“For from the time when Rome cast its net to the ends of the earth, it rearranged and transformed all in a single polity, and compelled all to be tributary to a single yoke. But when the net began to decay and split, the state of affairs and the names also changed.” (trans. Haldon 2021, p. 166)1

This sombre judgement on Rome’s failure occurs in a 10th century description of the military dispositions of the Eastern Roman Empire compiled by the emperor Constantine VII Porphyrogenitus in the preamble concerning the Balkan themes (military commands). The text of De Thematibus can be viewed more as a reflection of Byzantine attitudes to the Roman past than a literal account of contemporary dispositions, unlike another of Constantine’s texts, the De Administrando Imperio, concerned with the contemporary political geography and diplomacy of the Empire’s neighbours.

The region may be defined as a wide triangle of lands, limited by the Danube in the north, the Dinaric Alps to the west, and littorals of the Black Sea and the Aegean to the east and south, with some reference to the wider Black Sea; in effect the late Roman dioceses of Macedonia, Dacia and Thraciae (Fig. 1). The creation of the new capital city of Constantinople, located in the province of Europa, came to dominate the economic, cultural and above all strategic geography of the region, perhaps even more than the provinces of Asia Minor to the east. From the time of the Roman defeat and death of the Emperor Valens at Adrianople in 378, remedied by the construction of the city’s Theodosian land walls, the vulnerability of the capital became apparent. A crucial development in the 6th century was the decision under Justinian to link the south-western coastal provinces of Asia Minor and the islands of the Aegean with the lower Danube garrisons to create a new administrative circumscription under the command of the quaestor exercitus.2 While this move can be seen as a sign of increased connectivity it seems also to be a response to generations of barbarian incursions and devastation to Danubian frontier territories of Moesia I and Scythia Minora. But so long as Constantinople maintained supremacy in the Black Sea it did, however, ensure a strategic advantage over control of the river frontier, even when the riparian garrisons were weakened or dispersed.

The lower Danube under Anastasius and Justinian

The first half of the 6th century witnessed imperial-sponsored programmes of defence and refortification on the Danube ripa and its hinterland on a scale not seen since the tetrarchy and reign of Constantine the Great. Starting under Anastasius and continuing with Justinian, inscriptions and written accounts, notably Procopius’ The Buildings book 4, present a picture of extensive investment, not just on the river frontier (ripa) but across the interior of the provinces as far as the suburbs of Constantinople. The text of Procopius has been frequently discussed3 but there is still no convincing attempt to explain how the numbers of toponyms known in the text map on to the archaeologically defined defences known across the region, let alone explain their role. Needless to say in the absence of clear diagnostic archaeological evidence derived from excavations the default mode from many historians and archaeologists is to uncritically attribute a Justinianic date on the basis of Procopius’ testimony. Along the ripa fort sites known since the principate had been refortified in the time of Diocletian and Constantine with further restoration under Valentinian and their legacy of high walls and massive towers of strongholds like Iatrus or Dinogetia sustained the 6th century garrisons. Set back from the river were a series of small heavily defended towns such as Abritus or Zaldapa (possible birthplace of the usurper Vitalian, or his father).4 These were foundations of the tetrarchy or like Tsaravets replaced the Trajanic foundation of Nicopolis ad Istrum which itself was partly re-fortified and seems to have retained largely a military role. In this instance greater security was found at nearby and naturally fortified Tsaravets at Veliko Turnovo, the site of the medieval capital of the second Bulgarian kingdom, identified with Zikideva. In addition to what Slobodan Ćurčić has described at the late antique ‘miniature’ cities,5 there were also numbers of rural defences, in some cases set on hilltops often succeeding earlier Iron Age settlements. The character of these sites remains enigmatic. Many display sophisticated defences including double walls or pointed towers characteristic of the major new urban fortification of the 5th and 6th century such as Thessaloniki or Refasa. Most of the rural sites are situated in the foothills or passes of the Haemus mountains (Stara Planina) which separate the Danube plain from Thrace to the south. See the recent distribution map published by Rizos,6 note however there are more examples in the eastern Stara Planina south of Turnovo (shown as Zikideva) than he shows. Some are situated overlooking passes on the south side of the watershed such as Sayganskoto west of Kotel with angled towers and extensive brick construction.7 Both from the design and construction techniques using hydraulic mortar and brick this and other phrouria reflect centrally planned and executed imperial works.

