C. M. Whittaker, "The Collapse of the Frontiers," chpt 6 from his Frontiers of the Roman Empire, Johns Hopkins University Press, 1994, pp. 192-242.

A story was told about the emperor Theodosius the Great, who was returning to the East in A.D. 391, after defeating the western usurper Magnus Maximus. At Salonica in Macedonia he found the land in turmoil because barbarians were attacking the inhabitants by night, but during the day they hid in the marshes. They were, says the story, "like ghosts instead of men." Theodosius's remedy was to select a band of five horsemen with spare mounts and to ride continuously around the countryside incognito, eating at country inns. At one such inn, where he was received by an old woman, he noticed another traveler who kept to himself and said nothing. Surprised by this, Theodosius asked his hostess who the stranger was. She replied that she did not know but that he had been with her for some time, going out all day and returning home at night exhausted. Theodosius thereupon seized the man, tortured him, and discovered he was a spy for the barbarians hidden in the marshes, giving them information about the movements of the Roman army and where they could attack. So the man lost his head and Theodosius found the barbarians.

I do not ask you to believe this tale of Zosimus (4.48), which has all the marks of folklore, not history. But it contains authentic characteristics Of life in the later Roman Empire, when the frontiers were coming under pressure from populations beyond. These particular barbarians were probably, deserters from the Roman army, concerning whom Theodosius had just issued an edict. But they could just as easily have been invaders. The extraordinary history of Charietto, the German barbarian raider, bears this out. Crossing the frontier in about 350, he lived at Trier with a band of robbers, but he also made profits by killing Germans who raided across the Rhine as they slept in the woods at night in a drunken stupor. So successful was he that the future emperor Julian employed him and his men as an irregular military unit, mixed with some Salian Franks Julian had just defeated, against invading Chauci and Chamavi- peoples who were themselves often counted among the Franks.

Charietto went on to hold a high command as a Roman general on the Rhine, which is extraordinary enough. But the important point here is that dealing with the invasion was, as Ammianus the soldier knew, often better done "through undercover piecemeal action and through banditry" (particulatim perque furta magis et latrocinia; Amm. Marc. 31.7.2). The word "banditry" (latrocinium) crops up again and again in our sources, as we shall see. It often refers to irregular, as opposed to formal, military operations. The disruption these raiders caused was selective but continuous. And yet the rural community coexisted with them and, in our story, could not necessarily identify them by dress or language. They were absorbed into the countryside "like ghosts." After the Goths had won their shattering victory at Adrianople in 378, they disappeared, says Themistius, "like a shadow" (Or. 16.206d). Many of them had been living in the Thracian countryside round about for several years (Amm. Marc. 31.6.5-6). Charietto apparently had no difficulty living with his bandits even in the territory of the imperial capital of Trier. The frontiers of the barbarians, whatever the Romans may have imagined, were in fact extending deeper into the Roman Empire.

It is this aspect of the social and economic reality of the frontiers in the later Roman Empire that I shall examine in this chapter. Indeed, the question we should ask is whether the concept of the collapse of the frontiers is not a by-product of stereotyped preconceptions about linear frontiers that I discussed in my first chapter. Conservative Romans, like Ammianus, may have viewed the invasions "like a river that has burst its banks" (31.8.5). But it would be interesting to know how the inhabitants of the frontiers viewed the process. Or for that matter, what German Franks and Goths, Arab Saracens, or African Austuriani, who were coming and going across the frontiers all the time, thought about it. One of the problems of our sources is that from the fourth century to the sixth they all come from the Roman side. Not a single text is written by a Vandal or a Burgundian or a Frank. The Saracens (or rather, various language groups of those Arabs we call Saracens) have left hundreds of inscriptions but little to tell us about their history. The Goths, some parts of whose Bible have survived, wrote nothing about their history until Jordanes in the mid-sixth century. And his history was in Latin, often following stereotypes of Roman tradition-as when, for instance, he describes invaders as only just human (quasi hominum genus; Get. 122).

Roman Ideology

Even if we will never know how most of the inhabitants of the later Roman Empire perceived events on their frontiers, it is worth looking more closely at two pronouncements from the latter half of the fourth century A.D. made by two more or less contemporary writers who were close to the court and the center of power. The first is Ammianus Marcellinus, who is often quoted because he wrote by far the best history we possess of his own day up to the Battle of Adrianople in 378: "At that time [he says], as if the trumpets were sounding the war note throughout the Roman world, the most savage people roused themselves and poured across the nearest frontiers" (26.4.5). The second is the writer of an anonymous tract called "Concerning Matters of War" (De rebus bellicis), who writes: "We must recognize that the madness of tribes baying all about are hemming in the Roman Empire and treacherous barbarians ... menace every stretch of our frontiers" (6.1). The impression these two writers give is one of chaos and pessimism, while Rome cowered behind its defenses.

Before we accept such gloomy descriptions, however, we must remember what was said earlier-that the classical Roman ideology of frontiers was based on the idea of expansion and of control of the barbarians beyond the provinces, which created an unwillingness to accept any other power as an equal. Ammianus, a soldier and minor imperial functionary, was classically conservative in his opinions and could only regard any change in the balance of power as a disaster. The nameless pamphleteer, probably a minor civil servant, has produced a strange document containing advice about the ills of the empire, but in military matters he is no more than a "genial dilettante." Both men were writing soon after the shock of Adrianople; both were entirely negative about the character of the "barbarian" pressures.

Nor does either author help us much to understand what was happening to the complex network of Roman limites in the later empire, with all the. historical differences that existed between the de facto frontiers of provincial administration, the military frontiers of control, and the political frontiers of influence. Although the empire had evolved over three centuries, it is uncanny how from the Roman upper-class point of view, which most of our fourth-and fifth-century sources reflect, nothing had changed. The main themes of Roman upper-class ideology persisted from the earlier empire with an even greater intensity. The Goths took on the traits that Scythians, Getae, and Sarmatians had formerly possessed. It may have been a kind of defiance, suggests one recent writer, who says: "The menacing shadow of the barbarian Goth allowed writers to reaffirm the grandeur of Rome in the face of the barbarian world".

Perhaps so. An anthropologist in Africa has noted that ideology tends to be at its purest on the frontier, where it is most under pressure. But my own impression is that many Roman attitudes from the days of imperial expansion were simply frozen into stereotypes, as so often happens when people are afraid to face change. It can be seen every day in the racist myths of our own world, and the charges are similar. Prudentius in the fifth century, like Velleius Paterculus in the first, still portrayed Germans as animals (Prud. C. Symm. 2.816-19; Vell. Pat. 2.117). The story of Charietto illustrates that in Roman minds Germans were still perpetual drunkards, as they had been in the first and second centuries when Seneca and Tacitus wrote (Tac. Germ. 23.2; Sen. Ep. 83.22).

The barbarians' lack of construction skills was standard diet for Roman readers. Cassius Dio, for instance, told them that Domitian and Trajan had had to teach Dacians the civilized arts, and the same was still apparently believed by the imperial chancellery of Germans and Goths, since in 419 Honorius and Theodosius II threatened to execute those who betrayed the arts of shipbuilding to the barbarians (CTh 9.40.24). This was declared even though the Scythian Borani had subjected the empire to devastating sea attacks in the mid-third century (Zos. 1.32-34) and though Saxon raiders had long experience of seafaring. Anyone who thinks the Huns were primitive nomadic raiders should read the personal account by Priscus of Pannium about the elaborate machines they built to capture Roman forts and cities (pp. 230-32 Blockley).

The classical ideology of the Roman Empire from the Republic onward was quite simply that Rome ruled the world, either directly in the provinces or indirectly by political influence. One does not have to look far to see the persistence of this claim in the later Empire. A Gallic panegyricist praised Constantine's campaigns against the Frankish tribes for having "rooted them out from the very homes of their origins and from a farthest shores of barbary" (Pan. Lat. 6 [71.6.2). Even more sober inscriptions, like that commemorating the building of Constantine's fort at Deutz, on the right bank of the Rhine opposite Cologne, announced "the subjection and control of the Franks" (CIL 13.8502); and his imperial coinage showed Francia prostrate and weeping.

The evocation of early Empire and even Republican ideology is often explicit. Another Gallic orator, in a deliberate reference to Augustus, praised Theodosius because the terror of his name had reached India, Arabia, and the icy north and his rule had gone beyond the "terminal points of nature" (ultra terminos rerum metasque naturae; Pan. Lat. 2[12]. 22-23, 33). Alexander the Great's dream of conquering Persia, apparent in earlier emperors like Trajan and Severus, was alive in Constantine (Joh. Lyd. De Mag. 3.34). Julian's claim to be invading Persia only for revenge was no more than the kind of rationalization for conducting a bellum iustum used by Republican heroes, whom Julian actually recalled in the speech reported by Ammianus (23.5.16 ff.).

The ideological tension between the Greek concept of an empire surrounded by fortified frontiers to separate provincials from the irredeemable savagery of the barbarian exterior and the more Roman vision of universal rule also continued in the later Empire. A tetrarchic inscription on the Danube in Lower Moesia claims that the emperors had "established an eternal guardpost" (praesidium), suggesting the notion of a protective screen .6 But despite the increasing pressure, it does not appear that the Romans ever seriously contemplated walls as a means of frontier defense. It is true that in the later Empire a large number of examples of walls are known, especially in mountain passes, such as those in the Balkans and Caucasus. With the possible exception of the Caucasus passes, which were used to block the Huns, these walls were all behind the frontiers and served as internal tactical controls. In this respect they resembled city walls. Indeed, the makron teichos that ran for forty-five kilometers between the Black Sea and the Sea of Marmora was probably built in the later fifth century, just like city walls, to protect the aqueducts of Constantinople. In the Julian Alps a complex system of fortifications and wall sections across valleys, known as the claustra Alpium Juliarum, was constructed in the later Empire, possibly begun under Diocletian or Constantine, to control traffic between Italy and Illyricum. But not a single example is known when this system was used to keep out barbarian invaders, not even the Gothic incursions of the early fifth century. The walls, in fact, appear to have been intended as tactical battle defenses against rival political pretenders, not as frontiers of empire.

