IN THE SUMMER OF 357 AD, a huge army of Germani, led by various kings of the Alamanni, collected itself on the western, Roman, side of the River Rhine near the modern city of Strasbourg. As Ammianus Marcellinus, in the most detailed narrative history to survive from the later Roman period (c.275 onwards), reports it:
All these warlike and savage tribes were led by Chnodomarius and Serapio, kings higher than all the rest in authority. And Chnodomarius, who was in fact the infamous instigator of the whole disturbance, rode before the left wing with a flame-coloured plume on his helmet . . . a doughty soldier and a skilful general beyond all the rest. But the right wing was led by Serapio . . . he was the son of Mederichus, Chnodomarius’ brother . . . who had for a long time been kept as a hostage in Gaul . . . These were then followed by the kings next in power, five in number, by ten princes [regales], with a long train of nobles [optimates], and 35,000 troops levied from various nations, partly for pay and partly under agreement to return the service.
Ammianus’ description beautifully captures the absence of unified kingship among the Germanic Alamanni who dominated the southern sector of Rome’s Rhine frontier in the late imperial period. On the basis of it, indeed, historians have sometimes argued that little had changed in the Germanic world since the first century AD, when Cornelius Tacitus wrote his famous gazetteer. One of the central points brought home by even the quickest read of Tacitus’ Germania is just how fragmented, in political terms, the Germanic world was at that date. His work – and in this it is well supported by the slightly later, second-century, Geography of Ptolemy – records far too many primary political units to list by name: well over fifty. The best thing to do is place them on a map, which, even if the geographical placements are approximate, gives an excellent sense of first-century Germanic political fragmentation (Map 2).1 A closer look suggests, however, that it is a mistake to suppose that, just because the Alamanni had plenty of kings in the fourth century, nothing important had changed since the first.
A first indication of the extent of the change emerges from a quick glance at a situation map for Germanic Europe in the mid-fourth century, which looks very different from its counterpart in the time of Tacitus (Map 3). The south-east has a completely different complexion, with the rise to dominance of various Gothic groups in, around and to the east of the Carpathian Mountains. In the west, too, much had changed. In place of the multiplicity of smaller units known to Tacitus and Ptolemy, four larger groupings dominated the landscape on and just behind Rome’s Rhine frontier: Alamanni and Franks on the frontier line, Saxons and Burgundians just behind. In trying to understand the workings of these fourth-century entities, we are, as usual, struggling against the general lack of interest in things ‘barbarian’ among Roman authors, but, thanks overwhelmingly to Ammianus, we have much more evidence for the Alamanni than for the rest. And, as Ammianus’ narrative makes clear, politics among the Alamanni were complex.
Politics Transformed
In the surviving books of his history, which cover the years 354 to 378 in considerable detail, Ammianus never offers us an analytical account of how politics worked among the Alamanni, or of the institutional structures that sustained rulers in power. What he does provide is a more or less connected narrative of a quarter-century of the Alamannic confederation, as it can best be called, in action, from which it emerges clearly that Germanic politics in this part of the Rhine frontier region had seen two fundamental transformations since the time of Tacitus. First – and this much is uncontroversial – leadership within the different groups that made up the Alamanni, those headed by the various kings and princes who showed up for Strasbourg and who feature elsewhere in Ammianus’ narrative, was more solidly constructed than had been the case in the first century. Tacitus’ works – not only his Germania, but also the narratives of his Annals and Histories – give us a fair amount of information about pretty much the same regions of Germania in the first century. At that date, some Germanic groups, such as the Usipii and Tenchteri, functioned entirely happily without any kings at all: group policy, where necessary, being decided by an oligarchy of leading men (Latin, principes) discussing matters together in council. And even where royal authority was achieved by a single individual over a particular group, this was never accepted without resistance and usually proved transient, power not passing on to a designated son or heir. In the second decade of the first century AD, the two dominant monarchical figures of contemporary Germanic politics were Arminius among the Cherusci and Maroboduus among the Marcomanni. Arminius’ dominance was extremely brief. It was based on leading the famous revolt that had destroyed Varus’ three legions at the Teutoburger Wald in 9 AD, but he was only one of several leading figures among the Cherusci. Victory gave him a brief pre-eminence, but it was always contested, not least by Segestes, a second leader of the Cherusci who even aided the Romans. Arminius’ power was already ebbing away, in fact, before his death in 19 AD at the hands of a faction of his own countrymen. Maroboduus’ dominance had deeper roots, but was eventually undermined both by Rome and by internal rivals, so that in the time of Tacitus, at the turn of the second century, the Marcomanni were no longer ruled by Maroboduus’ heirs.2