One unusual site has been subject to detailed excavation and publication at Dichin, west of Nicopolis ad Istrum on level ground close to the river Rositsa. It does not secure a kleisura or pass, but is a well fortified enclosure containing dwellings and in its first phase granaries.8 The defences again reflect military involvement, brick and mortared stone walls with a double wall along one side, another characteristic of the ‘new age’ defences at Constantinople and elsewhere. The assemblage of finds including cow-bells and scythes points to a soldier-farming community, the excavator suggests Gothic foederati. But a main role in phase I (before the destruction of c. 500) was the storage of grain, presumably for garrisons of the limitanei on the ripa, but also for field army units who might utilize the late defended enclosure at Nicopolis nearby. There is clear evidence for destruction by fire c. 500 and the interior was cleared and rebuilt with less provision for grain storage. As an indication of the workings of the quaestor exercitus it is only from the second half of the 6th century that imported finewares and amphora reappear in the stratified assemblages.9 The coin series ends around 580 when the settlement was abandoned, the excavator notes there was little of the debris associated with the earlier destruction deposits.10

This pattern of new fortifications, some attesting a military presence or role, extends further south and west of the Haemus mountains. The main cities strung along the strategic diagonal road leading from Singidunum (Belgrade) on the confluence of the Danube and the Sava rivers across the south-east Balkans to Serdica (Sofia), Philippopolis (Plovdiv) and Adrianopolis (Edirne) towards Constantinople, all received major fortification throughout late antiquity. Notably Serdica had an inner circuit with brick walls and triangular towers,11 a feature also found at Philippopolis and Augusta Traiana—Beroe (Stara Zagora).12 Set back from the diagonal road was another new foundation Justiniana Prima (Caričin Grad) south of Niš, a heavily defended “model city” at Justinian’s birthplace.13 Intended as a new regional administrative foundation set back from the Iron Gates the city displayed a tightly defended core, with an outer walled area, enclosed by earthworks set in a landscape surveyed by watchtowers.14

Defence across the Balkan provinces and the hinterland of Constantinople was not restricted to those provinces bordering the Danube (Fig. 1). Throughout the interior a sequence of barrier walls controlled strategic passes and pinch points. These included the Isthmian walls at Corinth restored under Justinian, the walls at Thermopylae, a barrier wall at the pass south-east of Serdica, the wall across the Gallipoli peninsula, a long wall and forts in the eastern Stara Planina west of Obzor, and the longest of these walls, the Long Walls of Thrace, situated 65km west of Constantinople and stretching between the Black Sea and the Sea of Marmara, at 58 km it is as long as the 2nd-century Antonine Wall in Scotland (Fig. 2-3).15

The Invasion of 559

Events in 559 offer important insights into the Roman defensive strategy into the second half of the 6th century, the challenges of long-distance raiding and the internal defences. The two major Hunnic groups, the Kotrigurs and Utrigurs and the new Kotrigur leader Zabergan invaded across the frozen Danube in March 559. The invaders split in three directions, the first towards Macedonia and Greece where it was halted and repulsed at the Thermopylae Wall. The second force made for the Chersonese Wall at Gallipoli which also stood firm and where an attempt to outflank the barrier was defeated by a naval force; after heavy losses the Kotrigurs withdrew to Moesia and the Danube. The third force under Zabergan headed towards the suburbs of Constantinople and was able to penetrate through the Long Walls which according to Agathias:16

Age and neglect had in fact caused the structure of the great wall to crumble and collapse in many places. Some parts of it the barbarians themselves knocked it down, setting about the task with the air of demolishing their own property. There was nothing to stop them, no sentries, no engines of defence, nobody to man them. There was not even the sound of the dog barking, as would at least have been the case of a pig-sty or a sheep-cot. (trans. Frendo 1975, p. 148)

Despite the failure of the outer defences of the city Justinian was able to call on his old general Belisarius to outwit and defeat the Kotrigur threat.