Frontier ideology, therefore, far from adapting to assimilation, became more extreme in its praise of traditional Roman values and superiority. African inscriptions from Maximian to Theodosius reveal all the themes of mastery of barbarians (dominatio gentium barbarum) and expansion of the empire (propagatio Rornani imperii) that were typical from Augustus to Septimius Severus. The Gallic panegyrics echo them precisely. The tetrarchic emperors were exalted because "they had carried forward the boundaries of Roman power by their courage" (Romanae potentiae terminos virtute protulerant; Pan. Lat. 8[51.3). The final unreality arrived when the Goth Theodoric the Great was proclaimed in Italy as propagator Romani nominis, dornitor gentium (CIL 10.6850).

We might say that the ideology of an empire without limits almost prevented Romans from admitting that a frontier ever existed. This does not mean that all Romans were completely blind to the problems of populations beyond the frontiers or that they refused to make accommodations. But the compromises included the dreams. For example, Themistius, a consular at Constantinople in the fourth century and a firsthand observer of the formidable reality of peoples moving from central Europe and off the Russian steppes into the lower Danube and Balkans, recommended that Rome should peacefully absorb the Goths and in this way, he adds, would expand its frontiers. Libanius, a contemporary teacher of rhetoric in Antioch, had the same idea of the providential role of a universal monarchy under Rome.

The unpopularity among conservative Romans of a more realistic policy of accommodation with barbarians was manifest against emperors and military leaders such as Constantius 11, Jovian, and Stilicho. When Jovian conceded the city of Nisibis in Mesopotamia to the Persians after the disastrous campaign and death of Julian, Ammianus Marcellinus indignantly claimed (quite incorrectly) that this was the first occasion since the foundation of Rome when any territory had been lost to the enemy (Amm. Marc. 25.9.9). Even in the fifth century, with invasions threatening the gates of Rome, Stilicho lost support among Roman upper classes because he recruited barbarians to preserve the unity of the empire rather than crushing them. What could he do in face of the sheer absurdity of intelligent men like Synesius, later bishop of Cyrene, when the latter demanded that Rome employ no Goths or Huns but instead send all philosophers, craftsmen, merchants, and idle circus spectators to serve in the army (De regno 14)?

There were de facto boundaries, of course, both administrative and political, but modem historians find them difficult to define and contemporaries were never inclined to do so. Justinian's edict ordering the recovery of Africa from the Vandals in the sixth century commanded his generals to go "to that point where the Roman state had its boundaries [fines] before the invasion of the Vandals and the Moors" (CJ 1.27.4). The order was followed by detailed instructions concerning the maintenance of the frontiers (limites). They gave the illusion of certainty. But Justinian was in reality making a propaganda statement for public consumption to announce the restoration of the empire, and whether he really believed that boundaries and frontiers could be so easily discovered, we cannot know.

It is difficult, also, to agree with those who argue that Christianity enriched Roman. opportunities for assimilating barbarians "by opening up new ways to them". Ambrose, bishop of Milan, followed the traditional view of drunken enemies when he encouraged the sale of wine to barbarians, in order, he said, "that they may dissolve in drunkenness and thus be weakened" (De Elia et ieiunio 54). To him barbarians, Christian or not, were the enemy. For instance, he considered that usury, forbidden to Christians, was quite proper in deals with barbarians, since "where there is a right of war, there is also a right of usury" (De Tobia 15.51). Both examples, incidentally, assumed that trade and the capacity to enforce contracts existed beyond the frontiers, despite whatever bans the state may have wished to impose.

The conservatism of such perceptions was hardly likely to produce a frontier policy that recognized the reality of a new kind of coexistence with the population pressing in upon the Roman Empire. The treachery of barbarians who served in the army is taken for granted by almost every contemporary source. Yet it was not, by any modem, objective assessment, particularly frequent or striking. The savagery and deceit of Franks or Alamans, which our sources recount, was easily matched by Romans. It was not that the Romans were simply responding in like manner to "barbarians" because they lay beyond the moral frontiers of behavior. Even within the Roman world judicial and administrative savagery, such as mutilation and pouring molten lead down the throats of deserters, robbers, and criminals, was now a regular feature of Roman law. As Romans and barbarians became more alike, the dominant upper-class ideology of Rome became more shrill in its chauvinistic refusal to recognize the fact.

The Concept of Limes

The reality of frontiers and their historical development is complicated: first, by the Roman ideology of power that always claimed to reach beyond the formal lines of administered territory; second, by the strategic design of control that varied according to conditions from frontier to frontier; but third, by our own interpretation of the vocabulary Romans themselves used to describe various forms of military and civil limits, which perhaps took on different meanings in the later Empire. Since many of the terms had become fixed in Roman perceptions by the later Empire, this may be the right time to digress for a moment about the meaning of limes. It has been the subject of much study recently, and I shall only give a summary of the argument.

The word limes itself was originally a surveyor's term, adopted for military purposes to mean a road, and this technical sense was preserved throughout the later Empire in Roman surveyors' manuals. In practice, too, we can see how in Algeria, for example, many of the sectors of the limes in the fourth century A.D. were simply lines or roads through the difficult mountains, such as the Grand Kabylie, which lay well to north of the line of military control. The road could even go forward at right angles to the front. In Jordania a recent inscription found in the Wadi Sirhan oasis records a road, although in this case it is called a praetensio, built by Diocletian's soldiers from the camp at Bostra southeastward down the Wadi Sirhan toward Saudi Arabia and far beyond the trunk road of the earlier limes. The term praetentura, linked to praetensio, was often used to indicate a line of points for observation rather than protection (e.g., Amm. Marc. 31.8.5).

Not surprisingly, a military road linking camps on the fringes of empire, often lined with watchtowers and staging posts, was bound to invest the term limes with a sense of boundary or border in both earlier and later Empire. And not only roads, since rivers were just as often used as means of transport to camps that, for that reason, were often situated along their banks. The term ripa, riverbank, therefore also took on some of the notions of frontier, even though it signified a border region rather than a precise river line. This does not mean, as I have tried to stress, that the river thereby became considered a defensive barrier. On the contrary, it could hardly have served as a line of communication if that were so. In the later Empire the Danube, which was regularly considered a limes, was used in 361 by the emperor Julian for the transport of three thousand troops from Raetia to Singidunum (Zos. 3.10.2), just as the Rhine was used by the same emperor for the supply of grain (Amm. Marc. 18.2.3; Lib. Or. 18.83, etc.). But because soldiers occupied the forts on the ripae of the limites, the name ripenses probably became an alternative for the term limitanei to mean troops on the frontiers.

Use of the rivers for transport, however, obviously meant control of both banks, so it is not surprising to find that the emplacement of forts in solo barbarico ("on barbarian soil") remained a tactical necessity throughout the period. But the need for political control beyond the administrative boundary, either through direct military occupation or through alliances, explains why limes came to mean a frontier zone as well as some sort of boundary. In fact, it is unclear whether the word limes was ever official terminology except on the African frontier, where special conditions came into force by the early fifth century. Otherwise the term seems to have been used informally to describe a region within which military buildings were constructed both in advance of and behind the line of administered frontiers.

Possibly this idea of a deep zone of forward control explains the description given by a sixth-century writer, John Malalas, about the flight of a Saracen phylarch from a Roman frontier officer of the Syrian frontier, first through an "outer" and then an "inner limes toward the Indias." The notion of control without administration also corresponds closely to the curious topographic lists contained in such works as the Sphaera or Cosmographia of Julius Honoratus. These cosmographies, which named territories such as India or Gaetulia (in Africa) as "provinces" of the Roman Empire, were compiled in the form in which we have them today and put in circulation during the later Empire. As we saw, the term limes never occurs in those lists. The ambiguity of the word limes in the later Empire, therefore, and its sporadic use show that the idea of a fixed military or political frontier was never strong and that Roman strategic aims had not radically altered from those of the earlier Empire. What changed was the extent to which the aims were achieved and how far the claims to control corresponded to reality.

The Strategy of Late Imperial Frontiers

Parallel with the propaganda, frontier strategy remained simplistic tic and unchanged. just as in the earlier empire, conquest of the enemy and control beyond the limites were all that emperors desired, if the sources are an accurate guide. Although Constantine claimed to have recovered the empire of Trajan (Jul. Caes. 329c), he made his proselytizing zeal for Christianity into a vehicle to claim power beyond the frontiers, insisting on religious clauses in treaties with Sarmatians and Goths and proposing himself as protector of Christians within the Persian Empire. His actual control of southern Dacia west of the Olt and his impressive bridge over the Danube, publicized on coins, show that the Danube was in no sense recognized as a military barrier or a strategic limit.

There is one other revealing feature about the Danube frontier that has received little attention. The military units listed on the Notitia Dignitatum are relatively complete, since they were not stripped away in the early fifth century as they were on the Rhine. That allows us to make some judgments about the military strategy and tactics of warfare on the riverine frontier, even if we know very little about the way the field army of the interior operated in relation to the frontier troops. Broadly speaking, it appears that the frontier limitanei contained units both for defense (auxilia, equites, cohortes) and for attack (cunei, auxiliares), the latter having a higher status and perhaps the special name of ripenses to distinguish them from the units of the line.

This shows that Roman strategy on the Danube-and quite probably on the Rhine-was never intended to be defensive. The conclusion is reinforced, furthermore, by the evidence of double camps at major river bridgeheads, where artillery units were often situated on the far bank. Examples of these doublets were noted earlier at Transmarisca/ Daphne and Oescus/Sucidava on the lower Danube and at the two left-bank forts opposite Aquincum (Budapest) and Bonnonia on the middle Danube. On the Rhine we have the evidence of double forts at a number of sites, such as at Zurzach/Reinheim, Kaiseraugst/ Wyblen, Basel/Robur, Altrip/Neckerau, and Horburg/Breisach. They are thought to be Valentinian, but we must not forget that the most famous doublet of left- and right-bank forts at Cologne/ Deutz was Constantine's work. The system on the Danube, according to those who have studied it, was not the work of a single emperor but developed from Diocletian over the fourth century.

Constantine's successors could not appear weaker or more defensive than the founder of the dynasty. Ammianus admired Julian for his "completely republican outlook" that criticized even Trajan and Severus for seeking personal glory before defense of the res publica (Amm. Marc. 23.5.17-18). In practice, however, Julian behaved like the great propagatores imperii: Ammianus stresses that it was Trajan's fort in Alaman territory that Julian rebuilt; much stress is laid on the fact that food was requisitioned from Alamans beyond the Rhine as from other "contractors" (susceptores). Julian advanced beyond the lower Rhine and attacked Franks by moving up the river Lippe (Amm. Marc. 17.1.11, 17.10.4, 20.10.2). His war with Persia, although defended by Ammianus as not of his seeking, was on Julian's own admission (Caes. 33a) conceived long before with Alexander and Trajan in mind. The strategy was almost aimless, beyond the desire for glory and titles, but it was probably not unaffected by the fact that Roman interests beyond the Tigris were challenged. On the Danube, where any decent information service would have forecast trouble, Julian contemptuously dismissed the Goths, since he was "looking for a better enemy" (Amm. Marc. 22.7.8). That gives us some idea of the prevailing mentality.

