vernacular traditions.

Where is the archaeological evidence?

This is the evidence assembled by Blake and Lloyd in their reassessment of Offa’s Dyke and their attempt to show that it was built by the Roman emperor Septimius Severus. Their evidence consists almost entirely of documentary discussion, much of it using dubious sources, many of which are presenting information at second or third hand. The various holes in the treatment of the evidence have been pointed out, but it is necessary to go further, to look first at the documents, then the archaeology and finally the historical probabilities and possibilities. The case for renaming Offa’s Dyke turns out not just to be weak, but wrong in the final analysis.

The Roman walls of Britain

The two Roman walls – those of Hadrian and Antoninus Pius – are well known monuments that have been studied for several centuries. Consequently, they are understood in great detail and a vast amount of archaeological evidence has been assembled not just about the walls but about their supporting infrastructure and general context. If the earthwork we know as Offa’s Dyke is in fact of early third-century date, we would expect it to exhibit many, if not all, the features of these two walls, especially the Antonine Wall, which is an earthwork embankment. On the other hand, if Offa’s Dyke is early medieval, then it ought to display features consistent with other early medieval earthworks, such as Wansdyke.

The Antonine Wall consists of a turf rampart at least 3 m (10 feet) high and perhaps as much as 3.7 m (12 feet) high; this was laid on a stone base usually 4.3 m (14 feet) wide. In this respect, it resembled other linear frontier works (such as the German and North African limites). To the north of the wall, at a usual distance of 6.1 m (20 feet), lay a ditch 12.2 m (40 feet) wide and 4 m (13 feet) deep in the eastern sector and averaging 8.4 m (27½ feet) wide and as little as 1.8 m (6 feet) deep in the western sector. On top of the wall stood a wooden palisade and walkway.

Immediately south of the wall ran a road about 5.5 m (18 feet) wide; this was an innovation, as Hadrian’s Wall was served by the existing Stanegate, some distance to the south. The wall was built in segments by detachments from the three legions serving in the province, who recorded their work on highly decorative distance slabs. Numerous temporary camps housing the troops involved in the building work have been located. Finally, some nineteen forts were placed at intervals along the wall, at an average distance of about 3.25 km (although this varies considerably). The forts themselves vary in size but contain the usual range of buildings (headquarters buildings, commandants’ residences, barracks, stores, granaries, stables and so on). With one exception (Cadder), the forts faced north, towards hostile territory. Fortlets and beacon platforms have also been recognised on the wall.

What we have in the Antonine Wall is a complex and integrated system. There is a great deal of archaeological evidence for its construction in the form of temporary camps and building inscriptions, then of its garrison. We do not have to rely on the fourth-century Historia Augusta to tell us that it was built under Antoninus Pius, as the distance slabs record the name of the emperor.

Offa’s Dyke

How does Offa’s Dyke compare? If, as Blake and Lloyd assert, it was actually built by the emperor Septimius Severus, who was in Britain from 208 until his death in 211, it ought to show many similarities to the Antonine Wall, built almost seventy years earlier. It ought also to show some innovations based on the experience of that wall.

The earthwork construction of the Dyke varies considerably along its length and, unlike the Hardrianic and Antonine frontiers, it is not continuous throughout its length, with many original gaps, including one of 96 km (60 miles) where the River Wye appears to have formed the frontier. A total of 130 km (81 miles) were built as an earthwork, standing at least 7.3 m (24 feet) high. At Llanfynydd, at the northernmost end of Offa’s Dyke, the ditch was found to be at least 4 m (13 feet) wide and 1.5 m (4.9 feet) deep; there was no gap between the ditch and the base of the bank. In places, the combined width of the bank and ditch is 20 m (66 feet). There is no evidence of a continuous palisade on top the Dyke, although there seems to have been a stone wall in places and timber fencing in others, while some sections of Wat’s Dyke appear to have had a timber frontage to the rampart (as at Sychdyn, near Mold). Both Dykes lack the infrastructure seen at the Antonine Wall: there is no military road, no garrison stationed in forts attached to the Dyke (although Cwrt Llechrhyd, a moated site at Llanelwedd in Powys had been claimed as an Offan fort), no temporary camps to house the builders, no building inscriptions. Moreover, the Antonine Wall is full of Roman artefacts recovered during excavations: Offa’s Dyke has only scraps of abraded Roman material culture within its structure. Blake and Lloyd talk about the “Roman artefacts… found within the Dyke” as if they date its construction. The concept of the terminus post quem ought to tell us that the Dyke is of Roman or later date. This is entirely possible because of residuality.

The Dyke is also completely unrelated to the pattern of early third-century military sites in the region. This includes legionary fortresses at Chester in the north (undergoing considerable refurbishment early in the third century) and Caerleon in the south as well as auxiliary forts, such as those at Leintwardine, Caersŵs and Forden Gaer. The road system shows no sign of being aware of the Dyke. Moreover, the only dating evidence from stratigraphy proves it to be later than second-century occupation at Ffridd; how much later cannot be determined on archaeological grounds alone.

Monuments in context

The most devastating argument against regarding Offa’s Dyke as a Roman defensive work is that of context. What possible function could it have performed? To the east of the Dyke, the Midlands of England were a reasonably prosperous civil province of the Roman Empire. To its west lay further areas under civilian rule (notably in the south) as well as areas under military control (predominantly in mid and north Wales); it was every bit as much part of the province as the area to the east. Linear defensive works elsewhere in the Empire mark the boundary between civilised, Roman life and barbaricum, the uncivilised world outside, which might, at best, be home to a few outpost forts. Third-century Wales can in no way be thought of as anything other than part of Britannia.

There is, moreover, a Saxon context for the Dyke. Apart from Wat’s Dyke, which marks a slightly different boundary in the north and continues to the Dee Estuary, there are numerous post-Roman earthworks across England. The majority of them are to be found in eastern England and most are short structures lying across major routes. One possibly relevant earthwork, though, is the Wansdyke, an earthwork boundary south of the Thames, defending the area to its south. This seems to have been built in the fifth or sixth century to defend the British kingdoms of the southwest against attack from the Thames Valley, where Saxon kingdoms had been established. Unlike the Roman linear frontiers, these dykes were not provided with garrisons, but often appear to be more in the nature of boundaries imposed by a militarily dominant power. They were not located to defend the areas behind them but to act as a line of demarcation. The tradition of earthwork frontiers was a long one in Anglo-Saxon England and it provides a good context for the construction of Offa’s Dyke in the late eighth century as one of the last and certainly the greatest of these structures. Thousands of men were needed to build the Dyke, proof that the kingdom of Mercia was highly organised and under strong central control. The ninth-century history of Mercia, with the destruction of its bureaucracy and ecclesiastical structure by the Vikings means that its place in the history of Britain has often been undervalued; to regard its people as barbarians incapable of such works (as Blake and Lloyd do) not only ignores their long tradition of dyke building but also shows woeful ignorance of the political sophistication of Mercia. Offa regarded Charlemagne as an equal (even if Charlemagne did not reciprocate the compliment); this was more than self-flattery.

All in all, The Keys to Avalon is a muddled and amateruish book; its authors show little critical sense in evaluating documents, no understanding of the principles of etymology and utter ignorance of archaoelogical methods and theories. The ‘Wall of Severus’ is a complete non-starter and Offa can rest safely in his grave secure in the knowledge that he can still be accorded the honour of constructing the single largest archaeological monument in Britain.