Other sources, notably Theophanes, specifically note that:17

Having discovered that some parts of the Anastasian Wall had collapsed from the earthquakes, they got in and took prisoners as far as Drypia and Nymphai and the village of Chiton. Everyone fled with their possessions into the city. On being informed of this, the emperor conscripted many and sent them to the Long Wall. They engaged the enemy there and many Romans, especially the scholarii, were killed. Then the emperor ordered that the silver ciboria and silver altar tables that were outside the city be removed while the scholae, the protectores, and the numeri, and the whole Senate guarded all the gates of the Theodosian wall. (trans. Mango et alii 1997, p. 341)

From Theophanes’ source it is clear that part of the Wall had been damaged by the great earthquake of December 557 which also weakened the first dome of Hagia Sophia. It was this event that allowed the bands of Kotrigurs to penetrate it, who clearly had little or no expertise at siege warfare. After Belisarius’ successful counter-attack the Kotrigurs were forced to withdraw beyond the Long Walls to western Thrace and eventually under the threat of the Danube fleet to retreat across the river Danube. Theophanes’ account also provides evidence for the role of the praesental units in defence of the city.

Almost immediately the emperor responded with a programme to repair the damage to the Long Walls which commenced after Easter (11 April 559). Justinian had rarely left his capital over the course of his long reign but the recent failure of the Long Walls was sufficiently grave for him to move his court to Selymbria (Silivri), to oversee the reconstruction works. On the completion of the works to the Long Walls, Justinian returned on 11 August in triumph, a text preserved in another work by the emperor Constantine Porphyrogenitus.18

The chronology of this invasion and the imperial responses are unusually precise. The Kotrigurs crossed the frozen Danube early in March. The two columns heading for Greece and for Gallipoli (with the potential to cross the Dardanelles into Asia Minor) were repulsed. The barrier walls were a sufficient barrier and effectively manned with the additional support of naval forces in the latter instance. There is only limited evidence about the garrisons, but like the Thracian Long Walls, it is likely to have been a combination of a local force with elements from the field armies of Thrace and Illyricum. In the approaches to Constantinople the Kotrigurs were able to bypass the strong points such as Adrianople (Edirne) or Heracleia (Marmaraereğlisi) but having broken through the ruined Long Walls the Kotrigurs faced strong resistance and withdrew back to the ripa where they were harried by the Danube fleet. We do not know where Zabergan crossed the Danube, but later in the 6th century Theophylact Simocatta records that the journey from Heracleia (less than a day’s distance from the Long Walls) was 24 camps to Dorostolon (formerly Durostorum, modern Silistra), so just over three weeks seems a reasonable estimate.19 As a recognition of the continuing security risk in February 562 seven units of the scholae stationed in north-western Asia Minor were transferred to Heracleia and surrounding cities in order to bolster the western defences of Constantinople against future assaults.20 Significantly the provincial garrisons do not figure in these accounts which is largely due to the character of the sources with a strong Constantinopolitan bias. However, it is worth pointing out that where there is evidence from excavations at ripa forts such as Iatrus; the internal structures dateable to the second half of the 6th century appear to reflect significantly reduced occupation indicative of a smaller garrison and decreased strength.21 Changes of a similar nature, or at least with similar implications, to the internal arrangements at Dichin have already been noted.