Valentinian was no less traditional in his "Augustan efforts". That means he advanced Roman claims etiam ultra flumen (Amm. Marc. 28.2.1) deep into transliminal territory of the Rhine and Danube. The Roman outpost at Hatvan-Gombopuszta, sixty kilometers beyond the Danube, gives substance to Ammianus's claim that he built forts in the middle of Quadi territory"as if already annexed to Roman rule" (29.6.2). On the lower Danube there seems little doubt that, whoever was responsible for the war that broke out between Rome and the Goths in 367, Valens was seeking not a strategic first strike to stem the pressures from the Russian steppes, but simply to win a military reputation. The sloppy inattention to the infiltration of Goths between this date and Adrianople shows how little military matters counted.

Even after the damaging usurpation of Gaul by Eugenius and Arbogast in 395, Honorius and Stilicho were praised by Claudian for imposing laws on the Chauci and Suebi across the Rhine, as though nothing serious was amiss. (In Eutrop. 1.378-84; De cons. Stil. 1. 218-3 1). The momentous strategic decision by Stilicho, and before him by Theodosius in the East, to use federate troops to man the frontiers goes virtually unnoticed and could be argued to have been a policy adopted faute de mieux . Yet the frontier had been effectively stripped of its regular troops, leaving the Rhine exposed for the great invasions of New Year's Eve 406.

There is no need to pile on examples. If there was a "more realistic vision," a strategy of accommodation and peaceful coexistence, that found form in the policy of Constantius II or Stilicho, that vision remained as unpopular and as unacceptable to most Romans as it had been in Hadrian's day. Stilicho lost support disastrously when he was prepared to treat with Alaric's Goths in order to protect Italy, since the Roman upper class still believed the only way to treat barbarians was to crush them into unequal subjection.

I draw two conclusions. The first is that at no stage and in no source after the recovery under Diocletian was it ever publicly admitted that the frontiers of the high Empire were diminished. Even in the obvious cases of Dacia, north of the Danube, and of the Agri Decumates, east of the Rhine, which had been lost in the troubles of the third century, it is extraordinary how difficult it is to find any mention of them. Not only was the withdrawal from Dacia not total, in that many provincials were left behind, but Dacia restituta was already proclaimed by the panegyricists before the end of the third century (Pan. Lat. 81513.3; cf. 10.4), and the name Dacia was given to a new province south of the Danube. Furthermore, the Roman emperors of the fourth century seem to have regarded much of the lowland region of old Dacia as still their fief, even if there was no longer a trans-Danubian province.

As far as the loss of the Agri Decumates is concerned, the strange silence of our sources is significant. An interesting suggestion has been made recently that much of the population was transferred across the Rhine to a region known as Decem Pagi near Metz, creating perhaps the illusion, as with Dacia, that the Romans continued to control the region. Certainly "Alamannia," the territory that covered the old Agri Decumates, was claimed by the same panegyricist cited above to be totiens proculcata: "often crushed." The Augustan history, written in the late fourth century, probably reflects what Romans would have liked to believe, when it says that the emperor Probus recovered and settled the land up to the river Neckar and the Schwabian Alb (HA Prob. 13.7). We saw earlier something of the same mentality when the emperor Justinian in the sixth century proclaimed with elan that the fines antiquae of Africa would be recovered in their entirety. No emperor could afford to admit that Roman territory had been lost, since that would have contradicted the ideology of the sacred termini.

It is wrong to think this was all bluff. Emperors believed their own propaganda. Valentinian died of apoplexy because the Quadi dared to claim it was provocation when he built fortifications in their territory (Amm. Marc. 30.6.2-3).

My second conclusion is more tentative. I have doubted the sophistication of strategic planning by Roman emperors. I am equally skeptical, therefore, about the development in the fourth century of some new rational system-what has been called "defense in depth" or "elastic defense"-and the abandonment of the old system of what is called "perimeter defense." If it was right to reject the existence of a linear strategy for earlier frontiers, we cannot now assume that there was a conscious modification of a nonexistent concept. Zosimus's attack on Constantine for ruining the security of Diocletian's frontiers by withdrawing most of the army to the cities in the hinterland (2.34) is, as far as I can see, the only direct ancient evidence that can be used to defend such a supposed change in strategy.

But it is also demonstrably rubbish. There are as many, or more, military buildings on the perimeter limites of the Rhine and the Danube dated to Constantine's reign as to that of Diocletian. The disposition of forts on the Arabian limes has been misunderstood as providing strategic defense in depth rather than protective centers for the local population. Furthermore, there is positive evidence that Constantine did not withdraw Diocletian's advanced posts in the Wadi Sirhan. The history of the later frontiers does not reveal any new grand strategy on the Danube. Africa, Arabia, and Mesopotamia in the fourth century remained more or less as Diocletian left them.

Indirect evidence of new strategic thinking has been claimed from the changes in military organization that took place during the fourth century. What is meant, of course, is the new provincial field armies of the comitatenses. This is not the place to discuss the birth of these mobile armies and the quality of the remaining frontier troops (variously called ripenses or limitanei), which have not lacked commentators. Calculations from the army lists con tained in the Notitia Dignitatum, which are necessarily approximate, suggest that between 50 and 60 percent of all troops remained on the frontiers until the end of t6e fourth century. If the army at least doubled in size in the later Empire, as most historians believe, that must mean there was no absolute decline of numbers on the frontier. As to the quality of the limitanei, there is one irrefutable argument to show they had not declined; that is, regular upgrading of units from frontiers to pseudo-comitatenses must prove that they had not degenerated into a peasant militia. The growth of regional field armies (as opposed to Constantine's central, mobile force) could just as well have been the consequence of the divisions of the empire under the sons of Constantine. The army, that is, was divided regionally for political, not strategic, reasons.

It is sometimes suggested that the building of city walls and new types of urban defenses in the later empire is proof of a new frontier strategy of defense in depth, which was linked to the weakening of the frontier troops and the creation of the mobile field army of comitatenses. In other words the cities, as in the medieval period, were part of a calculated policy of nodal defenses, which allowed the enemy to enter Roman territory but delayed them while Roman forces grouped for counterattack. This is the interpretation given to the passage of Zosimus (2.34.1-2) already referred to, which says, "Constantine allowed the barbarians to penetrate land ruled by the Romans without hindrance ... by withdrawing the majority of the soldiers from the frontiers and installing them in the cities, which had no need of protection."

Quite apart from the accuracy of Zosimus's statement about the frontiers, which I doubted earlier, there are serious objections to such an interpretation of the role of cities as part of a supposed new strategy in the fourth century. In the East it is clear that city fortresses had always served as the basis for the frontier system, particularly in Mesopotamia and on the desert fronts, where any other form of frontier alignment is difficult to detect. But this was as true of Diocletian's strategy-or for that matter of Trajan'sas of that of Diocletian's successors. In the West, while it is true enough that in Gaul and Britain many cities in the later empire were equipped with new walls, including projecting or external towers and other defensive features, the great majority date from just after the crisis of the mid- third century, when the frontiers had been badly breached and the cities of the interior exposed. Current research tends to regard the emperor Probus, who ruled from 276 to 282, as the author of this activity in Gaul, Britain, and perhaps the Danube provinces, too.

By contrast, there is no evidence of a systematic or coherent plan of city fortifications in the fourth century, such as one would expect had there been a new defensive strategy. A modem comprehensive study of fortifications in the later empire barely discusses the policy attributed to Constantine by Zosimus and makes no reference at all to cities as an element of some new strategy. Another study of Gaul and Germany concludes from the great variety of defenses that even if one can see some regional similarities, "there can have been no central directive covering fortifications in the north-westem empire."

In Britain, where most cities had already acquired walls between the late second century and the third century, some of them to form part of the Saxon Shore system of defenses in the later third century, there may have been something more like a systematic restructuring of urban and small-town walls in the later fourth century by the addition of external bastions. The development has often been attributed to Count Theodosius, although without any archaeological evidence, because Ammianus (28.3.2) says, "He restored the cities and forts." But it can have had nothing to do with a strategy linked to Hadrian's Wall, which remained intact after Theodosius, since many of the northern forts did not receive any of the new external towers. Again, those who have studied these defenses in detail talk of "a studied lack of standardization" and "an indifferent understanding of the tactics involved. The pressures had shifted from Hadrian's Wall to the sea raiders of the Saxon Shore.

In fact, we know very little about the garrisons of the comitatenses in the provinces, nor can we tell whether the special units of laeti and gentiles, which are recorded as being stationed in civitates in documents like the Notitia Galliarum, were inside the town walls or encamped in the countryside. The spectacular Langmauer just north of Trier probably marked out a huge imperial estate within which were stationed troops, including units of the comitatenses, who are recorded as helping to build the wall. But Trier was an imperial capital in the later fourth century, and the emperor normally had a "praesental" army (as it was called) attached to the court. Otherwise the only explicit reason I can trace in the sources for stationing military units to the rear of the frontiers in the cities is that given by Ammianus, himself a soldier, who speaks of their "being distributed throughout the cities in order to be better supplied" (ut commodius vescerentur; 16.4.1; cf. 31.16.8). Obviously such dispositions were closely linked to the greater insecurity in the countryside, which made it necessary to store food in the cities (Amm. Marc. 31. 8. 1). Troops were required, therefore, in castella and burgi or along the major routes to protect the collection and transport of supplies. It may be, too, that extensive settlement of foreign communities within the frontiers had increased the danger of sudden raids on cities, as we hear of on more than one occasion. Ammianus (16.11.4) tells of one such incident when a group of laeti made a sudden attack on Lyons, which just managed to keep them off by closing the city gates.

But this was hardly a new principle of defensive strategy in depth. it was merely prudence and tactical common sense. it supports the thesis I have been proposing, that the main problem on both the western and eastern fronts was dealing with smallscale raiders and infiltrators, not large invading armies. A good example of the sort of conditions that prevailed in Gaul in the later fourth century is described in the Life of Saint Martin of Tours (Sulp. Sev. V. Mart. 18), when a sudden rumor of an attack by raiders put the city of Trier in a panic until Martin identified the devils who were spreading the false news. It was this, if anything, that produced new strategic thinking, which had begun even in the later second century A.D. But that is a very different matter from a defensive frontier policy.

