The Saxon Shore Forts - a misleading title !!

During the Crisis of the Third Century (AD 235-284), the Roman Empire nearly collapsed under repeated foreign invasions, civil wars, and economic disintegration. Britain was not immune to this chaos. As central authority weakened and Saxon sea-raids began to increase along the eastern coastline, the Roman authorities constructed the first generation of coastal forts. Caister-on-Sea was built around AD 200, Reculver between 200-225 AD, and Brancaster between 225-250 AD. These were relatively modest installations, designed to provide early warning and basic coastal patrol capabilities against small-scale raiding vessels. They represented the Empire's reactive response to growing instability.

In AD 260, the western provinces including Britain broke away from Rome to form the Gallic Empire under Postumus, which lasted until AD 274. During this fourteen-year period, Britain operated independently with its own fleet, ironworks, oak shipyards, and coastal defences. This experience proved crucial: it demonstrated that Britain could function as a self-sufficient entity, controlling its own resources and defending its own shores without direct support from Rome. When Emperor Aurelian forcibly reincorporated the Gallic Empire back into Roman rule in 274 AD, the memory of independence remained fresh. The infrastructure of the Classis Britannica—the British Fleet—and its control of Wealden iron and Andredsweald timber had already been established. What remained was the question of who would control it.

In AD 285, Marcus Aurelius Mausaeus Carausius, a Menapian naval commander, was appointed commander of the Classis Britannica with orders to suppress Frankish and Saxon piracy. However, Carausius was a man shaped by the Crisis of the Third Century—he had watched the empire fracture and reassemble within his own lifetime. The accusation that crystallised Maximian's fury was that Carausius was intercepting pirates after their raids and retaining the seized cargo rather than surrendering it to the imperial treasury. But the scale of what was subsequently built suggests a different interpretation: the seized cargoes of iron, timber, coin, and trade goods were precisely the materials a major fortification project would consume. Carausius was not enriching himself personally; he was systematically redirecting intercepted wealth into a pre-planned construction programme.

Between the 270s and 290s, a massive second generation of stone forts was constructed: Richborough, Dover, Lympne (Stutfall Castle), Bradwell-on-Sea, Burgh Castle, Pevensey (Anderida), and Portchester. These were not the small wooden palisades suitable for deterring pirate raids. Pevensey featured walls 14 feet thick and 27 feet high with massive external artillery bastions—the longest oval enceinte in Roman Britain. Lympne enclosed 10-11 acres with walls 12-14 feet thick. This was overkill for Saxon raiders. These were fortifications designed to withstand siege engines, professional legions, and the full military might of the Roman Empire. The dendrochronological evidence from Pevensey places the felling of foundation timbers around AD 290, consistent with a construction programme that began when Carausius declared his rebellion in 286 AD.

The location of these forts reveals their true purpose. Pevensey sat at the entrance to the Andredsweald forest, with the tidal estuary reaching deep into the woodland, giving direct water access to the oak timber supply that built the fleet. The fort's name—Anderida—literally meant "the entrance." Lympne controlled the original estuary of the River Rother (the Limen) and the iron-producing Brede valley. Richborough and Dover guarded the Channel crossing points. Portchester controlled the western approaches. These were not random coastal defences; they were a coordinated network of naval bases, supply depots, and strategic chokepoints controlling Britain's industrial heartland. Carausius had inherited not just a fleet but an entire military-industrial complex, and he fortified every critical node.

In AD 289, Emperor Maximian assembled an invasion fleet to retake Britain. He built it inland on the Moselle and Rhine rivers using softwoods and light riverine timber, away from Carausius's reach. The result was catastrophic. Carausius's fleet, built from seasoned Wealden oak using heavier hull designs refined by the local iron industry, destroyed Maximian's invasion force. Roman panegyrists attributed the failure to bad weather, but the seven-year gap before the next invasion attempt tells a different story. Rome had lost not just ships but trained crews, experienced pilots, and navigators who understood Channel tides—specialists who could not be quickly replaced. The destruction of Maximian's fleet established Carausius's dominance so completely that Rome required seven years to rebuild a credible invasion force.