Political life among the fourth-century Alamanni, by contrast, was all about kings (reges) and princes. The evidence we have suggests that the region called Alamannia by Ammianus was divided into a series of cantons or sub-regions (the Germanic for which was probably already gau), each of which (or all that we know of) was ruled by a rex or regalis. This royal power seems also to have been at least partly hereditary, if not necessarily in a simple pattern from father to son, then through royal clans. Chnodomarius and Serapio, the dominant leaders at Strasbourg, were uncle and nephew, and Serapio’s father Mederichus had been important enough to enjoy a lengthy spell as hostage among the Romans, during which time he developed a penchant for the cult of the Egyptian god Serapis, which led to his son’s strikingly non-Germanic name. In Ammianus, we also come across a father and son, Vadomarius and Vithicabius, each of whom was king in turn. It is important not to generalize unwisely. Alamannic kings could be overthrown. One Gundomadus was killed by his own followers because he wouldn’t join the army that fought at Strasbourg. Likewise, some people other than kings remained important in Alamannic society. Ammianus, again, picks out the optimates who were present at the battle. Nonetheless, kings are a much bigger and more stable presence among the fourth-century Alamanni than they had been in the early Roman period.3
Second, the Alamannic confederation as a whole operated as a much more solid political entity than its first-century counterparts. This is a more controversial point because, as we have seen, the Alamanni did not function as a centralized entity, with a single, unchallenged ruler. There never was just a single king of the Alamanni at any point in the fourth century. And first-century Germania had itself been perfectly capable of throwing up larger confederations, incorporating a number of the small primary political units. For some scholars, therefore, the evidence does not suggest substantial change. But the supratribal confederations of the first century were either longer-lived (though not unchanging) and primarily religious in function, or else, where they were political, highly transient. Tacitus mentions three ‘cult leagues’ (groups of tribes who shared an attachment to a common religious cult in addition to their own individual ones): the Ingvaeones, nearest the sea, the Herminones of the interior and the Istvaeones of the west. We don’t know all that much about them, and I would not want to underestimate their overall importance to the early Germanic world, since Ptolemy knew of them too, meaning that they persisted on through the first century into the second. But the narratives of attempted and actual Roman conquests in the region demonstrate that the cult leagues never operated as the basis for political or military response to outside assault. Whatever their significance – and it might well have been substantial in other areas of life – the cult leagues were not political organizations of importance. When resistance to Rome or – at a slightly earlier period – attempted Germanic expansion into the Celtic-dominated world took a confederative political form, it was organized around dominant individuals: Ariovistus in the time of Caesar, Arminius and Maroboduus at the time of projected Roman conquest of the area between the Rhine and the Elbe; or, later, the great revolt organized by the Batavian leader Julius Civilis. These leaders all knitted together – using a mixture of attraction, persuasion and intimidation – large confederations, which drew on a warrior base from across extensive reaches of the Germanic world; that is, from many of the small political units listed by Tacitus and, later, Ptolemy. In every instance, however, these confederations crashed with the defeat of their leaders, and are never heard of again. That of Maroboduus lasted a little longer than the others, but even that unravelled quickly after his death.4
It is on this level, the stability of the overall confederation, that a comparison between the first century and the fourth is so telling. The outcome of the battle of Strasbourg was a thumping defeat for the Alamanni:
There fell in this battle on the Roman side two hundred and forty-three soldiers and four senior officers . . . But of the Alamanni there were counted six thousand corpses lying on the field, and heaps of dead, impossible to reckon, were carried off by the waters of the river [Rhine].
Chnodomarius himself was captured as he tried to flee back across the river. The Caesar Julian, ruling in the west in the name of his cousin Augustus Constantius II, then exploited victory to impose terms of his own choosing on the various Alamannic kings who survived the engagement; and, in fact, Chnodomarius had only been able to assemble his forces with such freedom because a Roman civil war had generated a power vacuum in the Rhine frontier region in the first place. Nonetheless, and this is where the fourth century is completely different from the first, the defeat of Chnodomarius did not mean the total destruction of the alliance at whose head he had stood, as the defeats of his first-century counterparts such as Arminius and Maroboduus had done three centuries before. Not only were many of the lesser Alamannic kings who had participated in the battle left in place by Julian’s diplomacy, but, within a decade of the battle, a new pre-eminent leader, Vadomarius, was worrying the Romans. He was skilfully removed by assassination, but then a third appeared in his place: Macrianus. Ammianus records three separate attempts by one of Julian’s successors, Valentinian I, to eliminate Macrianus by capture and/or assassination, but eventually, pressed by events further east, the emperor gave in. Roman and Alamann met in the middle of the Rhine for a water-borne summit, where the emperor acknowledged Macrianus’ pre-eminence among the Alamanni.5 Unlike in the first century, even major military defeat was not enough to destroy the larger, Alamannic confederation.