The Nature of the Barbarian Invasions

It is easier for us to see with hindsight that the reality of frontier pressure in the long term was very different from the ideology. But I find it also quite difficult to believe that the emperors (and those who wrote about them), who boasted of their exploits, were totally self-deluded as to the nature and size of the happenings in front of their eyes, which we have subsequently termed the "great invasions." That is the problem. On the one hand, the language of panegyric, inscriptions, and doomsday predictions necessarily demanded bombastic phrases like those Ammianus uses when he proclaims that "the barricades of our frontier were unlocked" (nostri limitis reseratis obicibus) or that "barbarian columns" were flowing like lava from Etna (31.4.9). Others used the image of storms that raged and tore the empire (Claud. Cons. Stil. 1.282-83). But on the other hand, that is no good reason for us to be seduced into using the same hyperbole of "floods and waves" of barbarians just because they do.

The fact is that, when we come to look in detail at the ancient sources, they seem to contradict themselves. On the eastern frontier, says Ammianus, the Persians made a strategic decision that large-scale campaigns were profitless and that the best procedure was per furta et latrocinia (16.9.1). That was the phrase I picked out at the beginning of this chapter in discussing the small-scale raids and infiltration that were taking place on the western front. It also corresponds closely to Ammianus's opinion of the mode of warfare conducted by the Saracens, a people who were "like rapacious kites" and "fitted for clandestine acts of war" (ad furta bellorum appositi; 14.4.1, 23.3.8). Ammianus's perception was shared by other contemporaries with direct experience of the East, such as Jerome, who told of kidnapping and raids on monasteries by small bands of marauders. And even the emperor Julian ' said that he regarded the Saracens as no more than "bandits" (lestai; Or. 1.21b).

All this finds interesting parallels with other frontiers, although obviously no one would want to say the threats were exactly the same. On the Danube the fleet patrolled the river to stop infiltrators (Amm. Marc. 31.5.3). Themistius, a contemporary orator, described the marshy regions of the, Dobrudja where raiders came in small groups "not as soldiers but as robbers" (Or. 10. 136-38). Julian thought the Goths no more than a petty nuisance (Amm. Marc. 22.7.8). In Valentinian's day the Quadi were judged to be .a people [natio] not to be feared very much" (Amm. Marc. 29.6. 1). The Sarmatians were considered "better suited for robbery than open war" (Amm. Marc. 17.12.2). Charietto's value to Julian on the Rhine was that he knew how to cope with the small raiding bands of Franks. And so on.

We can get some idea of what this meant from the archaeology of rural establishments in northern France, Belgium, and the left bank of the German Rhineland. There some settlements were being abandoned even before the destructive invasions of the later third century, perhaps for economic reasons. But there was a considerable revival in the early fourth century, more marked in some places than others, that subsequently suffered a serious setback after the usurpation of Magnentius in the mid-fourth century. Thereafter the settlements underwent progressive change and impoverishment of buildings until the end of the fourth century and the early fifth century. The rebellion of Magnentius, we must remember, was a civil war, not an invasion, in which both sides invited Franks and Saxons to enter Roman territory.

Indeed, we might ask, on how many occasions were the "barbarian invasions" really invasions as we would think of them? This is not, of course, to disregard the seriousness of the great set piece, field battles like Strasbourg or Adrianople, or the devastating effect of the irruptions into Gaul and Italy in the early fifth century for those inhabitants who lay in the path of the invaders. The real problem is to assess the importance of these dramatic but isolated events in comparison with the more banal but continual pressure of the small bands of infiltrators.

I have not seen anyone challenge the arguments put forward by Delbruck in his famous Geschichte der Kriegskunst in Rahmen der politischen Geschichte, published in the earlier part of this century, concerning the strength of Roman and barbarian armies. After careful analysis of the figures given. by our sources for the battles of the fourth century and the Volkerwanderung of the fifth century, Delbruck-who was a pupil of Clausewitz and no fool on military matters--came to the conclusion that most of the tribal movements never consisted of more than 5,000 to 15,000 fighting men, and in some cases, like that of the Burgundians in the mid-fifth century, the numbers were as low as 3,000 men.

Even the figure of 80,000 Vandals (including old men, women, and children), which has been considered more authentic than others because they had been subject to a census before taking ship to Africa in 429-even that figure has been challenged on the grounds that it was "black" propaganda put out by Geiseric himself to deceive the Romans as to his strength." Ammianus's figure of 243 Roman dead at Strasbourg (16.12.63), Delbruck thought, was a more accurate guide to the size of the forces engaged in battle than the thousands upon thousands of Alamanni reported by Libanius and Ammianus. So much for the rivers of lava. We have to break away from the stereotypes of "tribal" history and mass movements of tribal migrations, which, when we can trace them archaeologically (as we can in the case of the Goths), seem to be slow movements of infiltration by small groups of warriors. Aetius's glorious victory over the Salian Franks at vicus Helena, enthusiastically hailed by Sidonius (Carm. 5.219-29) as a great victory, turns out to be no more than a "minor skirmish" when the Romans broke up a wedding party.

Of course, from time to time there were larger but also more ephemeral federations that, typically in our reports, lacked cohesion and organization. News of the great Gothic coalition in 365 was not apparently serious enough to divert Valens from his expedition against Persia (Amm. Marc. 26.6. 11);3' and the Gothic migration in 376 was in Ammianus's simplistic account divided under the leadership of at least seven or eight chieftains with their followers. There were reports of a "mob" (multitudo) of unknown nationes "wandering around the Danube," says Ammianus, "scattered with their families" (31.4.2). The terms "Goths"-similar to "Franks" and "Alamans"-is a generic term meaning "men" and gives a false impression of unity to what was basically a society fragmented into subdivisions that rapidly disintegrated after rare shows of unity.

The same applies to other invaders. Among the Alamans, it was their hydra-headed, multitribal organization-or lack of it-that made them so tricky to contain, according to Ammianus's narrative (e.g., 27.10.5; reparabilis gentis motus timebantur infidi). As for the organization of the Franks, Gregory of Tours was baffled by the constant references to petty chieftains (reguli and duces) in Sulpicius Alexander's history of the fourth-century Frankish invasions (HF 2.9). The "barbarian conspiracy" against Roman Britain in 367 turns out be difficult to document and almost certainly exaggerated by Ammianus for ulterior motives. Even Attila's much feared Huns included the Akatziri, subjects described by Priscus, who "had many rulers by clans [phyle] and families [gene]". And after Attila's death his whole kingdom collapsed like a pack of cards into rival groups.

On the eastern frontier, Persia, the only really unified force that might have launched a coherent attack on the Romans, was content to keep a relatively low profile. An important recent study has concluded that it is impossible to prove that the Persians had any general aggressive intent to occupy Roman territory.

But the same was not true of the Arabs. It is also the "multiple structure" of the Arab federates that makes it impossible to identify the many phylarchs who lived on the eastern borders of Roman rule. Indeed, one recent study goes so far as to argue that the whole Saracen menace was a myth invented by Ammianus and his contemporaries because of their obsessive paranoia about brigands on the borders of Roman civilization. Since most Safaitic and Thaumadic inscriptions have been found within the borders of Roman territory, we should, says the author, regard Saracen attacks as "cases of internal strife," not as "intrusions of distant tribes.

I have to say I find such an extreme view difficult to accept. It is hard to talk away all the references to attacks from beyond the frontiers. Mavia, queen of one Saracen group, withdrew extra limitem (Rufin. HE 2.6) when she fell out with the emperor Valens before making a devastating attack on the provinces. The transhumance movements of the Jordanian border make it almost inevitable that there were symbiotic relations between folk on either side of the frontier that, unless controlled, could become destructive. I am not concerned here whether Ammianus did justice to the Saracens or understood the differences between the relatively stable federations, such as the Lakhmids and Tanukhids, and nomadic bands of pastoralists. The important thing to realize is that-just as in Africa-the border was open and the differences between external raiders and internal dissidents were difficult to disentangle. The same might be said of the lower Rhine border between Frankish Holland and Toxandria.

in other words, we must be careful about what we are told. Take, for example, the description of the tough lives of the "Scythian" Huns and their terrifying cavalry, by which Ammianus does his best to stir up panic in Roman hearts. Not only is this a stereotype of "permanent nomads" that Herodotus would have recognized, but our suspicions are roused still further when we are told by a study of the ecology that it was impossible for any land west of the Carpathians to sustain a truly nomadic force of more than fifteen thousand warriors. In fact the eyewitness, Priscus of Pannium, found Attila living not on horseback but in a large "village" where the aristocracy had enclosures adorned with wooden walls, towers, and even a Roman bath. Archaeologists also report that on the Hungarian pusztas where the Hun's empire flourished "not a single usable horse bone has been found".

The Demography of Invasions

If the impression we get from the ancient sources, therefore, is that there were no sudden intrusions of great new populations, that impression is reinforced by archaeology, insofar as it is possible to generalize from the uneven spread of evidence. The most detailed information available comes as a result of the program of the Deutschen Forschungs-Gemeinschaft in Lower Saxony over the past four decades, and through a number of studies of the Low Countries by various Dutch archaeological institutes. Enough settlements on the coastal clay terpen/wurten (like Feddersen Wierde) and on the sandy geest (like Flogeln), as well as at inland sites (like Wijster), have now been investigated for us to be reasonably certain that these areas were the heartlands of Saxons and Salian Franks in the fourth century A.D. Since these results are well known, I need do no more than repeat the conclusions that are relevant to my theme of frontiers.

First, the growth of population. On virtually every site there is evidence of a dramatic increase in the size of settlements between the first and the fifth centuries A.D., rising to a peak in the later fourth century. At Feddersen Wierde on the coast near Bremerhaven, the number of cattle stalls and farmhouses allows us to calculate an augmentation of population and herds of between 250 and 300 percent by the end of the third century." Inland at Wijster, in the Dutch Drenthe province, the population "reached its greatest expansion," say archaeologists, in the period between 360 and 395.