By AD 291, Pevensey was complete—genuinely impregnable provided it was not isolated by the loss of the fleet. In AD 293, after Constantius Chlorus captured Boulogne, Carausius's treasurer Allectus assassinated him and seized power. Allectus had waited patiently for the fort to be finished, then taken it. But he made the fatal error of treating the walls as a guarantee rather than as a resource. He held his fleet passively behind the walls instead of using it offensively as Carausius had. In AD 296, Constantius launched his invasion in two divisions. Using fog to slip past the Carausian fleet anchored off the Isle of Wight, the Romans landed on the Hampshire coast. Allectus marched out to meet them rather than holding the fort, and was killed in the engagement that followed. Pevensey, the impregnable fortress, was surrendered without a siege. Not a stone was chipped.

The term Litus Saxonicum appears only in the Notitia Dignitatum, a document usually dated to around the end of the fourth century — a hundred years after most of these forts were built. The name implies the whole system was conceived as anti-Saxon coastal defence, but the dating and scale of the southern, Channel-facing forts sit uneasily with that. The first-generation forts of the early third century probably were built as emergency anti-raiding measures. But Pevensey, Lympne, Richborough, Dover, and Portchester — built in the 270s–290s, sited to control the Wealden iron and timber supply, and scaled well beyond anything small raiding parties required — look, on the geography and the timing, like the infrastructure of a breakaway naval power holding the Channel against Rome rather than a defence against Saxon ships. This argument is confined deliberately to those southern sites. It is not extended here to Burgh Castle or Brancaster, which sit outside the Carausian/Allectan network and for which an anti-raiding role remains the better-supported explanation. The "Saxon Shore" label may be a misnomer for the Channel forts specifically, without being wrong for the system as a whole.

The fleet command itself likely did not survive AD 296 intact either. The Classis Britannica had just produced two successive usurpers operating from these forts, and Rome had every reason to dismantle so concentrated a naval command rather than hand it to a new admiral. The dispersed, garrison-based structure that appears later in the Notitia Dignitatum, under the Comes litoris Saxonici, looks like the likely result. That argument is developed at greater length on this site's page on the Classis Britannica boundary.

The British half of the Saxon Shore command comprised nine forts, listed in the Notitia Dignitatum under a single officer, the Count of the Saxon Shore in Britain. They range along the coast from Brancaster in Norfolk to Portchester in Hampshire, and span roughly a century of construction: Brancaster, Caister-on-Sea (or Burgh Castle — the two are disputed as the site of Gariannonum), and Reculver belong to the early 3rd century and look like a first-generation response to raiding rather than part of any later, coordinated system. The remaining six — Othona, Richborough, Dover, Lympne, Pevensey, and Portchester — were built or substantially rebuilt in the 270s to 290s, several dated by dendrochronology or coin evidence to the Carausian and Allectan usurpation specifically. It is this later group, concentrated on the Channel-facing southern coast and tied together by CL·BR-stamped tiles and shared construction methods, that this page's central argument concerns: that their function as a rebel fleet's naval infrastructure has been obscured by a label — "Saxon Shore" — that only appears in a document compiled after the fact, describing a command structure imposed once Britain was back under central imperial control.

Built around AD 225–250, Branodunum guarded the approaches to the Wash and is among the earliest of the nine British forts. Excavation has established a roughly square plan enclosing about 6.5 acres (2.6 ha), with walls some 10 ft (3 m) thick backed by an earthen rampart and fronted by a single V-shaped ditch. The Notitia names its garrison as the Equites Dalmatae Branodunensium, though a tile stamped "Cohors I Aquitanorum" hints at an earlier, different unit. As a first-generation fort, Brancaster sits outside the Carausian/Allectan argument made elsewhere on this site — its role looks more straightforwardly to have been anti-raiding.

The Notitia's single mention of Gariannonum, garrisoned by Stablesian cavalry, has never been securely pinned to one bank of the Yare/Waveney estuary rather than the other, and this remains a genuine, unresolved dispute among specialists rather than a settled identification. Caister-on-Sea's fort dates to around AD 200 and enclosed roughly 3.5 ha; Burgh Castle, on the opposite shore, was probably built later, around AD 260–280, with walls tapering from 3.5 m at the base to 1.5 m at the top and standing some 4.5 m high, enclosing around 2.4 ha. Both sites are treated in the excluded, first-generation category on this page rather than folded into the Carausian argument.

Othona is late 3rd century in style — probably contemporary with Dover, Lympne and Gariannonum — and garrisoned, per the Notitia, by the Numerus Fortensium. Its walls were substantial, around 4.2 m (14 ft) thick, but the site has suffered badly from coastal erosion: only the west wall (522 ft) and fragments of the north and south walls survive, so the true extent of the fort — originally at least 4 acres — can no longer be established. It remains the most poorly evidenced of the nine British forts, with no modern, in-depth excavation on record.