This suggests that the confederation had a political identity that was much more firmly rooted than its first-century counterparts, in the sense that it did not rise and fall with the careers of single individuals. As circumstances changed, one canton king or another might rise to pre-eminence, but the confederation as a whole could survive the vagaries of individual political careers more or less intact. The strength of these ties is also suggested by some of the nuggets of information Ammianus preserves, not least about the canton king Gundomadus who was overthrown by a faction of his own followers for not participating in the larger group action that led up to Strasbourg. For these men, at least, group identity could at times be a more powerful determinant of political behaviour than loyalty to their local king. How this group identity worked, Ammianus does not tell us. He does report that the kings of the Alamanni feasted each other, and that ties of mutual support bound together at least some of the kings who fought at Strasbourg. In the details of such agreements and of royal feastings would lie the information we would require to understand the fourth-century Alamanni properly, but Ammianus, unfortunately, does not tell us what we need to know.
Like many late antique and early medieval confederative entities, the Alamanni had, I suspect, an established repertoire of political and diplomatic conventions which defined and bound together their various kings in positions of overking and underking, the latter owing allegiance and some duties to the former, while still retaining direct day-to-day control of their own cantons. In these kinds of systems, political continuity could never be absolute. No overking exactly replicated, nor could he usually directly inherit, the patterns of power enjoyed by a predecessor; but, once a new pecking order had been established, there was an accepted type of relationship between kings of varying statuses that could be used to orchestrate and define the rights of both parties – senior and junior – to any new agreement. Such a system was clearly operating, in my view, among the fourth-century Alamanni, and a visible sign of its overall importance is the general ‘shape’ of Roman diplomatic policy on this part of the frontier. Whenever Roman attention was distracted, usually by events on the Empire’s Persian front in Ammianus’ time, an Alammanic overking would duly emerge, and Roman policy on the Rhine was largely directed towards removing the succession of such figures who appeared in the course of the period he covers.6
Unfortunately, again, Ammianus does not give us any indication whether similar systems operated among the other large entities of the Rhine frontier – Franks, Saxons and Burgundians. Like their Alamannic neighbours, the fourth-century Franks certainly had a plethora of kings, but we simply do not see them in action often enough to know whether a Frankish political identity could, likewise, act as the basis of collective action even after the shock of heavy defeat. And there is no reason to suppose that all fourth-century Germanic groups had to operate on exactly the same basis, any more than their predecessors of three hundred years earlier had done, when, as Tacitus records, some groups had kings and others did not. Confirmation that broader and politically more solid group identities were not confined merely to the Alamanni in the fourth century is provided by the Tervingi, a Goth-dominated confederation which operated at the other, eastern extreme of Rome’s European frontiers in the foothills of the Carpathians. The Tervingi are the one other group amongst Rome’s Germanic neighbours, apart from the Alamanni, about whom the sources preserve a substantial amount of information.
In its political operations and durability, the confederation of the Tervingi shows three characteristics which strongly resemble the Alamanni and firmly distance it from any first-century ancestor. First, central control of the Tervingi seems to have been handed down through at least three generations of the same dynasty between c.330 and c.370, and their official title was ‘judge’. As was the case with the kings of individual cantons among the Alamanni, therefore, power in this eastern Germanic world had become much more hereditary. Second, also like the Alamanni, the judges of the Tervingi ruled a confederation, which involved a number of kings and princes. And third, the Tervingi confederation was bound together by ties strong enough to survive even heavy defeat. We first encounter the coalition in the early 330s, when a massive defeat was inflicted upon it by the Emperor Constantine. Not only did it survive that defeat, but the same dynasty retained power and, a generation later, plotted to overthrow the most burdensome aspects of the terms that Constantine had imposed.7 It is important to stress that the Alamanni and Tervingi are the only two fourth-century Germanic entities about whom we are at all informed, and that you cannot just assume that every large Germanic grouping of the period worked the same way. Between them, however, the two cases provide excellent evidence that larger and more coherent group identities had emerged in fourth-century Germania than could have been found anywhere within its limits three hundred years before.
How had this come about?