I do not wish to overdramatize the impact of this evidence. We cannot produce census statistics to prove an overall increase in the population, although it seems probable. But expansion was not continuous, and site settlements plus pollen analysis suggest an actual decline of agricultural activity on both sides of the Rhine in the third century. At Wijster-though not in Lower Saxonythere was a hiatus about 225, but it was followed by a settlement that was culturally identical. It also seems certain that the climatic change of the Late Roman Transgression caused a rise of sea levels and salinity of lowlands that forced abandonment of pastures and greater nucleation at the period of greatest prosperity. That probably accounts for the fewer but larger settlements that have been noted, despite a lack of reliable dating on many sites in the later empire.

What has to be stressed, however, is the continuity of the material culture. When the site at Feddersen Wierde was finally abandoned because of the rising sea level in the early fifth century, the silt covered early types of distinctively "Saxon" handmade pottery, which means almost certainly that they evolved locally out of the preexisting culture. Stutzarmfibeln, too, were found. These are a type of brooch looking like a crossbow (en arbalete) common to all frontier societies along both sides of the Rhine, and for that matter along the Danube too in the same period. Although they are often associated with Germanic intruders and are illustrated as such in the chronological typologies of German grave goods that the archaeologist H.W. Bohme drew up, these same brooches are found in the Sintana. de Mures culture, which we usually associate with Goths, along the Danube. In other words, they look like a frontier cultural development that took place over a long period of time.

How far this picture of apparent population expansion and nucleation is repeated over the rest of the lands beyond the Rhine is difficult to judge from limited evidence. it has been suggested, in the absence of positive indications, that the Alamanni were a relatively unimportant force in comparison to the Franks, despite their ferocious attacks in the early fourth century, and that generally they had had little contact with Roman territory before the occupation of Schwabia in the third century. Perhaps so. But Roman cultural influence on the Alamans looks strong in the fourth century. There is quite a lot of archaeological evidence to support Ammianus's statement (17.1.7) that the Alarnans built their houses in Roman style and that rather more "villas" of a Roman style were still occupied than might be thought. One such example studied at Bondorf has produced distinct, though limited, signs of continuity in graveyards.

"The overriding impression conveyed by the excavated sites," says one archaeologist, therefore is "of stable and enduring communities" over "decades or even centuries". That means that we must reject the myth of a late German migration. The unimportance of the Alamanni in the fourth-century invasions, even though they were described as a gens populosa, was surely because their population was not under pressure until finally pushed by the Burgundians in the late fourth century. They lived in a part of the world that was described in chapter 3 as the empty 14 quarter of Europe. But it was otherwise with the Franks. Theirs was the country of the "Germans" that Procopius describes as polyanthropos-full of people-in contrast to the Visigoths or the Thuringii (BGoth. 5.12.20-21). Whether he was talking of the lower or the middle Rhine Franks is impossible to know. Possibly both. The Meuse-Rhine basin has already been discussed, but we may note that the only real concentration of sites in the fourth to fifth century north of the middle Rhine was in the land of the Chatti Franks just north of the Lippe.

Apart from the Germans, changes were also taking place on other western frontiers. I have already noted that on the British frontier there is the well-known puzzle of what had happened to Hadrian's Wall. It is generally agreed that there had been a reduction in numbers of troops, especially in the western British province, perhaps by as much as 40 to 50 percent. Regular occupation of posts beyond the wall declined after 310, and military demands and exchanges probably ceased. This is thought to be one reason British grain was now available for the Rhine army (Amm. Marc. 18.2.3, etc.), which in turn might explain why the farms of Picardy seem to decline.

Quite simply, the threat by land from the north had disappeared and could be contained by native scouts (arcani), until they disgraced themselves (Amm. Marc. 28.3.8), together with the old auxiliaries on the wall. The main preoccupation was raiders from the sea. Like the Alamanni, therefore, the British population in this region was neither under pressure to migrate nor stimulated by frontier exchange to expand, which seems to be confirmed by the lack of any increase of native sites north of the wall. Unlike the Alamanni, British troops and officers were also a negligible part of the later Roman army. The absence of disruption in pollen diagrams, despite the apparent abandonment of wall sites by 420, makes one think the region remained relatively static.

The African frontier was not one of the hot spots of the later empire-not, at least, in the fourth or fifth century. But it is a useful case to cite alongside the case of the Rhine, because of the rise in pressure on the Tripolitanian frontier, which began in the fourth century and came to a climax in the Byzantine period. During the period we can see something of the same developments that occurred on the Rhine, although on a lesser scale. The attacks began with the Austuriani in 363. They probably accounted for the reinforcement of the forts system attached to the clausurae in the mid-fourth century.

But a recent hypothesis argues that the attacks by the southern Laguatan of the sixth century were not the result of a single migrating population, as has often been thought, but the conclusion of a long period of raids, in the course of which long-standing border populations, such as the Arzuges and the southern Tripolitanian farms discussed earlier, were absorbed by the invaders. The attacks, it is further argued, were originally stimulated by overpopulation of the oases and loss of productivity owing to soil salinity, reverting back to an earlier hypothesis of "a desert momentarily overpopulated."

In other words, we have here all the ingredients of frontier dynamics: surplus population, cultural homogeneity, and a long period of symbiotic exchange. The marginal lands of Tripolitania were by definition a poor region where there was competition for scarce resources. This bred endemic rivalry and factions that increased as the frontiers became more stable and the population increased. Recent evidence suggests that there was a quite distinct, though small, temperature rise in the mid-third century, which put the ecosystem of the marginal territories under additional strain. If that is true, then we can understand how the stability might have become disturbed. For the moment, however, we must await further research to determine whether the climatic variation was the major determinant or whether, as seems more probable, it only -contributed along with other factors like soil exhaustion and overgrazing.

The Arabian frontier was perhaps not dissimilar, although there is a major dispute here about whether there was an external or an internal enemy. The debate at least highlights the fact that the seminomadic or sedentary inhabitants within the Roman provinces-the Safaitic language group of the Hauran in southern Syria and the Thamudic language group on the Jordanian plateauwere extraordinarily closely linked politically and culturally to Arabs beyond the frontiers, irrespective of which direction the seasonal migrations took place. Clearly, where there was such an open frontier, the inhabitants on both sides always tended to be a law unto themselves, and it was impossible for the Romans to control them other than by political pacts and payments-an extreme example of frontier "pull."

Of particular interest, however, is what effect the Roman occupa tion of the key transhumance borderlands had upon the populations. And in this we can see some comparison with Africa. Information is limited, but those who work in the area speak of the prosperity of the frontier strip, possibly increased sedentarism, and increasing population and therefore increased clashes with pastoralists. There is some evidence of scarce resources that drove pastoral Saracens into Roman territory and of clashes caused by climatic change, although it is difficult to judge how serious it was in the long term. And it may be, if we are to judge by the increasing references to Saracen kings, that there was a rise in social and political organization, although whether this had anything to do with the "bedouinization" of the Arabs is a thorny and unresolved problem. The Arabian frontier, like the African, was slow to develop, but the sixth century Procopius believed that the Lakhmid Saracen king Mundir was Rome's "most difficult and dangerous enemy" (BPers. 1.17.45).

At the extreme opposite end of the spectrum of stable and unstable frontiers lay the Gothic lower Danube. It is no accident that we know most about new intruders in this part of the frontiers and that it most closely resembles the Rhine frontier. Once again we can see the same ingredients as before that rendered the mass critical-an increase in population, scarcity of resources, and rapid assimilation through transfrontier contacts. The rise in population hardly needs comment, given the intrusive population into the Danube plain of Goths, succeeded by Huns, Alans, Slavs, Bulgars, and Avars. But it is right that we should be reminded that individual groups were small, and that the main problem was lack of resources, which together with population pressure produced chronic food shortages. On this frontier we have more recorded cases than on any other of voluntary "surrender" and receptio into Roman territory of groups, such as Sarmatians, Marcomanni, Taifals, Carpi, and of course Goths. The specific figures are large, always in the thousands, and probably exaggerated, but they illustrate the growth of pressures in the struggle for resources.

The social and cultural dimensions of the demographic changes of these frontier societies are the key to understanding the barbarian invasions in the West. The German and Dutch Herrenhofe and the Tripolitanian gsur each illustrate growing social hierarchies on the frontiers, which controlled the scarce resources of imports and production as the various nuclei of population grew. At Feddersen Wierde the southeastern sector of the village was dominated by a single large longhouse surrounded by an oak pallisade and ditch, to which were attached storehouses, stables, and a kind of assembly hall. Not surprisingly, this complex has been termed the "manor" of a chief. But it might almost be a description of the Tripolitanian gsur, the fortified farms around which large settlements of dependents developed. The export of Roman commodities, as I have tried to stress, whether they were acquired by trade, raids, or foreign troops, accentuated the social divisions.

The effect of the Roman frontier policy in encouraging alliances with local border chiefs and their employment in ethnic units must have reinforced their political power. Hence the significance of the Saracen Thaumadic federation, whose temple dedication to the Antonine emperors was found at Ruwwafa, far beyond the frontiers in the Hejaz. And hence the plethora of phylarchs and "kings" we find cropping up in the confused history of the region, serving either Rome with their Saracen cavalry, or Persia, or neither. When Julian was retreating up the Tigris after his ill- fated expedition against Ctesiphon in 363, he found the Saracens hostile because "he had forbidden payments and many presents to be given, as in the past" (Amm. Marc. 25.6.10).

This growth of political control of economic resources by local elites was repeated on the Danube. Although the social grouping of retainers around reiks in Gothic society was a basic clan unit, the word reiks was an adaptation from rex, which may indicate the increased political power of chiefs after contact with Rome, no doubt often through service with the Roman army, There was also a growth of the importance of the Gothic baurg - the fortified residence of a Gothic lord and his retainers, at the expense of the free peasant villages. The Passion of Saint Saba tells the story of how Saba in one such village was killed in 372 by the retainer of a reiks. We saw earlier the considerable evidence on the middle Danube of chieftains' "villas" with their enclosures for retainers and their rich graves. They were quite clearly a development in relation to Roman trade goods from the second century onward and show by their contents the control such elites had of resources. One wonders how much the oikemata of Attila-the royal residence visited by Priscus-resembled such buildings.