After Justinian

After Justinian’s restoration of the Long Walls and the defeat of the Kotrigurs, a new pan-Eurasian presence appeared north of Danube, the Avars who replaced the Gepids and were concerned with the north-eastern Balkans.22 The Avars initially posed no significant threat to Moesia and Thrace. In reviewing the situation of the Balkan frontier at the time of Justinian’s death in 565, Lemerle argued that the frontier remained on the Danube, although the fortified places played little role, in comparison with the significance of the river fleet. Despite a number of raids and invasions few cities were captured, since most of the raiders were unsuccessful at besieging cities, he concluded “Les armées byzantines, en rase campagne, ne les effraient pas; mais une muraille les arrête”.23 Here it is important to differentiate between the Eurasian peoples like the Hunnic Kotrigurs and the Avars who fought in coherent large war bands under single leaders with whom the Romans could negotiate and bribe, and on the other hand the Sklaveniai who operated in smaller tribal units who were often allied with Huns or Avars but independently presented the imperial forces with an asymmetrical opponent with differing ambitions.24

It was only when the Eastern Empire became preoccupied with the eastern frontier after 572 that the Avars themselves became a more serious threat, but not until 577 is there any record that the Thracian Long Walls were attacked when an Avar force was stopped at the Wall (other walls like Gallipoli and Corinthian Isthmus no longer figure in chronicle accounts). In 583 having been refused a huge rise in subsidies from the Empire, the Avars destroyed Sirmium (Sremska Mitrovica), one of the critical strongholds of the Western Balkans and other cities,25 and went on to capture Anchialos on the Black Sea coast from where the Chagan declared to Roman envoys that “he would destroy the Long Walls as they are called”. In the following year the Avars loosed their Slav subjects to pillage Byzantine lands as far as the Long Walls and the emperor Maurice personally organised their defences. In another passage (possibly 587) Theophylact Simocatta states that the Chagan released his army through the borders of Messembria and slaughtered the forces of 500 who were guarding the strong points. Most probably an indication of those units stationed in the kleisurai such as Sayganskoto kale noted above or perhaps the forts associated with the so-called Haemus Gates.26 Throughout this decade it is evident that raiding continued deep into Thrace up to the Long Walls barrier. Although Singidunum was lost other cities including Justiniana Prima and the cities from Serdica eastwards along the major routes remained secure behind their walls. On the Danube at Viminacium (Kostolac) there is evidence for destruction at this time but also Roman recovery in the 590s indicated at the nearby site of Svetinja where there are remarkable finds including lamellar armour.27 Further resistance is shown by Theophylact’s report when Diocletianopolis and other cities remained secure with their artillery defences (Fig. 4).28

Peace on the eastern frontier allowed Maurice to campaign vigorously up to and beyond the Danube against the Slavs and Avars and the history of these campaigns is recorded in some detail by Theophylact.29 Although the historian’s account may display bias towards the commander, the patrician Peter, his campaign in 594 provides valuable insights into aspects of the frontier. The field army was active along and beyond the Danube including mention of Novae, Iatrus and Zaldapa.30 One anecdote concerning the fort at Asemus is of note. Like Iatrus it is described in the history as a city and it was located close to the Danube west of Novae by the Osom river.31 An earlier historical mention in the mid-5th century recorded how the fort and the garrison described by their toponym as Asimountioi had most effectively resisted Attila’s Huns.32 Theophylact’s account one and a half centuries later presents the survival of Roman military practices but also a singular military community which prised its security over the demands of the distant state. Given its significance and the insight it gives of the relations between the field armies and the residual frontier garrisons and their communities it deserves quotation:33

On the third day he (Peter the general) established his quarters at the city of Asemus. But when the inhabitants of the city had learned that the general was expected, they came out of the city to meet Peter, and made his arrival at the city splendid. From bygone times a garrison had been organised in this city for the protection of the citizens, since the barbarians swooped down like lightning around this city quite frequently. Accordingly, when the garrison stationed in this city learned that the general was about to arrive, they took up the standards, which Romans call banda, and went out of the city; then, arrayed in armour, they welcomed the general most gloriously. And so Peter, on seeing the magnificence of the city's soldiers, attempted to remove them from the city and include them amongst his own forces. And so the citizens and the city's garrison produced a decree of the emperor Justin which granted the city this successive armed protection. On the morrow the commander made objection and hastened to remove from the township those posted for its protection. For this reason the soldiers in the city took refuge in the city's church. (Whitby, Whitby 1986, p. 217-218)

The account continues with an attempt by the Patrician to demand that the soldiers are expelled from the church but this fails, the bishop is forcibly brought before the commander but the citizens resist further and close the gates of the ‘city’ ensuring the safety of their garrison, followed by acclamations to the emperor Maurice, who was Peter the Patrician’s brother! The general and the field army depart in disgrace.

