Birdoswald Roman Fort: dating the post-Roman use of the site

In another article, I described the post-Roman timber halls constructed on the site of the north granary in the Roman fort of Birdoswald on Hadrian's Wall. We know the halls are post-Roman because late Roman coins of Valentinian, dated to the 380s, were found beneath the relaid floor. How long did the site remain in use after the end of Roman occupation?

Dating the post-Roman halls

Unfortunately, there are no surviving timbers from the post-Roman timber buildings that could be dated by dendrochronology or radiocarbon, and there are no datable objects from the site after the coin of Theodosius found under the collapsed roof of the south granary (Wilmott 2001 p. 123). So dating the post-Roman halls is a matter of interpretation.

There are two distinct timber building phases to accommodate, the first one built more or less directly on top of the north granary and the second one built partly on the north granary and partly on the adjacent street. Both must post-date the coins of Valentinian found under the flagstone floor, so both must post-date the 380s. If the timber buildings replaced the function of the south granary, whose roof collapsed some time after the worn coin of Theodosius (388-395) was dropped on its floor, they also post-date this collapse.

Tony Wilmott suggests that the south granary roof collapsed in the early fifth century, say around 420, as the coin of Theodosius was very worn and thus had been around for some considerable time. If the first timber building on the north granary site was constructed around this time, and if each phase had a life of 50 years, this would suggest the second timber building was constructed around 470 and lasted until around 520 (Wilmott 2001 p. 123-4).

Needless to say, these are very approximate estimates. Timber buildings can and do last a lot longer than 50 years if the structural timbers are supported on stone foundations (as both the Birdoswald buildings were), as witness the medieval and early modern timber-framed buildings that still survive in Britain today. Conversely, a timber building can have a life considerably shorter than 50 years if bad weather, accidental damage, poor construction or bad luck intervene, or if the owner simply decides he feels like replacing it. And although a collapsed building is unlikely still to be occupied, the reverse is not necessarily true; a building could be abandoned for social, political or personal reasons long before it actually falls down. The issue date of a coin can be identified accurately, and this gives the earliest possible date for its loss or deposition, but the actual date of its loss or deposition could be later, perhaps much later, than its date of issue, depending on how long the coin was in circulation (or in storage).

Furthermore, it's difficult to be sure whether a site was continuously occupied or whether it went in and out of use, if the periods of disuse were too short for archaeologically recognisable deposits to form, or if any such deposits were cleared at subsequent re-use rather than built over. Ken Dark reviews the same sequence of evidence from Birdoswald and proposes a break in occupation between the residential use of the south granary and the buildings constructed on the north granary. In this scenario, he suggests that the south granary was in use as a residential hall during the last decades of the Roman occupation (roughly 350 to 400), that the fort was then abandoned for a period after the end of Roman government, and that the fort was re-occupied some time during the later fifth or sixth century, at which time the site of the north granary was cleared and the timber buildings constructed (Dark 2002, p.198-199).

Another possibility that occurs to me is that there may have been a break between the first and second timber building phases. Since the second timber building was partly on top of the site of the first, the first must have collapsed or been demolished before the second one could be built. Demolition and building might have been quite quick, if all the required materials were assembled and pre-prepared on site beforehand, but that requires efficient logistics and even then the replacement of one building by the other could hardly have been the work of a day. Abandonment of the site, perhaps only briefly, allows for site clearance and the subsequent construction of the second timber building.

It is worth bearing in mind that abandoning the area of the granaries does not necessarily mean abandoning the entire fort site. Areas of occupation may have moved around according to the condition of existing structures and the availability of sites for rebuilding. A barrack-type building constructed in the late Roman period (later fourth century) had re-used an inscription from the commanding officer's house (praetorium) and another inscription commemorating the rebuilding of the granary, indicating that these buildings had gone out of use while other areas of the site were still occupied and in use.