One of the earliest Saxon Shore forts, built around AD 200–210, Reculver's design has been compared directly to forts on Hadrian's Wall — unsurprising, since its Notitia-attested garrison, the Cohors I Baetasiorum (confirmed by tiles stamped "CIB"), had previously served at Maryport in Cumbria. The walls stood around 15 ft high, roughly 10 ft thick at the base tapering to 8 ft at the top. Much of the fort — including the entire north wall — has been lost to the sea.

Richborough's stone Saxon Shore fort, built around AD 275–290, replaced an earlier signal station built into the ceremonial arch marking the traditional site of the Claudian invasion landing. The walls are 11 ft thick and still stand up to 25 ft in places, enclosing around 6 acres. The Notitia names the garrison as Legio II Augusta, previously stationed at Caerleon — a full legion, unusual among the Saxon Shore garrisons, most of which were smaller numeri or cavalry units.

Dover's Roman history has two distinct phases: an earlier Classis Britannica base (from the 2nd century, rebuilt around 130–140 AD, over 2 acres), demolished around AD 270 and replaced by a later Saxon Shore fort. Much of the site now lies beneath the modern town, and the garrison specifically attested for the Saxon Shore phase isn't something I could confirm with confidence from the sources checked — several accounts blur the earlier Classis Britannica unit with the later Notitia entry. What is solid is the sequence itself, and Dover's place as one of the two best-attested CLBR bases in Britain, alongside Boulogne on the continental side.

Built in the 270s–290s, contemporary with the other southern, Channel-facing forts, Lympne enclosed 10–11 acres within walls 12–14 ft thick. The Notitia records its garrison as the Numerus Turnacensium, a unit raised at Tournai. Like Pevensey and Portchester, CL·BR-stamped tiles here point to Classis Britannica activity predating the Saxon Shore designation itself — part of the basis for reading these southern forts as continuous naval infrastructure rather than a sudden 4th-century anti-raiding response.

Pevensey's dendrochronological date of around AD 290 makes it one of the more securely dated forts in the whole system. Its walls, 14 ft thick and originally up to 27 ft high, enclose the longest oval circuit of any Roman fort in Britain. The Notitia's garrison, the Numerus Abulcorum, and the CL·BR tile evidence tie it into the same Carausian/Allectan naval network as Lympne, Richborough, Dover and Portchester — the core group this page's central argument concerns.

Coin evidence dates construction to after AD 268, and tradition (not direct proof) attributes the fort to Carausius, around 285–290. Portchester is the best-preserved Roman fort north of the Alps: a 9-acre square enclosed by walls 10 ft thick and 20 ft high, with 16 of an original 20 bastions still standing. CL·BR-stamped tiles again link it to Classis Britannica activity, reinforcing its place in the southern, naval-infrastructure reading of the "Saxon Shore" label.

Unlike Britain, the Gallic coast had no single unified command — the Notitia splits it between two duces, and this shorter northern stretch, covering Flanders and the Pas-de-Calais, fell under the Dux Belgicae Secundae. Only three units are named here — Marcae, Locus Quartensis sive Hornensis, and Portus Aepatiaci — a strikingly thin list given that this coastline also included Boulogne, the confirmed principal base of the Classis Britannica, and Oudenburg, an excavated fort with a long occupation history. Neither appears in the Notitia's text for this command at all, which is one of the more puzzling gaps in the whole document — scholars generally assume both fell under this dux regardless. Of the three named sites, only Marcae is explicitly called "on the Saxon Shore" in the text itself, and its location, like the other two, remains genuinely unresolved in the conventional literature. The identifications offered on this page — Bergues, Abbeville, and Pont-Audemer — are working hypotheses built from the CTL tidal model and are explicitly theories, not established fact; none currently has independent archaeological support, and each depends to some degree on the one before it.

This identification is a theory, not an established fact. The Notitia places Marcae explicitly "on the Saxon Shore," garrisoned by Dalmatian cavalry, but its location is unresolved in the standard literature — Marquise and Marck are the usual candidates, with Marck preserving the ancient name almost unchanged, which remains the strongest single piece of evidence for the conventional view. The identification with Bergues proposed on this page rests on the 4.5 m tidal hypothesis and the site's position midway between the confirmed forts at Oudenburg and Boulogne, rather than on any excavated or textual evidence at Bergues itself: no Roman occupation is recorded there in standard historical accounts, and the Digital Atlas of the Roman Empire carries no entry for the site at all. It may be that a genuine Roman station lies undetected beneath later sediment and building — or there may simply be nothing there. No one appears to have tested this either way.