No bishopric is otherwise attested at Asemus but recent research has identified the site of the late Roman fort, on a hilltop overlooking the river Osam; there have been no excavations.34 Whilst details in Theophylact’s account cannot be corroborated the maintenance of a strong military community and the tensions with central authorities present comparisons with other frontier provinces in the final decades of Roman control from Noricum to Britannia.

After a period of calm in 598 the Avars once more gained the ascendancy and again threatened Constantinople, but the Thracian Long Walls ensured the security of the capital’s inner hinterland and the emperor Maurice himself led troops from the city with the circus factions up to the walls as the defeated troops of the field army retreated. The Avars withdrew in part due to an outbreak of plague and payment of a large subsidy, but as with other reverses before Constantinople, there was a consequent loss of prestige amongst their allies and in subsequent years Maurice and his generals were able to embark on a programme of aggressive defence against both the Slavs and Avars.35 A contemporary military handbook, the Strategicon, recognised the advantages of campaigning against the Slavs in winter when there was less cover in the forests and in 602 the emperor ordered Roman armies to overwinter beyond the Danube. Unlike in the time of Nero when Corbulo’s legions endured the hardships of an Anatolian winter near Satala in the upper Euphrates, six centuries later the trans-Danubian Roman armies mutinied at such a prospect and Phocas led a murderous coup against Maurice and his family. Unpopular in Constantinople, Phocas faced a further coup directed by Heraclius and was deposed in 610, but, during his reign, the Empire experienced continuing Avar and Slav raids across the Balkans and most seriously the breakdown of the treaty with the Sasanians which initiated a long and draining war in the east to last for nearly three decades.

The Avar siege of 626 and the end of Danube ripa

The final assault on Constantinople in 626 by the Avars again ended in failure despite mobilising siege engines and potential collaboration with Sasanian forces on the Asiatic side of the Bosporus. On this occasion the Long Walls were no longer mentioned in the written accounts of the siege. A recently discovered inscription attests the restoration of the wall under Heraclius directed by the patrician Smaragdus, a former exarch of Ravenna. This restoration probably followed Heraclius’ nearly disastrous ambush by the Avars in the so-called “Avar surprise” probably in 623 when Heraclius had returned from his first campaign against the Persians. On this occasion Heraclius’ aim was to secure a treaty in the Balkans to ensure security for his deep strike against the Sasanians in the heartlands of their empire. The disaster at Heracleia and the Long Walls meant he had to pay off the Avars with 200,000 solidi in order to redeem previously captured citizens.36 By the time of the siege of 626 the emperor was campaigning in the Caucasus with troops drawn from across the Empire, probably including the Thracian field armies and the defences of the city were too stretched for any effective defence of the Long Walls. The year 626 is a pivotal year for the East Roman world and marks the end of the era of deep defence across the Balkans based on barrier walls. It also marks the end of the known coin series for many of the Danube sites. For smaller forts such as Iatrus the cut-off appears to be around 600, at Dichin the last coins are in the 580s but for the fortified towns like Zikedava it is the 620s, similarly at Justiniana Prima, although this follows evidence for major destruction recognised on the site.37