To conclude, we cannot doubt the pressures of growing populations and the demand for food on some of the frontiers, although not on all or at the same time. The sources are explicit, and I have studied them elsewhere in detail. Fifty years before Adrianople Visigoths were dying of hunger, which drove them into treaty with the Romans (Exc. Val. 1.6.31; Jord. Get. 112). Vandals and Alans were forced by hunger to move westward to Frankish Germany, increasing the pressure on already scarce resources (Procop. BVand. 3.3. 1). The warfare this generated only made the conditions more acute. Julian's invasion of Alamannia or Valentinian's attack on the Quadi devastated crops and must have created "the scarcity of food" that Ammianus notes (17.1.7, 30.6.2). Emperors were prepared to capitalize on hunger and disease by using economic warfare to control rebellious "barbarians". Ambrose, as a Christian, thought this quite proper (Expos. in Luc. 10.10 (PL 15.18061), and it had some effect. But the normal permeability of frontiers by their very design, plus the long habitude of three centuries of symbiotic exchange of goods and manpower, ensured that in the end these vain expedients were unenforceable. It is that aspect I turn to now.

The Symbiosis of Frontier Societies

In earlier chapters I have tried to underline the differences between juridical or ideological frontiers, and the absence of such barriers in commercial, political, or cultural terms during the first two centuries of the Roman Empire. Interaction was not perfectly even but was limited to certain social groups, whose power and identity were defined by the exchanges. The result, according to my hypoth esis, was a commercial and perhaps a cultural affinity between barbarian elites beyond the frontiers and the Roman inhabitants of the cities or camps on the frontiers, while the ordinary rural settlements on either side remained relatively untouched by Roman influences and, in terms of their artifacts, seem to have resembled each other more than their own elites. As we all know, the Roman urban and military system was almost designed to increase this sort of stratification in the provinces. While the economic gap between urban aristocracies and peasants grew wider care was taken-particularly on the British and Rhine frontiers-to separate native recruits from their native leaders by ending the employment of ethnic units locally or by mixing local recruits with recruits from other provinces.

In the later Empire of the fourth century A.D. this separation was forgotten. It became increasingly common to employ native inhabitants from both sides of the frontiers as ethnic units in the army, frequently under their own chieftains, while at the same time more and more barbarians were admitted and settled within the frontier provinces, often on the borders. Social, economic, and cultural exchanges inevitably continued across the frontiers, despite-or perhaps because of-the use of trading privileges as a political weapon. Within the Roman province, therefore, the .pull" of exchanges increasingly created a frontier society that was fast becoming indistinguishable from that beyond.

Every frontier has had to face the difficulty of recruiting native irregulars. Holdich tells the story of how Sir Mortimer Durand, creator of the Durand Line on India's North-West Frontier, visited a hospital for soldiers and was astonished to find there two "wazirs" of the Mahsud Afghans enjoying the good life of the hospital at a time when Britain was waging a frontier war against their tribe. When he mentioned this to the two men, they roared with laughter and "expressed the pious hope that their own people were putting up a good fight". The story illustrates the ambiguity of frontier societies, repeated again and again in history. The Muscovite principality of the sixteenth century used border troops of "Cossacks"-a word as vague as Gaetuli or Goths-who were often peasants drawn from the enemy Mongols.

In theory, of course, the Romans had recognized the dangers of employing recruits locally ever since the rebellion of Civilis in A.D. 69, and I can find no evidence of change in official policy in the fourth century concerning stationing native ripenses troops in frontier forts. The law permitted a Roman dux to enlist recruits locally. But such recruits did not make up the whole of the local units, and the experience in 324 of the Christian monk Pachomius, who was locked up to stop him from running away while awaiting posting, suggests that even individuals were often sent abroad (V. Pach. 4). As so often, the Historia Augusta reflects fourth-century ideals when it recounts that the emperor Probus scattered sixteen thousand recruits in units (numeri) and frontier troops (limitanei) in different provinces, saying "that when the Roman was helped by barbarians it must be felt but not seen" (HA Prob. 14.7). That, I think, was meant to be official policy, and the only exception was in the occasional use of scouts and spies, like the Batavian exploratores at Roomburg near Leiden (CIL 13.8825) or the British arcani beyond the wall (Amm. Marc. 28.3.8).

So in principle the policy of foreign postings for ethnic units was continued. Despite the problems we have in discovering precise frontier dispositions from the Notitia Dignitatum, particularly about those on the Rhine, it is remarkable that Alamanni units are never recorded in the West and that in Britain there is not a single identifiable ethnic British unit, although clearly some soldiers in the various mixed units came from Britain. Even in Africa, where the third Augustan legion had been recruited extensively from within the province in the high Empire, not a single legion, or vexillation, of those units that appear on our army list as being upgraded from frontier to field army can really be identified as an ethnic unit recruited in Africa. We can, of course, assume that many individual soldiers in the Tertioaugustani (ND Oc. 5.254)the former frontier legion listed in the field army-were still Africans but were from the most Romanized part of the community.

What changed this frontier policy in practice were two accidents that are linked. The first was local usurpations and wars of succes sion. On the Rhine frontier this perhaps started with Postumus in 259, but the most publicized example, which probably introduced the greatest number of transborder soldiers into the frontier forces, was that of Magnentius's challenge to Constantius II in 350. Born of barbarian parents settled in Gaul, he rose high in imperial service and recruited Keltoi and Galatai. But his most enthusiastic followers, according to the emperor Julian himself, "were by virtue of their ties of kinship Franks or Saxons, the most warlike of whom live beyond the Rhine and along the shore of the Western Sea" (Or. 1. 34c-d). At the same time, just to add to the complications, the orator Libanius tells us that Constantius had invited "barbarians" (almost certainly other Saxons and Franks) to cross the borders to make life difficult for Magnentius. One cannot miss the implications of an army-or armies-drawn from both sides of the frontier. It was suggested many years ago that the unusual number of gold coins in Westphalia dating 365-70 could be linked to this event

. The Keltoi, who were probably "Celtic" Germans, and Galatai (Gauls of unknown origin) had served in brigades together in the Roman army since Constantine (Jul. Ep. 4.579). But here we come to the second accident-the formation of the comitatenses by Constantine, as a huge field army that fought the wars of succession from 306-24. To do so he stripped the Rhine frontier, which was his power base, of many of its frontier units, to which were added any local recruits that could be mustered. The local manpower included barbarian units under their ethnic leaders, such as Bonitus the Frank and Crocus the Alaman. This is all well known and is proved by the privileged position subsequently occupied by the Franks, especially their "kings," at Court (Amm. Marc. 15.5.11) and by the unusual grant of ius matrimonii, the right of legal marriage with Roman citizens, to them alone of foreign troops.

From this central field army evolved-probably under the sons of Constantine-the regional comitatenses; that is, different field armies stationed in various dioceses of the empire. This was the origin of a Gallic field army, which incorporated, of course, many of Constantine's Rhine frontier recruits. Their units could now be I stationed at strategic garrisons, like Cologne, on the frontiers and controlling beyond (e.g., Amm. Marc. 15.5.15-16, 28.5.11), or 6 in the provincial hinterland fortifying burgi and supply routes.

In practice, therefore, by the time of Julian we are left in no doubt that many of his units, including those that took a leading part in his usurpation in Paris, were locally recruited with families living nearby, among whom some were explicitly voluntarii from beyond the Rhine (Amm. Marc. 20.4.4; Lib. Or. 18.95). The regular brigading together of German and Gallic units strongly suggests to me that these were soldiers drawn from assimilated populations on both sides of the frontiers, whom Zosimus describes as "a vast army of young men, both barbarians living near the Rhine and farmers from the Roman provinces" (Zos. 4. 12). Many in the army knew the territory across the river and passed on information to both sides about tribal politics or Roman dispositions. Of this we have plenty of examples (e.g., Amm. Marc. 14.10.7, 16.4.1, 16.12.2, 27.2.9, 29.4.7, 31.10.3).

It does not matter here whether the soldiers serving in these armies were those who were called gentiles and laeti- that is, peoples settled within the frontiers-or whether they were soldiers from outside serving by terms of a treaty- those who, after 380, were also increasingly settled within the frontiers under their own kings. It is not always easy to tell the difference between them, and they all derived broadly from the culture of the frontier zones. Nor is it easy to imagine exactly how Goth, Saracen, or Mauretanian chieftains fitted into the imperial court as high officers and officials. But the fact that they did so illustrates how far Romanization had progressed among the elites on both sides of the frontiers. Vadomar, the Alaman king near Augst, is a good example. We are told that he had often seen the Roman army in his youth because he lived near the frontier (utpote vicinus limite), and he ended up in the Roman service as dux Phoenices.

The extraordinary history of the Saracen queen Mavia and her family in the later fourth century shows the same easy interchange. One moment she is ravaging the provinces, the next appointing an Arab hermit, Moses, as bishop of Alexandria, marrying her daughter to a Roman commander, Victor the Sarmatian, and sending her Saracen troops to defend Constantinople. But the lesson of Silvanus, the Frankish magister militum, who was told he would be either killed or surrendered if he fled to his barbarian compatriots in 355 (Amm. Marc. 15.5.15-16), is that Frankish nationalism as such did not exist. It underlines the point I began with, that invasions were launched from a society made up of petty regules, temporarily united in confederations that were riddled with factions (cf. Amm. Marc. 18.2.16 discussing a treaty with the Alamanni).

What is striking at this stage in history is how far the economic and cultural interchange, partially created by the frontier, continued and even increased. To the barbarian soldiers and princes who passed from one side to the other we must add the Roman deserters or "collaborators" who voluntarily crossed into barbaricum, about whom we have many references. There is a significant proviso in a law of 366, which granted returning prisoners of war postliminal rights of recovering their property, as long as they could prove they had not gone to the barbarians of their own free will (CTh 5.7. 1). The huge figures given for prisoners of war who were resettled either in Persia or in Roman lands show that many stayed. In one case we hear of a captured Persian soldier who had begun in the Roman army, was taken prisoner, married a local girt, and then served Persia.

The stability, even stalemate, on the Persian frontier must, I think, be explained by the fact that, after the experiences of Constantius and Julian, there was no real possibility of Rome's regaining the initiative in Mesopotamia or Armenia. From the point of view of frontier assimilation, however, -we have a lot of information about the interchanges that took place-hardly surprising in view of the way territories of the eastern Mesopotamian and Transtigritan states changed hands. The virtual absence of any defined geographic frontier line is underlined by the fact that the so-called Transtigritan states actually seemed to extend to both sides of the river Tigris. This permitted a flow of traffic between Persian and Roman lands that is most interestingly illustrated by the hagiographic sources. They portray holy men or their clients crossing between the two without undue difficulty, though sometimes in disguise as merchants.