This identification is also a theory, and it rests on a further, unproven assumption. The name plausibly derives from quartus, "fourth," implying a numbered sequence of coastal stations — but the Notitia itself lists only three units under this duke. Treating this site as the fourth requires assuming Oudenburg and Boulogne, though absent from the Notitia's own list, were nonetheless counted as the first and third stations in the same informal sequence — a plausible but unverified assumption that everything else here depends on. The documented silting of the Somme (some 12 km of recession since Roman times) is solid, independent evidence of a lost Roman-period harbour somewhere upstream of the modern coast; it does not by itself confirm that harbour was Locus Quartensis, or that it sat at Abbeville specifically.

This is the most speculative of the three continental proposals, since it depends entirely on the previous identification being correct. The conventional candidate, Étaples, sits comfortably on the coast between Boulogne and the Somme, and is displaced here only by the argument that Locus Quartensis occupies the numbered "fourth" position, leaving Étaples oddly placed as "third" in a sequence it shouldn't head. If that numbering argument doesn't hold, there's no remaining reason to prefer Pont-Audemer. The supporting detail — the Risle's tidal limit, the Nervian garrison, two associated Roman roads, and a speculative pont au de mer etymology — is individually reasonable but not independent confirmation.

Not listed in the Notitia, but archaeologically confirmed
Unlike the three sites above, Boulogne's status as the Classis Britannica's principal base rests on convergent, independent evidence: sailors' tombstones (one Claudio-Neronian in date), CL·BR-stamped tiles whose fabric analysis traces to clay sourced up the Liane valley near Boulogne itself, and an excavated fleet camp — located in 1967 beneath the Ville-Haute, roughly 400 × 303 m, with 1.8 m-thick walls and rectangular towers. The Tour d'Ordre, a Roman lighthouse traditionally dated to Caligula's invasion preparations around 39 AD, stood on the cliffs above the port until it collapsed into the sea in 1644. Boulogne's absence from the Notitia's Gallic list is a genuine puzzle, but its Roman military importance is not in doubt in the way Bergues's is.

Not listed in the Notitia, but archaeologically confirmed
Along with Boulogne, Oudenburg is one of the two continental sites scholars generally assume fell under the Dux Belgicae Secundae's command despite not appearing in the Notitia's own text for this command. I haven't yet gathered the same level of excavation detail on Oudenburg specifically as for Boulogne, and would want to check its archaeological record properly before writing more than this placeholder paragraph.

The longer, western command covered the whole Norman and Breton coastline, from the Seine to the Loire, and lists ten units under the Dux Tractus Armoricani et Nervicani — twice as many as the Belgic command, reflecting the greater length of coast and, arguably, the greater administrative reach of this dux's authority inland. The list mixes genuinely distinct kinds of site: purpose-built coastal castella comparable to the British forts (Alet, Brest, and Vannes, each with real excavated defences and reasonably firm dates), and major existing cities — Rouen, Coutances, Avranches, Nantes — that simply had garrison units attached to them, several in direct response to the crisis decades of the 270s. One entry, Grannona, is named twice in the list (or possibly duplicated in error), and its location is among the least resolved in the entire document — candidates span the length of Normandy and Brittany, and at least one proposed site has been excavated and found to contain no Roman trace at all. Two further entries, Blabia and Benetis, rest on tentative modern identifications (Hennebont or Port-Louis; Vannes) that carry real uncertainty of their own. Taken together, this command illustrates, on the continental side, the same point argued for the British forts: that "Saxon Shore" as a single descriptive label conceals what was in practice a patchwork of different site types, built at different times for different immediate reasons, only later gathered under one administrative heading.

Port-en-Bessin, Granville, Guérande, Graignes and Guernsey have all been proposed) Grannona is named in the Notitia as being explicitly "on the Saxon Shore," under the tribune of the cohors prima nova Armoricana — making it the western counterpart to Marcae as one of only two Gallic sites given that designation. Unlike Marcae, though, Grannona has attracted a genuinely wide scholarly spread of candidate locations along the whole Norman and Breton coast, and at least one candidate (Guérande) was excavated as early as 1894 and found to contain no Roman-period trace at all — a rare case where a proposed identification has actually been tested archaeologically and come up empty. This is one of the least-resolved place-names in the entire Notitia, on either side of the Channel, and nothing currently marks any candidate as more than a guess.