Within a few years of Heraclius’ triumph over the Sasanians both empires confronted a new aggressive enemy from Arabia, with the armies of the followers of Muhammad. Persia was crushed and the Romans withdrew from Egypt and Syria up to the eastern mountains of Anatolia. In this context we still recognise the presence of the Thracian field army outside the Balkans when the magister militum per Thracias Marianus was defeated fighting Arab armies in Egypt, probably dated to 639.38 Locating the field armies remains a problem in late antiquity and beyond, especially with the decline in the epigraphic habit in comparison with the wealth of inscriptions from the earlier imperial age. Recent study of finds from destruction levels at Aphrodisias identified bronze fittings identified as stirrups. These deposits are dateable to Sassanian raids on the west coasts of Asia Minor in the 620s.39 Stirrups are also known from Avar contexts but these examples studied from Aphrodisias and elsewhere differ from the recognised Avar forms. Significantly the closest comparisons are from excavations in the Balkans at late 6th century contexts at Justiniana Prima and other Bulgarian sites including Karasura and Krasen both south of the Stara Planina.40 The innovative use of the stirrup can be associated with heavy cavalry and serves as a proxy for the Roman comitatenses. Another proxy is expertise in torsion artillery, as noted before Theophylact specially mentions the use of artillery in the defence of Diocletianopolis against the Avars and notably at that site excavation has revealed buildings constructed against the inner curtain wall, interpreted as barracks, a rare example in an urban site (Fig. 4), although the precise distinction between civil and military becomes blurred throughout the period.

How far is it possible to define Constantinople’s military presence in the southern Balkans?

The failure to besiege Constantinople in 626 and consequent loss of prestige led to increasing weakening of Avar power throughout the Balkans.41 With the sudden loss of the eastern provinces to the Arabs, Constantinople was forced to turn towards the east for much of the rest of the century, and apart from sources such as the Miracles of St Demetrius there are fewer references in the surviving chronicles to the empire’s Balkan territories. The Miracle texts continue to demonstrate settlement and raiding by Sklaveniai into the 7th century and most maps illustrating the period reflect the almost total loss of territory apart from eastern Thrace and the coastlands of Macedonia.42 A key text is the reference in an account of the exile and trial of the theologian Maximus the Confessor in 655. Maximus was held in the kastron of Bizye (Vize) (Fig. 5) not far from the source of the long-distance aqueduct for Constantinople. His disciple was held at nearby Perberis which was described amongst those places where “the Roman Empire does not extend one step beyond” (trans. Allen, Neil 2002, p. 180, n. 54).43 Frustratingly the latter toponym cannot be located with any certainty as it is rarely attested, although it appears in other contemporary writings associated with Maximus’ trials for heresy, and one wonders if it is an alternative for the coastal defended town of Salmydessos (Midye/Kiyiköy) (Fig. 6) which is north of Bizye, with another place of exile noted at Mesembria (Nessebre) further north along the Black Sea coast. In the following year Maximus was taken to Selymbria (Silivri) (closer to Constantinople) and there is reference to a unit stationed in a camp nearby described as phossaton, the Greek derives from fossatum and the term was used for earthworks, a rare reference to the location of a camp outside a town.44 Importantly this shows that the field army still employed encampments outside towns and also reflects the small size of many of the Thracian towns. The strategic hub of Adrianople was only 36 hectares; comparable to Diocletianoplis at 30 hectares,45 and significantly smaller than the greater circuits of Philippopolis or Beroe.

How far does the apparent precarious situation of ‘Perberis’ define the limits of Roman control in the mid-7th century or is it simply a perspective on the hazards which lay to the West? Contrary evidence derives from near contemporary bishops lists. For the Balkans and for Asia Minor the list of bishops before the 7th century crisis can be expected to correspond to the known lists of cities such as Hierocles Synekdemos. But in later lists for both regions there are clearly significant lacunae which in the east may be interpreted as reflecting the impact of Arab raiding on urban centres.46 Notitia I dating to the monothelete council of 662 includes sees of Marcianopolis on the Danube and its hinterland as far west as Novae and Zikedava, and in the province of Thracia with Philippopolis and Diocletianopolis. Significantly, Dacia Mediterranea and Serdica, located to the West on the military road, are not represented.47 How far the records of ecclesiastical administration map on to military dispositions is difficult to assess since for Moesia II and Scythia there is little specific numismatic or ceramic evidence for any continuing military presence after the 620s and the evidence suggests that direct control was limited to the Pontic coast especially at Odessus and Mesembria48 whilst inland Slavic tribes remained as Roman foederati. Of equal significance was the crucial diplomatic and military outpost at Cherson (Sevastopol) in the Crimea which ensured Constantinopolitan dominance over the Black Sea.49