The Rise of Military Kingship
This is not a story that can be told directly. No major narrative sources survive between the first and fourth centuries to give a detailed account of any aspect of Romano-German relations in the crucial intervening period. Even such a major convulsion as the second-century Marcomannic War has to be reconstructed from fragmentary evidence. In any case, it is doubtful that any – even lost – Roman historian would have covered a broad enough time frame to be able to chart the long-term transformation that culminated in the Alamanni and their contemporaries the Tervingi. The first-century sources document plenty of power struggles between tribes. We even hear of whole tribes being created and destroyed. The Batavi, for instance, were originally an offshoot of the Chatti, while Roman observers witnessed the destruction of the Bructeri, and Tacitus tells us about a fight to the death between the Chatti and the Hermenduri and the eventual destruction of the exiled and unfortunately landless Ampsivarii.8 Sometimes, too, if less frequently, we even hear of power struggles within tribes, not least that between Arminius and Segestes for control of the Cherusci. But there’s nothing in these bits and pieces of information that would lead you to think that Germanic political structures were heading off on a journey towards greater size and coherence. The most dramatic clue as to the kinds of process that really underlay their appearance emerged from one of the least likely places imaginable.
In 1955, a group of Danish workmen were cutting a drainage ditch at Haderslev in the northern Schleswig region of southern Jutland. Their work quickly came to a halt, however, when one small stretch of their ditch produced an astonishing haul of six hundred metallic objects, many datable to the Roman period. The low-lying meadow where they were working had in ancient times been a lake, if not a particularly deep one. Over the next nine years 1,700 square metres of meadow were carefully excavated, the site producing a whole series of startling finds, not least the remains of a boat. All of these materials had been dumped in the lake at different points in the Roman period, the clustering of objects showing that, on occasion, literally mounds of them were deposited at one go, emptied out from bags or baskets. This was by no means the first Germanic dumping-ground to be excavated. In the later nineteenth century a whole series of north European, particularly Danish, bogs had produced similar clusters of material. But Ejsbøl Mose, to give the Haderslev site its proper name, was the first of these sites to be excavated using modern archaeological methods. This made it possible to answer the big question left unanswered by the earlier digs. Had these dumping-grounds been created by successive small deposits, or a few much bigger ones?
With careful attention to stratigraphic detail, the answer emerged loud and clear. The items found at Ejsbøl Mose had been deposited at several different moments, but, occasionally, huge amounts of material had been sunk at one time. In particular, the excavators were able to identify as a single, unitary deposit, the entire military equipment of a small army of about two hundred men, which had been submerged in the waters at one go somewhere around the year 300 AD. Amounting to many hundreds of individual items, the equipment turned out to belong to a coherent, well-organized force with a clear leadership hierarchy. It comprised close to two hundred spearmen, each armed with a barbed throwing javelin and a lance for thrusting; the excavators found 193 barbed spearheads and another 187 barbless ones. Something like a third of the men also had side-arms. The excavators found 63 belt buckle sets, with 60 of the swords and 62 of the knives that the belts had originally housed. The military force had been led by ten or more commanders on horseback. Ten bridles and seven sets of spurs were all part of the swag.
Interestingly, all of this equipment had been ritually destroyed before being sunk in the lake. The swords had all been bent out of shape, and many fragments of wood came from smashed spear hafts. The obvious violence of the process makes it impossible not to associate the remains with the kinds of ritual act occasionally reported in the historical sources, whereby the weapons of an enemy were offered as a sacrifice to the gods.9 One or two horsemen may have escaped on foot, or maybe the missing spurs were just lost. Essentially, though, the excavators had found the last material remains of a military force wiped out in some long-forgotten, entirely unrecorded Vernichtungsschlacht (battle of annihilation) from the turn of the fourth century.
As an archaeological set-piece, Ejsbøl Mose is fantastic, but the finds have a broader significance. The clear image that emerges from them – of a professional, well-organized force with a well-structured hierarchy – coincides with a considerable body of literary evidence that, by the fourth century, Germanic leaders of royal rank – kings – had personal and permanent establishments of household warriors, on precisely this scale of magnitude. When the Romans eventually cornered him after Strasbourg, Chnodomarius’ retinue surrendered as well as their leader himself. Coincidentally, it also numbered two hundred men. These retinues had an obvious military function, but a few precious indications confirm what you would otherwise have to suppose, that they were also employed more generally as an instrument of social power. When the leaders of the Tervingi decided that they would attempt to enforce uniformity of belief among their subjects in the early 370s, retinue members were sent round to Gothic villages to demand compliance. The key point here is that Tacitus reports the existence of no institution of this kind for the first century. Retinues and warbands existed at that time, but they were not permanent, and prominent individual leaders received only occasional voluntary donations of food for the upkeep of the men in their service. Archaeological material from this earlier era has also thrown up nothing like the professional, variegated weaponry uncovered at Ejsbol Mose. In the intervening two centuries, Germanic kings had begun to dispose of an entirely new level of permanent military muscle.10 This immediately explains, of course, why they themselves should appear in our fourth-century sources as a much more permanent and prevalent fixture of Germanic society than their counterparts from the time of Tacitus.