Only the area around the west gate at Birdoswald has been excavated to modern standards, and it is possible that other early medieval structures may have stood elsewhere in the fort. Early excavations would not have picked up the slight traces left by timber buildings, and parts of the fort interior have not been excavated at all. Medieval ridge-and-furrow plough marks cover the southern third of the fort interior, with the exception of an area in the south-west corner where there is also a long flat platform against the fort wall. This has been suggested as the site of a medieval building (Wilmott 2001, p.133, 135); if so, perhaps the area might also have been in use earlier. The foundations of a square tower, possibly dated to the fourteenth/fifteenth century, were discovered in excavations north of the west gate (Wilmott 2001 p. 137). The west gate was probably still standing when the tower was built, as the pottery found under the rubble of its collapsed arch indicates it collapsed in the later fifteenth century, and Wilmott suggests that the first-floor storey of the gate was re-roofed and used as a hall building to go with the tower (Wilmott 2001, p. 138). If the west gate was sufficiently intact to be re-roofed and occupied as a hall in the fourteenth century, perhaps the same also applied several centuries earlier.


It's clear from the timber halls that occupation at Birdoswald continued beyond the end of Roman government in Britain, for an unknown period that was long enough to construct two successive large timber buildings on the same site. The size of the halls implies that whoever ordered their construction controlled substantial resources of materials and manpower. It's a reasonable inference that this was a local or regional ruler.

Although the halls are built on a Roman site, and the first one directly re-used surviving Roman walls as part of its structure, they are built of timber not masonry. This implies that Roman building techniques had been lost, either due to lack of knowledge or lack of the infrastructure to obtain the appropriate materials, or deliberately rejected.

The duration of post-Roman occupation is unknown, but must surely extend until at least well into the fifth century, unless the timber buildings had absurdly short lives, and could easily extend well into the sixth century or beyond, particularly if there was a break in occupation. The long cist grave found nearby is consistent with a burial of early medieval date, and is consistent with its occupant having been a Christian. One could speculate that the long cist grave represents the burial of a Christian ruler who lived in the timber hall, but this really is clutching at straws.

The eighth-century pin found nearby may indicate that the site was in use at this date, especially if the timber halls on the north granary site were succeeded (with or without a break in occupation) by other structures elsewhere in the fort, but this is unsubstantiated.

The place name, combining an Old English personal name with a Brittonic topographical description ("pen or farmyard of Oswald") is most likely to have been coined some time in the early medieval period in the centuries after the end of Roman rule (since no trace of the Roman name remains in the modern name), but before the eleventh century when Old English names largely went out of fashion following the Norman Conquest and perhaps more likely before the Norse/Viking influence of the ninth and tenth centuries. The absence of any trace of the Roman name of the fort, Banna, in the modern name is consistent with a period of abandonment at some time after the end of Roman rule, long enough for the name to be lost, followed by re-occupation some time in the early medieval period. However, it could also be consistent with deliberate replacement of the old name by someone who wanted to emphasise a change of ownership or political allegiance. Whether "Oswald" refers to the sort of individual who might have owned and feasted in the timber halls, or to a humbler farmer working the site as agricultural land, or anything in between, is anyone's guess.

On the whole, I would say that occupation until well into the sixth century is supported by the timber halls. This could represent continuous occupation without a break from the Late Roman period, or could represent reoccupation of a temporarily abandoned site. Further occupation into later centuries - with or without a break - is not unreasonable, but the only evidence for it is the eighth-century pin and perhaps the place name, which is not exactly definitive.

If Birdoswald was the base of an important local ruler in the later sixth century, this brings it into the same sort of period as the battle of Arthuret, entered in the Annales Cambriae in 573. Arthuret is traditionally located near Longtown in Cumbria, which is less than 20 miles from Birdoswald. I mentioned Arthuret in my article about Peredur. More about the battle in a later post.


Dark K. Britain and the end of the Roman empire. Tempus, 2002. ISBN 0-7524-2532-3.
Wilmott T. Birdoswald Roman fort: 1800 years on Hadrian's Wall. Tempus, 2001. ISBN 0-7524-1913-7.

Map links

Birdoswald - Streetmap
Birdoswald - Google Maps satellite image

Note for readers of Paths of Exile

I based the (fictional) warlord's hall at Navio Roman Fort (modern Brough, Derbyshire) in Paths of Exile on the second of the post-Roman timber halls excavated at Birdoswald. There is no evidence that Navio fort was occupied by a local warlord in 605/606 (nor is there any evidence that it wasn't).