Rouen (Rotomagus) was the metropolitan seat of Lugdunensis Secunda and had an established bishopric by 314 — evidence of major, well-organised city status rather than a purpose-built military site. I haven't been able to confirm specific fortification dimensions for Rouen's Roman-period defences from the sources checked; what's clear is that it fits the general pattern across late 3rd/4th-century Gaul in which existing cities took on reduced, defensible circuits rather than being purpose-built forts, and that the Notitia's garrison here — milites Ursariensii — sat within an already-established urban centre rather than a new coastal installation.

Coutances took its Roman name, Constantia, from Constantius Chlorus around AD 298, replacing the earlier name Cosedia — itself a marker of the town's elevated status under the Tetrarchy. The Notitia records both a praefectus militum primae Flaviae and a praefectus Laetorum Batavorum et gentilium Suevorum stationed here — Batavian and Suebian laeti (settled barbarian troops), alongside Bayeux, which points to a garrison built around irregular settled units rather than a conventional legionary or auxiliary force. An older (1823) source describes the town's fortifications as "quite demolished" even by then, alongside surviving traces of a Roman aqueduct — but I don't have modern excavation data on wall dimensions or dates and would flag this as needing a specialist source before publication.

Avranches (Legedia on the Tabula Peutingeriana, later Abrincae) prospered as a Gallo-Roman trading centre until Saxon raids from around AD 280 disrupted the region. Local heritage sources point to Roman-period traces surviving in later structures — the buttresses of the medieval keep, a tower of the Porte Baudange, the Tour de l'Arsenal — but these are popular/heritage attributions rather than confirmed archaeological findings, and I'd treat them with real caution rather than repeat them as established fact. The Notitia's garrison, milites Dalmatae, is the only specifically military detail that can be stated with confidence.

This is one of the better-documented Armorican sites. The Gallo-Roman city of Alet, former capital of the Coriosolites, was enclosed by a rampart built around AD 270–280 — 3–4 m high, following the clifftop for close to 1,400 m, reinforced with square towers. The settlement of the milites Martensii named in the Notitia is generally dated to the later 4th century, on the strength of associated finds — meaning the garrison arrived roughly a century after the walls themselves were built, which is worth noting if you're drawing parallels with the British forts' construction-versus-garrisoning timelines.

This is a real, unresolved dispute, not a settled identification with a popular alternative. Most researchers favour Brest, where a substantial late-antique castellum survives (in part) beneath the modern Château de Brest: a trapezoidal fort with walls in opus mixtum construction, one side around 200 m long, and roughly twenty hollow towers — possibly designed to carry ballistae. Others prefer Carhaix, the former Osismian capital (Vorgium), on the grounds that it may have retained the tribal name after losing administrative primacy to Brest — but Carhaix has produced no defensive remains at all and sits too far from the coast to plausibly serve a Saxon Shore-type coastal defence role, which weighs against it. The Notitia's garrison, Mauri Osismiaci (Moorish troops), fits either location without helping to resolve which one is meant.

As noted previously, the location is genuinely uncertain between these two candidates, garrisoned by the milites Carronensii. I wasn't able to find further excavation detail for either candidate site specifically in the searches carried out for this page, and the additional complication — that "Blabia" has historically been confused with Blaye on the Gironde, a wholly different site — is worth keeping as a caution against over-confident identification here.

Vannes has a well-documented late Roman defensive phase: at the end of the 3rd century, under Probus, a castrum was built on the Méné hill to the west of the existing Gallo-Roman town, with a triangular wall enclosing roughly 12.5 acres (5 ha) and a perimeter under 1,000 m. This is one of the more solidly dated and measured Armorican sites, comparable in documentation quality to the British forts. The Notitia's garrison, milites Mauri Beneti — again Moorish troops — fits the same pattern of deliberately non-local garrisoning seen at Osismis and elsewhere in this command.

Nantes' Gallo-Roman wall was begun after AD 276 (dated by milestones bearing the name of the short-reigned emperor Tacitus, found reused in the foundations) and probably completed under Probus. Before this, the city — then occupying the modern Bouffay quarter — had been undefended, and the wall was a direct response to the Germanic incursions of the 270s. The Notitia's garrison, milites superventores, again suggests a city with an attached military unit rather than a purpose-built shore fort, echoing the pattern at Rouen.