Further striking testimony to the importance of this development has also emerged from a totally different quarter. One of the more exotic and demanding disciplines within the field of humanities is comparative philology – the study of the linguistic origins of words and meanings, together with their transfer between different language groups. As a recent study has demonstrated, all Germanic languages derived their terms for ‘king’ or ‘leader’ from just three root words: thiudans (‘ruler of a people’), truthin andkuning. Of these, thiudans is certainly the oldest, being the only one with parallels in other Indo-European languages, but the pattern of its distribution across the different branches of the German language family also shows that it was falling, or had fallen, out of use by the late Roman period, when it was being replaced by truthin. Kuning came into currency only later. The striking point is that truthin originally meant ‘leader of a warband’, but by the late Roman period had come into use as the main term for ‘king’ or ‘leader’ right across the Germanic world. There is much more to this than merely a change of name. Thiudans meant ruler of a people, for whom any military function was only part of the job profile, and perhaps only a relatively small part. Famously, Tacitus remarks of Germanic societies of the first century that ‘they chose kings for their nobility, war leaders for their courage’, which seems to imply as much. By the fourth century, the new leadership terminology indicates that this distinction had disappeared, and that military command had become the primary function of contemporary Germanic leaders. It is hard to think of better testimony to the overwhelming importance of the rise, by the late Roman period, of a new kind of leader, who owed the strength of his position to having at his beck and call a permanent body of warriors.11 Archaeology, literary sources and philology all come together to bring to light the roots of the more solidly founded form of kingship that we meet in the fourth century.
What transpired between the first and fourth centuries, then, was broadly this: a class of military leaders developed a new kind of military muscle, and used it to put greater distance, in terms of social power, between themselves and everyone else. It doesn’t take more than a moment’s reflection to realize that this could never have been an entirely consensual process, since a small elite was busy asserting its dominance over everybody else. And this, of course, provides one possible context for the events that culminated in the weapons deposition at Ejsbøl Mose. What the archaeologists found there was the weaponry of an entire military retinue. And since the weapons themselves had been so thoroughly destroyed, it’s a pretty safe bet that this fate was shared by the men who had wielded them. In establishing their social dominance, the new military kings were playing for high stakes, and Ejsbøl Mose serves as a reminder that for every group that succeeded, another, or several others, failed. Two possible scenarios for this failure immediately suggest themselves. The warrior group unintentionally immortalized there may have been destroyed by another, rival warrior band, or by a group of ordinary, less military Germani, who didn’t appreciate the kind of dominance that the warriors’ leader had in mind. In Hollywood terms, we might be thinking Godfather – the ancient lake having been used by a dominant king to send to any rivals the message that they were likely to end up sleeping with the (in this case freshwater) fishes; or Magnificent Seven – a band of peasants having found enough military effectiveness to rid themselves of at least one predatory warband. There’s no way to be sure, although the fury of the destruction might suggest Yul Brynner rather than Al Pacino, since in some later instances we know of, victorious warband leaders tended to absorb the troops of a defeated rival to increase their own power.12 But this is a detail. The fundamental point is that the rise of the military kings can only have come about through a periodically violent process whereby rivalries both between different warband leaders and between that class of leader and those they sought to dominate, slowly worked themselves out.
Expansion and Development
But this is only part of the story. The kind of military retinue destroyed at Ejsbøl Mose, or employed by Chnodomarius, was a high-maintenance item. Not producing their own foodstuffs, professional warriors required feeding, and all the evidence evocative of Germanic warbands at play – mostly deriving, admittedly, from later heroic poetry, but bolstered by hints in Ammianus and anthropological parallels from better-documented but analogous contexts – suggests that we are talking about feeding literally on a heroic scale: lots of roasted meat and alcohol as brought to the big screen recently in Hollywood’s Beowulf. Military equipment was also not cheap. Admittedly, there is no sign of any body armour in the Ejsbøl Mose finds, and that was the single most expensive item of personal military hardware in the ancient and medieval worlds. Ammianus comments, for instance, that Chnodomarius was easily distinguishable on the battlefield because of his armour, suggesting that even in the fourth century it was not generally being worn by Germanic warriors. Nonetheless, swords were possessed by maybe one-third of the Ejsbøl Mose force. Most of the rest of a warrior’s distinguishing equipment was also made by highly skilled craftsmen from expensive raw materials.13 In other words, the retinues that made the new military kings such a powerful feature of the fourth-century Germanic landscape could not come into being without two preconditions. First, there had to be surpluses of foodstuffs and/or other forms of negotiable wealth being produced by the economy around them, and second, the kings had to be able to turn these surpluses, or a significant portion of them, to their own purposes.