Collaborators, fugitives, and prisoners constituted merely one class of transfrontier traffic, which also included traders, smugglers, slaves and technicians, and those who are simply called spies, sometimes posing as traders. Stachao, who came from one of the Mauri gentes contermini of the Tripolitanian limes, is described by Ammianus in 363 as quite normally "wandering freely in our land during the peace" until he was "proved'--unjustly, as the Libyans believed-to have been a spy (Amm. Marc. 28.6.3). Legislation in 323 envisaged the possibility that some provincials might give information to barbarians in order to share the plunder (CTh 7.1.1). So while we may agree with the view that "the course of the major invasions ... was [not] seriously affected by dissident Romans", what we really need to ask is, What about the effect of ordinary Romans and barbarians moving freely across the frontiers? The impression given by contemporary sources is that the traffic had increased considerably and had a major effect on the character of the so-called invasions.

The archaeological evidence of such transfrontier movements is unambiguous. The excavator of Wijster in Holland, north of the Roman frontier, talks of this century as the "golden age" for Roman goods moving through the native center, many of them the buckles, belts, and fibulae typical of the frontier society. In the German Westphalian village of Essen-Hinsel a third of the pottery is Roman, and at Westick in Westphalia the fourth to the fifth century was the most flourishing period for Roman imports. In Alamannia there has been found "an exceptional density" of Roman provincial terra nigra and Argonne pottery, reaching as far as the middle Main and Tauber. The profusion of bronze coins in this region up to 365 shows that it was the consequence not just of raids but of trade too, In old Dacia, now Gothic and Hunnic Dacia, the coin finds show a break just at the point when Valens imposed trade sanctions. Themistius (Or. 10.135-36) assures us about the free trading before Valens.

The third to fourth century was also a kind of golden age for the Tripolitanian gsur, the fortified farms of the chiefs, which are found filled with African Red Polished Ware and monuments of wealth. Beyond the frontier it is not always evident whether the treasures that reached such men came through trading or raiding. At Traprain in Scotland the spectacular hoard of late Roman silver might have been the result of raids, since it was discovered crushed into a heap, as though intended as bullion. But Traprain was the home of the Votadini, long-standing allies of the Romans, and it is more probable that these valuable goods reached them originally in the form of gifts. The buckles in the hoard could have been worn by either Romans or barbarians.

But those who believe that Roman legislation could impose trade blockades as a political weapon for any length of time underes timate the age-old tradition of such exchanges, including smuggling. Herodian's reference to secret imports by merchants over the Persian borders (4.10.4) recalls the parallel I noted earlier about how impossible it was to alter the rhythm of economic exchange in the lies et passeries of the Pyrenees by international treaties. If the Italians in the nineteenth century were unable to stop the huge smuggling of tobacco, sugar, and salt from Switzerland by putting up a fence of wire netting, I doubt if their forebears could have done better. The significant part of the story of Antoninus, the Roman officer who defected to Persia in 359, is that he bought a farm on the very frontier in order, says Ammianus (18.5.3), to avoid suspicion. The implication is inescapable: frontier farmers were expected to cross borders.

One obvious reason for such interchange, which is well documented, is that huge numbers were moving across and settling in the Roman Empire under negotiated terms in the fourth century along the Rhine, the Danube and-though apparently to a much lesser extent-the eastern frontier. It is difficult to estimate the scale of these movements from the vague statements of our sources-, phrases often simply talk about "a lot" (tot translati . . . in rura Romana cultores; Pan. Lat. 8[5]1.4). Ancient figures, where given, are inherently untrustworthy, such as when three hundred thousand Sarmatians are said to have been resettled by Constantine (Exc. Val 1.6.32). But over twenty-five such reports along the Rhine-Danube in the century from Diocletian to Theodosius must indicate a very large number indeed, at a guess well over one million foreigners along a frontier of some ten thousand kilometers. This figure does not include the federate settlements-the enclaves of foreigners under their own leaders-which became common practice after Adrianople. Nor does it take account of the sixteen distinct military units of laeti and gentiles recorded in the lists of the Notitia Galliarum, which though reconstructed is imperfect.

We know, of course, that not all such barbarians were settled precisely on the frontier, and that many were exploited on estates in the interior as laeti and dediticii under near slavelike conditions (CTh 10.10.2514081, etc.). But judging by what we are told of the Goths in the period before Adrianople and of the way they were distributed, we can be reasonably certain that many were settled in the frontier zone or in the vicinity of the frontier. Information was given to invaders in 378 by Goths who had earlier been sold as slaves in the Roman province (Amm. Marc. 31.6.5-7).

The truly remarkable fact about these huge, peaceful shifts of population is how difficult it is to see any real sign of them in the archaeological remains. They disappear "like ghosts" into the countryside. That is to say, from the point of view of artifacts, they became rural provincials. Naturally we can talk about the end of Roman villas and of new styles of weapon-burials. Many villas certainly ended violently. But the picture of wholesale abandonment of villas in Belgium and northwestern France has recently been considerably modified, since it is dear from archaeology that many were occupied in the fourth century, even if only by squatters when the owners had apparently left. in some cases desertion of rural sites in Picardy and even those near Trier had apparently begun in the late second century, long before the invasions.

And who can honestly detect the presence of laeti or gentiles among the peasants who remained in many of the vici adjoining the estates after the villa had collapsed? The nearest we can come is to identify wooden huts on the villa sites and Germanic handmade pottery or buckles and fibulae, which look like evidence of foreign settlers. In just a few cases in northern Gaul, such as the villa site of Donk (Limburg) in Belgium, we find unmistakable signs of Germans in the dugout huts or houses, the Grubenhauser, and in long wooden houses with stables, the Wohnstallhauser - the kind of houses found in free Germany. just occasionally, as at Haarf or Froitzheirn in the Rhineland, these huts date to the third century, although most begin after the mid-fourth century. In truth it is usually impossible to tell whether the artifacts are evidence of intrusive invaders, peaceful settlers, or even Gallo-Romans who had adopted the new frontier culture.

North and south of the lower Rhine and Waal in the late fourth century much of the pottery is Identical, even if Roman fabrics were fewer to the north. The Goths, as we believe from the appearance of the Sintana de Mures culture in the Transylvanian Alps and Wallachia in the late third century, had lived side by side with Romano-Dacian settlements, absorbing their culture, for over a hundred years before they crossed the Danube frontier in the later fourth century. They did not appear on the frontiers as raw barbarians, and once inside they cannot be traced archaeologically.

A good example of this kind of assimilation has been found by two Hungarian archaeologists working, on a graveyard at Mozs, north of Szeksgard. They discovered a group of graves they believe belonged to three generations of four families. The earlier bricklined graves were clearly Roman in tradition, some of them containing bird-head buckles. But at the same time the skulls were deformed in a manner associated with the appearance of Goths, Huns, and Alans. The later generations abandoned their Romanizing habits and adopted the artifacts associated with the period when Attila controlled the Roman province of Valeria on the Danube. Perhaps, the authors suggest, we can see here some evidence of how federated settlers from beyond the frontiers kept their religious customs but easily adopted Roman ways. In the Visigothic kingdom of Aquitania and Septimania even their distinctive art in the fifth century is judged to be exclusively the development of local Gallic artists. The settlements of Visigoths, Franks, and Burgundians in the fifth century, which are marked on maps, are really not distinguishable in terms of new cultures coming into Gaul from beyond the frontiers.

This is the conclusion to be drawn from the massive literature on inhumation graves containing weapons, throwing axes, belt buckles, and fibulae that appear in the later Empire of the West. A good deal of revision of long-held views about the character of the barbarian "invasions" has now taken place as a result of recent studies of these grave goods. We can no longer maintain, for instance, that these are all the relics of barbarian soldiers or paramilitary laeti and gentiles, since there are children's graves among them also containing swords and throwing axes. They have a strong military idiom but often appear to belong only to families of high status. Nor can we be sure they are all graves of Germans, since in Gaul many of these burials appear to be perfectly integrated in provincial communities, using the same graveyards without signs of disruption. Only occasionally do the burials indicate a discontinuity in the inhabitants of the villa.

On the other hand, it is impossible to deny their close Germanic connections with similar artifacts in Saxon, Frankish and Alamanic burials found beyond the frontiers, particularly in the case of women's graves. This applies to various types of dress brooches (the tutulus trumpet-shaped brooch, or the crossbow-shaped brooch), whose prototypes appear to have evolved in an area between the Elbe and the Weser. But so too was there a close Roman link, since belts, buckles, and military insignia were typical of both military and civilian officials of the late imperial army and court. And it is an interesting fact that weapon burials appear to be confined to people inside the Roman Empire and not to have been practiced by free Germans.

In other words, we witness here the development of a frontier culture among peoples who were closely associated with Romans, possibly through military service. Fibulae, like the crossbow type, seem to be used by the Gothic Sintana de Mures settlements. And it is certain that some of the goods like the dolphin buckle or the swords were products of Roman manufacture. The best illustration of this assimilation comes from a typical buckle of this type, found in a "Germanic" burial at Landifay (Aisne) in the north of France. But engraved on the dolphin buckle is the portrait of the man dressed like a Roman cavalry officer and his wife, dressed in Roman style without the fibulae normal for a German woman. That shows just how difficult it is to make any cultural separation between Germans and Romans in this period.

Quite a number of the graves and artifacts do, however, appear to be linked with Roman military settlements. just north of Valentinian's imperial capital of Trier is the famous Langmauer, a huge construction enclosing about 220 square kilometers of land that was almost certainly an imperial estate. Within the estate are two settlement sites where belt buckles, fibulae, weapons, and Germanic types of nonwheel pottery have been found. Plausibly these were the billets of German units serving in the imperial army of the emperor's garrison at Trier.

Easily the most remarkable military evidence comes from the huge necropolis at Krefeld-Gellep at the Roman fort of Gelduba north of Cologne, where about five thousand graves have been excavated. The majority and the richest date from the mid-fifth century onward, some of them clearly princely, suggesting that it became a center of Frankish power. But the most relevant fact here is that the necropolis demonstrates a continuity from Roman to Frankish periods, with a period from the mid-fourth century when the new style of "German" burial appeared and a new orientation of graves took place. There seems little reason to doubt that this is evidence of federate troops' taking up their station, although we must note that before the mid-fifth century there are few graves with weapons and many containing Roman pots. The change was not sudden.

