A re-examination of the campaign and battle of Adrianople, August 378 CE
© Peter Donnelly
Revised February 11, 2013
The Geographical Setting
The Nature of Combat
How Many Fought?
Locating the Battleground
The Next Three Days
The Morning of August 9th
What Determined the Outcome?
The Text of Ammianus
Appendix: The Text of Orosius
The battle fought near Adrianople in 378 CE surely qualifies as one of history’s decisive battles: the one that permanently lodged the Goths within the Roman empire as an independent force. Yet our understanding of it rests almost entirely on the account of one man, the contemporary historian Ammianus Marcellinus. This ex-soldier undoubtedly spoke to eyewitnesses, and his telling of the story is vivid, but he is maddeningly imprecise about the events of the battle. As a result, most modern accounts contain a good deal of conjecture (not always labelled as such), some of it based on a misinterpretation of the text or even on mistranslation. Accordingly my aim in this essay is to put the text back in the forefront, without attempting to bridge its gaps except in the most tentative way.
I have read most of the accounts in English that more than mention the campaign, including the two book-length studies. Of these, Alessandro Barbero’s (see References) is a popular treatment that adds nothing new, though it is a useful summary of the historical context. Simon MacDowall, on the other hand, makes a serious attempt to reconstruct the events leading up to the battle, and the fight itself. Some of the movements on his battle diagram are highly speculative (and based on what I consider an incorrect identification of the battleground), but he may well have judged correctly on many points. He also includes much background information, illustrated with maps, drawings, and photographs.
Hans Delbrück’s view of the campaign has been particularly influential, and indeed it was his foolish depiction of the formation of the wagon laager that first set me to thinking about what must really have happened in Thrace in those hot days of early August. He is one of the few historians to tackle the question of the preliminary movements of the opposing forces; but, as will emerge, I believe he is utterly wrong on some key points.
Roger Blockley, in his commentary on this passage of Ammianus, attempts to unravel the movements and phases of the battle. His remarks are perceptive, even essential, and this remains the only close analysis of the text in English until the great work of J.W. Drijvers et al. should reach its final volume. Blockley is, however, a little hampered by a misunderstanding of the nature of late Roman combat: thinking, for example, that swords were supposed to be drawn as soon as the lines came together.
T.S. Burns (1973) also examines the battle in detail. He too is somewhat hindered by the state of knowledge at the time he wrote his influential article; in particular, his argument depends partly on the tendency of infantry to drift toward their unshielded right side, a phenomenon attested for early classical warfare but surely not applicable to the infantry of the later empire. (Burns’s views of the campaign, if not of the battle, are modified in his 1994 book.)
I begin this study by examining some general issues about the campaign, paying particular attention to the mysterious three days preceding the battle, which are crucial to understanding where and why it took place. I then accompany the reader on a journey through J.C. Rolfe’s translation of Ammianus. An appendix contains the only other near-contemporary narrative to have come down to us, that of Orosius; if nothing else, this brief text serves to illustrate the paucity of our other sources.
The city of Adrianople – also known as Hadrianople, Hadrianopolis, and in modern times Edirne – lies at the center of classical Thrace. To the south and east are the rich lowlands that only the year before have been devastated by the Goths. To the southwest rise the Rhodope Mountains, with the broad Hebros River flowing along their northern flank, bringing with it the great highway from Philippopolis and all the west. At Adrianople, just before it turns south, the Hebros meets another considerable river, the Tonzos, which has its origins far north in the Haemus Mountains. Beyond that range is the Danube, marking the frontier that Fritigern’s people crossed, as legal immigrants to the empire, two years ago.
Between the Haemus and Adrianople, the plains are interrupted by a range of hills that present a sufficient barrier to form the modern boundary between Turkey and Bulgaria. This range broadens as it extends eastward before becoming Mons Asticus (today called the Yildiz), a largely impassable region. At least two paved roads cross the hills from the north: one alongside the Tonzos, leading from the town of Kabyle (modern Yambol) to a junction with the western highway at Adrianople, and another farther to the east connecting Marcianople and the lower Danube with the capital. From the plain around Adrianople, the ground rises gently toward the hills, across a landscape featureless save for shallow ravines and the occasional stream trending toward the Constantinople highway.
This much we can gather from the excellent maps in the Barrington Atlas of the Greek and Roman World. Today, judging from satellite photographs, almost the entire area under examination is settled and cultivated, except for the high ridges. It is reasonable to assume that in the fourth century, all the arable land near rivers, including that along the Tonzos, was cultivated and indeed had been for many centuries. The higher slopes would have been used for pasture if not for crops and orchards. The ridgetop was rougher ground, bare rock in places, no doubt forested in others.
Roman tactics in the late empire were very different from those of the classical age. In particular, the cavalry (including bowmen as well as heavily armored lancers) played a much more prominent role. However, the disaster at Adrianople did not signal the triumph of barbarian cavalry, as some earlier historians thought. By this time, the Romans had already developed the armored cavalry known so well from Byzantine times, and could certainly match anything the Goths put into the field.
Missiles of all kinds were more greatly relied on than in classical times. The heavy infantry marched behind a shield wall bristling with spears — not like the old hoplite or legionary shield wall, but a wall perhaps 6 feet high, the higher part being made up shields held over the shoulders of the men in front by the men in the second rank. Meanwhile the ranks behind them showered the enemy with javelins, arrows, and lead-weighted darts. It must have been very frightening to see, and something cavalry would not readily charge into.