This straightforward observation draws real historical bite from the fact that, up to the birth of Christ, substantial food surpluses and other forms of negotiable wealth were in short supply right across Germanic Europe. The place to start unravelling this story is agricultural production. The economy of Germanic Europe – as indeed that of Roman and every other kind of Europe in the first millennium – was fundamentally agricultural. There are, however, more and less productive types of agricultural economy. Archaeological research undertaken since the Second World War has demonstrated that Germanic Europe went through its own agricultural revolution during the four hundred years when the Roman Empire was its closest western and southern neighbour.
At the start of the period, agricultural practice east of the Rhine was generally ‘extensive’: ‘extensive’, that is, as opposed to ‘intensive’. This meant that a relatively large area was required to support a given population unit, because yields were low. It was entirely characteristic of this kind of farming regime that settlements tended to be small, widely dispersed, and to last for no more than a generation or two on any one site. Essentially, the populations of Germanic Europe did not, or did not have to, maintain the fertility of their fields so as to maximize crop production in any given year, or keep the same field in use over anything but the short to medium term. Once yields began to decline below a level that they found acceptable, they would move on to a new area. The evidence underpinning this interpretation comes in many and varied forms.
In large parts of central-northern Europe, the boundaries of the then widely prevalent ‘Celtic field’ system are still visible in the form of stone walls constructed out of debris cleared from the fields in the course of cultivating them. The fields are extremely large, reflecting the sheer amount of land that was required to keep a single family in business. Known settlement patterns confirm the point. Before 1945, few Germanic settlements belonging to the first two centuries AD had been identified; the early Germani were largely studied, in archaeological terms, from their cemeteries. That situation has now been reversed, with the ratio of settlements to cemeteries standing at 7:1 and growing fast, but the reason for the earlier imbalance has also become evident. All of the settlements now known from these early centuries were small and short-lived. Knowing that any settlement had only a limited life expectancy, in no instance did the inhabitants invest much time or effort in their construction. Therefore, the settlements were both large in overall number and were originally difficult to find. The little direct evidence of prevailing agricultural techniques that happens to survive confirms the point. The well-excavated Germanic-period cemetery at Odry in modern Poland, for instance, was established right on top of an old ‘Celtic field’. From underneath one of the excavated barrows emerged evidence of the ploughing and fertilization regimes employed. Both were rudimentary. Ploughing took the form of narrow, criss-crossed scrapings. This means that the soil was not being turned over, and hence that weeds and crop residues were not rotting back into the soil to restore its vital nutrients, particularly nitrogen. The only form of additional fertilization in evidence was some ash. Employing these kinds of techniques, arable fertility could not long be maintained.14
Conclusive evidence that something changed dramatically in Germanic agricultural practice over the course of the Roman period has emerged since the 1950s, starting in the muddy fields of coastal areas of modern Holland and Germany. By this time, when Ejsbøl Mose was being excavated to such good purpose, archaeological interest was turning generally to settlement, and techniques had advanced to such an extent that really useful results could be obtained. The first major excavations of early Germanic settlements focused on the characteristic manmade mounds of these coastal areas – called terpen in Dutch and Wierde in German – formed by many years of sequential settlement on the same, originally low-lying, site. Over the years rotted refuse, house timbers and other human debris caused the ground level of the settled area to rise. This made these sites an obvious target for archaeological excavation, but local farmers had also long realized that the mounds were piled high with fertile topsoil, so many had been fully or partially grubbed out before the archaeologists got there.