Less dramatic signs of possible military units, whether of federates or of laetic and gentiles units, have been found at other sites like that at Vermand (Aisne) or Oudenburg (Belgium). Let me stress again, however, the point made by archaeologists. The new practices of burying weapons, clothing, and jewelry went hand in hand with the old Gallo- Roman grave goods of pots, food, and coins (for Charon). Cemeteries arranged in rows, the so-called Reihengraberfeld, which were sometimes (but not always) a distinctive Germanic feature, occurred sometimes also in the later Roman period. In short, all these graves are evidence of what Bohme called a Mischzivilisation, although perhaps not as he adds "specifically Gallo-Roman." They are the specific product, rather, of the frontiers where cultures met.

But it is not only in the military sphere that this change was taking place. There is extensive evidence of graves and of settlements in a more rural or civil context. In northern Gaut and in the Lower German province the local population clearly lived side by side with new settlers without constant violence. At the villa of Famechon (Somme) in Picardy the villa urbana section, where the rich landowner used to live, was abandoned, but the rustica section continued to be occupied. At some of the large Rhineland Alas, like that at Konz or Leiwen, the urbana sector was transformed into baths in the later fourth century; in others the baths were used as kitchens and bakeries or as pottery kilns, and metalworks were built in the old cellars. But often alongside these peasant "squatters" there appear sunken-floor wooden buildings and wood longhouses, as in the villa at Donk in Belgium. So, as poor farmers and peasants were abandoning their Roman heritage, Germanic settlers were moving in alongside them, particularly in the later fourth century.

In Britain, where the frontier of penetration had effectively become the Saxon Shore, there has been a lot of controversy about whether Saxon settlers overlapped with Romano-British villa owners in the fourth century. As along the Rhine, foreign troops were stationed in and died in the country, and evidence of their graves can be seen in Roman cemeteries, such as that at Lankhills near Winchester or in the late cemetery at Mucking in Essex in southeastern Britain. But it is impossible to identify them as Saxons rather than as soldiers wearing the typical late- frontier style of dress and practicing inhumation burial rites for themselves and their wives, which makes them indistinguishable in archaeological terms from other German groups like the Franks. Their cemeteries at Dorchester- on-Thames, Milton Regis, and other late Roman sites along the upper Thames valley suggests they may have been employed as federates by the British civitates. If so, they would conveniently date from the first decade of the fifth century when the last Roman troops had been removed in the wake of an attempted usurpation from Britain by the pretender Constantine.

At some sites, such as Kelvedon near Colchester, Roman and Saxon artifacts are found side by side in fifth-century graves. At the rural site of Orton Hall Farm in Oxfordshire the Anglo-Saxon houses and artifacts seem to respect the earlier Roman buildings. Some early Saxon settlement sites, such as those at Mucking in Essex or at Catholme in Staffordshire, also seem to occupy the edges of Roman villas, which might mean that for a time Saxons lived symbiotically with the British owners.

But it is all very inconclusive, and in general most Saxon evidence, when it can be dated, is now thought to date from after the end of the Roman province and often after a gap in occupation. Most important, these later settlers practiced cremation; they did not bury their weapons, and they possessed a completely unRoman material culture that can be compared with that from northern Germany, Jutland, and southern Scandinavia. Almost certainly they were Saxons, Angles, and Jutes who had had very little contact with Roman frontiers and came in a second phase of settlement. Perhaps the most significant frontier lessons to be learned from Britain are counter factual. The first is that, unlike the Rhine or Danube provinces, the seacoast of the Saxon Shore did not permit steady infiltration and assimilation, so that the break, when it came in the mid-fifth century, was far more radical and the continuity with Roman culture far more superficial than elsewhere. The second lesson, as one recent writer points out, is that "the collapse of RomanoBritish culture ... cannot be attributed to a violent Anglo-Saxon settlement." There was continuity. But by the time the Saxons settled, the post-Roman British had already thrown off the veneer of Roman life and reverted to a subculture of rural subsistence and petty chieftainships that adapted quickly to Anglo-Saxon society. That is something we are only just beginning to understand.

In the rest of the western empire the cemeteries add particularly to our general knowledge of the social changes that were taking place on both sides of the frontiers. Almost all the studies of such burials describe them as being clustered in groups. The rural cemetery at Abbeville-Homblieres (near Saint-Quentin), for in stance, began in the mid-fourth century near the site of an abandoned Roman villa and contains a number of weapons burials. But there are three groups of particularly rich graves, separated from the others, and it has been argued that this is evidence of a German-owned villa worked by Gallo-Roman labor. in the cemetery adjacent to the Roman villa at Frenouville, near Caen in Normandy, it is impossible to detect a break in continuity of the population from the late third to the late seventh century, except in the mode of burial and in the orientation of the graves in the mid-fifth century.

The pattern is repeated all over the north-Vert la Gravelle in Champagne, Pry in Naumur, Mezerny in the Ardennes- small groups cups of graves belonging to twenty or thirty persons, including three to six rich graves containing weapons or women's fibulae. Sometimes they give the impression of being "familial concessions" or signs of noble Germanic families' taking over villas with their Gefolgsmanner while surrounded by "the local, rural, Gallo-Roman population who remained where they were". Small groups of graves arranged like this are also found in cemeteries beyond the frontier. At Kirchheim bei Munchen in Bavaria a cluster of seventeen graves is thought to be those of the inhabitants of a rural villa. At Rhenen near Utrecht a very large necropolis of mixed cremation and inhumation beginning in the fourth century contains graves arranged in three groups that are thought to represent social "clans".

The evidence of cemeteries and settlements, therefore, is broadly the same in its conclusions. just as the northern German settlements were nucleated and became increasingly differentiated socially, so now too a similar kind of community started to develop south of the frontier. But it was not just because Gallo-Romans were being displaced by German populations, although it looks as if that had happened in the mid-fourth century in Toxandria. But even among the later Roman communities smaller farms were disappearing while a few larger villas-Rheinbach-Flerzheim in the Rhineland, Saint-Ulrich near Metz, or Echternach in Luxembourg---survived and were fortified, often with protected grain stores. Many vici were abandoned or reverted to Iron Age hilltop sites and ramparts, but peasants at some sites appear to have taken refuge on the estates of the rich. Overall in Belgica, concluded a study a few years ago, "an increasing militarization of rural communities is clearly visible in the second half of the fourth century".

That statement must be qualified somewhat in the light of new information. The overriding impression of later fourth century developments in the north is that most villa owners departed, with a few notable exceptions like those cited above. But the peasant labor-or more probably tenants-moved in to share the villa buildings, obviously for protection. The same process has been observed on British villas, where one family among the peasants seems to be the richer. The increasing nucleation meant greater social control by the powerful. That is a theme I shall take up in the last chapter.

A "Barbarian" View of Frontiers

I have one last thought that may help put the frontiers of the later Empire in a different perspective and that anticipates some of what I will discuss in the next chapter. The trouble with all our frontier studies is that they are made from the inside outward. At the beginning of this chapter I wondered how those whom Romans called "barbarians"-those beyond the frontiers-would have viewed the Roman frontiers from outside.

In the fourth and fifth centuries we witness the reverse process of the conquests that had begun with Julius Caesar and Augustus. The frontiers of the Roman Empire in the early Empire were administrative boundaries in an "area inviting entrance" (to use a phrase of Walter Prescott Webb's), even though often portrayed like a rampart ringing the Roman world (Ael. Arist. Ad Rom. 82). Now they seemed to take on the same quasi-juridical character from the barbarian point of view. Valentinian was forced to negoti ate peace with the Alaman king on a ship in the middle of the Rhine, a symbolic but artificial compromise between two powers with conflicting territorial claims. Valens did the same on the Danube with the Goths (Amm. Marc. 30.3.3-6, 31.4.13). These are two of many examples of formal treaties recognizing the rights of outer gentes, in exactly the same way that the Romans had earlier recognized the Parthians by meeting, them on the Euphrates (Vell. Pat. 2.101; Tac. Ann. 2.58; cf. Amm. Marc. 27.5.6; Lib. Or. 18.88~ Greg. Tur. HE 2.9, etc.).

But just as Roman ideology-as opposed to practice-had claimed and never ceased to claim control beyond the formal termini, the reverse began to happen in the fourth and fifth centuries in respect to the barbarians. It was now the barbarian kings who permitted crossing of the rivers as a privilege of favored allies. The king of the Rugi permitted Roman traders to visit markets beyond the Danube from Noricum (Eugip. V. Sev. 9.1, 22.2), and it was the Huns and their boatmen who controlled the crossing of the lower Danube at Ratiaria.

Meanwhile, barbarian political power extended deep into Roman territory. The Alamans in 352 devastated a band of territory 180 kilometers beyond the Rhine and claimed the right to settle a zone 60 kilometers broad on the left bank (Jul. Ep. ad Ath. 279a; Amm. Marc. 16.12.3). The "core" of Frankish power, judging by gold hoards and princely burials, seems to have been the Rhine itself, not farther north, which means that they too extended well to the south. A Roman embassy wishing to visit Attila, whose kingdom theoretically lay well beyond the Danube, was met by a band of Huns long before reaching the river, in "Roman territory," says Priscus, who was himself present. But then he adds that the region was used by Attila for hunting, just as part of Roman Pannonia too was said to be subject to him.

So though the Romans still claimed to exercise political power over Germans beyond the great rivers, we must try to imagine what the Roman world looked like from the German point of view. Roman ideology still regarded access to barbarian territory as open and unlimited by formal frontier lines. But barbarian kings were coming to adopt the same view in reverse. The Germanic Rugi and Alamans still recognized the Danube as a Roman frontier but levied tribute and disposed of villages and castella south of the river, as Eugippus's Life of Saint Severinus in the fifth century so vividly describes (8.2, 22.4, 31.1, 31.6, 42. 1). Attila had forced the Romans to pay tribute (phoros), but, so we are told, it was concealed by the Roman court in the traditional manner as if it were payment of subsidies and supplies, in recognition of which they gave Attila the title magister utriusque militiae in the Roman army. Nothing could illustrate better differences in perspective of those who lived on each side of the frontiers. We see in this example the perfect contrast between ideology and reality.

Please enter your name:

and your E-Mail address:

Type in here which week this assignment is for:

Please answer the follwoing question in the space provided; then submit. Please type in the question first;follow it with your response.
1. To what extent did Roman frontiers, east and west, furvive the barbarian "invasions?" How did ideology shape the Roman response to these incursions during the fourth and fifth centuries?