If the major fortified Thracian cities fall below the Constantinopolitan literary radar, the ability of the state to retain and redefine control in parts of the Balkans is demonstrated through building and restoration. More than a century after Maximus’ verdict on Roman failure, later building works and restoration at strategic cities implies that Constantinople retained contact with certain urban centres. Serdica was the most western stronghold on the great military road linking the Illyricum with Thrace and Constantinople. Among its most interesting early Byzantine buildings is the church of St Sophia (set outside the walls) and a re-analysis of its structure suggests either major construction or at least reconstruction in the 8th century.50 Restoration of the walls by the emperor Nicephoros is attested in 808, before the capture of the city by Krum in 811. Nothing is known (or reported) archaeologically from the major city in the Thracian plain at Philipopolis, but following an imperial campaign Beroe was renamed as Irenopolis in 784 and an inscription of the same date records the restoration of the city wall.51

The Bulgars, a new steppe people and a new boundary

North of the Stara Planina the dynamics between the Sklaveniai and Constantinople changed in the late 7th century with the arrival of another steppe people, the Bulgars. Constantine IV campaigned against them, but in 681 a treaty recognised their presence south of the Danube in the region of Odessus and soon after at Pliska what was to become their first capital in an open plain suitable for the meetings of a semi-nomadic people. In the later 10th century the Byzantines were able to recover the lower Danube as a border but for over three centuries the Bulgar state dominated the region to the north and south of the river. The Roman state that survived the 7th century crisis ceased to be a territorial empire defined by set borders and frontier garrisons. The old Roman order as defined by Procopius for Justinian’s Danube frontier ceased. But the Empire still possessed a populous and securely defended hub reaching out through sea and land routes to territories from the Caucasus to the Balearic Islands, but as Constantine Porphyrogenitus observed “the net had become too frayed and split”.52 The Bulgars learnt from their East Roman enemies, the palace at Pliska had a bath house, treaties were agreed and the former steppe peoples came to the Empire’s aid in the second Arab siege of Constantinople in 717-18.53 Conflicts continued throughout the 8th century. A successful assault on Pliska by Nicephoros was crushed on his withdrawal over the Haemus passes when timber barricades trapped Roman cavalry in narrow defiles. A decade later khan Krum failed to besiege Constantinople but ravaged the cities of Thrace. In the treaties that followed one of his successors, probably Omurtag constructed a great earth bank called the Erkesija from the Black Sea at Debeltos, south of the naval bases at Anchialos and Mesembria, and south-west across towards the Hebros river (Maritza). 130 km long it survives up to 3 m high and 16 m wide with a ditch on the south side.54 In his study of the emperor Maurice, Michael Whitby notes a reference in the eastern chronicle of Michael the Syrian to an earthwork near Adrianople and suggested this could be a predecessor to the Erkesija.55 In practice the reference is probably to earthworks around Adrianople similar to those documented beyond the stone walls of Caričin Grad56 rather than a long earth bank to contain Avars, Slavs or later Bulgars. But neither was the Great Bank or Fence a diateichisma or barrier wall like Thermopylae or the Long Walls of Thrace, rather it defined a boundary and was a discernible statement of the power of the Bulgar state and their territorial acquisition of what had been Roman.57 According to a Byzantine chronicle the treaty was sealed between the Christian emperor Leo and the pagan khan Omurtag with a sacrifice “who poured on the earth a libation of water, swore upon a sword, sacrificed dogs and performed other unholy rites”.58 For students of earlier Roman frontiers this account may have resonances of the suovetaurilia shown on Bridgeness distance slab from the Antonine Wall, but in Thrace “the names had changed”, as Constantine Porphyrogenitus observed.