The most detailed work was done at a site that has become celebrated in the field, if little known outside it: Feddersen Wierde. Careful stratigraphic excavation over the best part of a decade, from 1955 to 1963, allowed the full evolution of the settlement to be established. It began in the middle of the first century AD, when five families established themselves there. They comprised a total of maybe fifty people at maximum and practised a mixed agriculture, with much effort put into the rearing of cattle. From the number of animal stalls constructed in the first phase, the five initial families possessed about a hundred cows. But this was only the beginning. The settlement prospered over the next three centuries, reaching its maximum extent in the later third century AD, by which time it numbered as many as three hundred inhabitants who, between them, possessed upwards of four hundred and fifty cows. Many detailed studies have been done of myriad aspects of daily life there, but, for our purposes, the key point is the settlement’s size and longevity. What these indirectly reflect is a revolution in agricultural practice. Under the old extensive agricultural regimes of the early Germanic world, this many people living in such close proximity for over three hundred years would have been inconceivable. Production could never have been that intense, nor fertility maintained for so long. Feddersen Wierde was only possible because its population had adopted a much more intensive agricultural regime, which allowed them to maximize the fertility of their fields to a much greater extent, and permitted a much greater concentration of population to thrive over many generations. The full details of the revolution are beyond reconstruction, but it certainly involved using the manure from all the cattle in a more integrated fashion to maintain the fertility of arable fields.15
It would be rash to generalize from this one example, nor is there any reason to suppose that Feddersen Wierde – based on a greater integration of pastoral and arable agriculture – provides the only possible model of Germanic agricultural intensification. A substantial number of other excavations of Roman-period settlements have made it clear, however, that it was by no means an isolated example of rural development. Nearly as famous as Feddersen Wierde is Wijster, also in north-western Germania. There, originally a single family began to farm in the middle of the first century BC. Grubbing-out by modern farmers meant that large parts of this site were too damaged to excavate properly, but by the fourth century the one family farm had grown into an extensive settlement housing between at least fifty and sixty families, who were busy exploiting the easily worked sandy soils overlooking the mouth of the nearby River Drenthe. Other large settlements of the Roman period excavated in this area beyond the Rhine frontier include Hodde, Vorbasse, Ginderup, Mariesminde and Norre Fjand.
Elsewhere, the picture is not so comprehensive, nor is the precise mode of agricultural intensification so well understood, but enough is known to document the fact that Germanic rural development was a general phenomenon of the Roman period. In what is now central Germany, and the eastern and south-eastern reaches of ancient Germania beyond the Carpathians, the evolving settlement pattern is known in much less detail, and there is no reason, of course, why agricultural practice had to have changed everywhere at the same time. Nonetheless, enough big settlements are known from all these regions – Barhorst, fifty kilometres west of Berlin with thirty families, for instance, or, in the far south-east, the many large settlements of the Goth-dominated Cernjachov system of the fourth century – to show that more intensive agricultural regimes had evolved right across Germanic-dominated Europe in the course of the Roman centuries. Some isolated finds of agricultural equipment indicate the same, iron ploughshares and coulters showing that the soil was being more effectively turned over by the fourth century. The greater size and longevity of settlements, combined with all the evidence for more effective ploughing equipment, document a major transformation of agricultural practice in Germanic Europe in the early centuries AD, even if its techniques remained considerably less specialized than on the other, Roman side of the frontier.16
Two observations follow. First, the massive increase in food production that this revolution in agricultural production must have generated goes a long way towards explaining how the new military kings could support their retinues. Before it unfolded, it must be doubtful that there was enough surplus food in the undeveloped Germanic agricultural economy to support permanent specialist warriors on the fourth-century scale. Second, and this is a much broader point, the vast increase in food production also implies that the population of Germanic Europe increased exponentially during the same period. There is no way to put a figure on the increase, but, as the demographers teach us, one of the key limits on the size of any human population is always the availability of food. The Germanic agricultural revolution, with its vast increase in food supplies, meant that the population must have grown accordingly. Demographic expansion also shows up in other evidence. In Germanic cemeteries occupied throughout the Roman period and excavated with due attention to stratigraphy, larger numbers of people are found interred in those areas in use in the third and fourth centuries compared with the preceding two hundred years. Pollen studies, likewise, provide an alternative view of the same development. Over the first four centuries AD, the proportion of pollen produced by cereal crops increased at the expense of grass and tree pollen, a further indication of agricultural intensification.17
This major increase in agricultural output not only explains how retinues were fed, but must also have been one basic source of the new wealth in Germanic society of this period, visible most obviously in the form of the retinues’ expensive military equipment. Food surpluses could be exchanged for other desirable items. But while perhaps of central importance, agriculture was not the only source of new wealth. Evidence has emerged in recent years to show that, over the first four centuries, the overall economic wealth of Germanic Europe was being increased dramatically by a marked diversification of production and an associated increase in the exchange of a whole series of other goods besides food.
The evidence for both metal production and its subsequent working is highly suggestive of a similar pattern of expansion in that sector of the economy. In particular, two major centres of production in the territory of modern Poland – in the 
Mountains and in southern Mazovia – are between them estimated to have produced upwards of 8,000,000 kilograms of raw iron in the Roman period, with exploitation increasing dramatically in the later centuries. For metal-working, the evidence is more fragmentary, but equally suggestive. When they were first excavated, it was thought that the sixty swords from Ejsbøl Mose represented the greatest find of Roman swords ever discovered in one cache. More detailed analysis has shown, however, that, though based on Roman models, the swords were actually copies forged in Germanic Europe. By c.300 AD, therefore, at least one centre was turning out standardized military equipment on a reasonably large scale, whereas the Germanic swords known from earlier eras were all individual products.18
Evidence for the working of precious metals is equally striking. A hoard of exquisite gold and silver vessels was found at Pietroasa in Romania in the later nineteenth century. Much of it dates to the fifth century, but at least one of the silver dishes was produced in the fourth century and outside the Roman Empire, in Germanic Europe. Moulds for making these kinds of item have been discovered in fourth-century Germanic contexts, and the general level of personal adornments made from precious metals increases over the Roman period. By the fourth century, intricately worked silver fibulae – safety-pins – by which the Germani customarily fastened their clothes, had become reasonably common, and the remains of workshops for producing them have been found at at least one royal seat among the Alamanni. In the first two centuries AD, fibulae had usually been made of bronze or iron. From the mid-third century, Germanic pottery began to change its modes of production. In the third and fourth centuries, Germanic potters for the first time – if not everywhere, and not at the same moment – started to use the wheel to form their wares. This development was combined with much improved kiln technology, allowing the pots to be fired at far higher temperatures, and led to a considerably higher quality of pottery becoming widely available across Germanic Europe. Switching to wheel-made pottery not only generates a higher-quality product but is closely associated with larger-scale, more commercial production. In some areas the transformation was total. In the Goth-dominated Cernjachov world north of the Black Sea in the fourth century, wheel-made tablewares, largely indistinguishable from their provincial Roman counterparts, became the norm (although cooking pots were still made by hand). Among the contemporary Alamanni, by contrast, several local experiments in wheel-made wares never managed to achieve either longevity or widespread distribution – in face, perhaps, of stiffer and nearer Roman competition than their Gothic counterparts. But before the late Roman era, all high-quality wheel-made wares found in Germanic contexts were, without exception, Roman imports, so even this much economic development represents a major transformation.19
Metalworking and pottery production are obviously major areas of the non-agricultural economy, producing both more expensive and cheaper, more widely consumed items. Increasingly professional production methods are visible in other sectors of the later Germanic economy as well, some of them again entirely new. One of the most dramatic is glass production. Before the fourth century, all the glass found in non-Roman Europe was Roman, imported across the frontier. But sometime after 300 AD, a glass production centre opened at Komarov in the hinterland of the Carpathians. Its products came to be distributed widely across central and eastern Europe (Map 3). The various contexts in which the glass has been found indicate that it was an elite item, often used as a mark of status. Though hardly a major employer, its production would certainly have represented a highly valuable addition to someone’s economy. An equally fascinating, though entirely different, example has turned up in an excavated village within the lands dominated by Goths in the fourth century. At Birlad-Valea Seaca in modern Romania, investigators found no less than sixteen huts devoted to the production of one item characteristically found in graves of this period: combs constructed from deer antler. Hairstyles were used by some Germanic groups to express political affiliations, and also to express status. The most famous example is the so-called Suebic knot described by Tacitus and beautifully preserved on one early Germanic skull (Plate 4). In this context, it is hardly surprising that combs were a significant personal possession. Within the huts, parts of combs in every stage of production were discovered, shedding light on the whole process. In this case, it would seem, an entire settlement was devoted to the production of one key item.20
Not only agricultural production, then, but other areas of the economy of Germanic Europe had begun to blossom – in relative terms – by the late Roman period. Right across the region, the early centuries AD witnessed an explosion of development and wealth generation. And like globalization now, at least as important a historical phenomenon as the new wealth itself was the much less comfortable fact that it was not being shared remotely equally. Development in the Germanic world generated clear winners but also clear losers, and it is at this point that military kings, their retainers, and economic development converge still more closely. Many of the items being produced, not just the food, were being consumed by the new military kings and their armed retinues. The iron was necessary for steel weaponry, obviously, but some at least of the glass, precious metal objects and even the higher-quality pottery was aimed in their direction. All of these items have turned up in burials, which careful analysis can show to have belonged to the Germanic social elite of the late Roman period.21 Just how big a social and political revolution had been set in motion?