Much work has been done in recent decades on the mechanics or actual experience of battle, and some old ideas have been overturned. For an up-to-date treatment of how men fought in the later empire, see Philip Rance’s fascinating chapter in The Cambridge History of Greek and Roman Warfare.
Ammianus tells us that the Roman scouts reported the number of Gothic warriors as 10,000, and that this proved to be an underestimate. It is futile to guess at the actual figure, but it is worth noting that more than once Ammianus stresses the vast numbers of the enemy.
Adding in noncombatants, we must certainly suppose the encampment to have contained no fewer than 40,000 souls, and perhaps considerably more. Delbrück and MacDowall think that families and goods might have been left behind at Kabyle, but this is inconsistent with the statement that Fritigern has concentrated his people and quickly left that vicinity for the wide-open spaces. Having had so much bad experience of being cooped up and harassed during the previous two years, the Goths intend to move in a body to the open country between Adrianople and Constantinople. Nothing in the text suggests that this is a purely military expedition.
MacDowall estimates the number of wagons in the train at 2,000 to 5,000. But when we consider that this was an entire people on the move, carrying much booty as well as provisions (including fodder) and all the material of daily life, one wagon among a dozen or so people is perhaps not enough. Of course, pack animals may have been used as well. A line of 5,000 wagons would stretch perhaps 30 miles along a road. Cross-country trips, such as the one Ammianus describes, might have been made in column abreast, or simply in vast herds like buffalo, but the trailing wagons must still have been miles behind the leaders.
Most historians speculate, reasonably enough, that Valens reckoned he had the barbarians outnumbered, and therefore must have commanded in the range of 15,000 to 20,000 soldiers — putting the dead and missing at over 10,000. In his 1973 article, Burns posits a far greater number on the field, some 60,000, based on A.H.M. Jones’s interpretation of subsequent recruitment as documented by the Notitia Dignitatum; but in his subsequent book (pp. 30-1) he acknowledges that Jones’s figures are far too high. Wolfram (p. 124) and Lenski (p. 339) say 30,000 to 40,000, but if the Romans had even this many men, why would they have debated whether to attack 10,000 Goths?
In deciding whether the emperor was foolhardy to offer battle when he did, we have to consider that Roman intelligence was in general excellent. Ammianus, a man not at all ignorant of army operations, makes it clear that the Romans were reconnoitering carefully, and we must remember too that they had been operating in the area against this enemy for two years. As to the nature of the blunder made by the Roman scouts, see Open Question.
The location of the battleground is unknown. Until archaeology comes to the rescue, as it has at Teutoburger Wald, the best we can hope for is to establish the right area to look in. To do that, we must examine carefully what has happened in the days before the enemies meet there.
In the earlier part of the summer, Ammianus tells us (31.11.5), the Goths concentrated near Kabyle, some 60 miles from Adrianople. He tells us nothing of what route they took south, but it appears that the main body — comprising the folk with their wagons — must have approached Adrianople by the direct road through the Tonzos valley. A route around the western flank of the hills would not put the Goths in a position to conduct the raid discussed in the next section, besides being dangerously exposed to Gratian’s forces. Delbrück’s lengthy argument (pp. 277-9) in favor of a more easterly route through Beuyuk Dervent may have merit, but it is based on some questionable assumptions, chief of which are that Valens is already west of Adrianople when the Goths leave Kabyle (more on this later), that the Goths have left much of their baggage behind, and that the Tonzos is unfordable. Hodgkin, while admitting that “it is not very easy to understand Ammianus’ account of the movements of the Goths,” thinks that they moved south along the Marcianople road, still farther to the east; but his theory rests on a misapprehension about the location of Nike.
Meanwhile Valens, having left Constantinople in a huff on June 11, has mustered a large mixed force, containing many veterans, at the villa of Melanthias (or Melantias) just outside the capital and is marching toward Adrianople.
At some point during their trek, the Goths detach a strong party of warriors and send them off to attack the Roman supply lines. In response, Valens sends two units, one of cavalry and one of foot archers, to hold key defiles, or “narrows” (angustiae).
Most historians either overlook the raiders or make no attempt to identify their route. MacDowall sees them as an advance party travelling along the same road that the horde will follow over the next three days. But I do not believe that the narrow passes defended by the two units can be in the Tonzos valley itself. For one thing, at no point does the valley extend less than 10 miles between elevations of 1,000 feet. As I suggested above, even in classical times the land on both sides of the river must have been settled and cleared. Wagons might have to keep to the road, but cavalry did not, and a raiding force strong enough to threaten the Constantinople highway was not going to be stopped by two units of Romans spread thinly across the ground.
Ammianus says that the action was taken competently, though he stops short of saying successfully (see the notes on 31.12.2). It would not be competent to send two units of whatever size — and specialist units undoubtedly comprised only a few hundred men — to hold a broad valley against a strong raiding force, especially if Valens was aware that the rest of the horde was coming up behind.
I stand to be corrected on the topography by anyone who has travelled through the valley by the modern road, but I think the last argument stands on its own. If the horde is coming down the Tonzos, a day or two behind the advance party, Valens surely knows it. I stress again the Roman intelligence network: not only do they have scouts abroad, but good information is to be had from deserters hoping to buy their lives and a fresh start. We cannot believe that Valens was informed of the planned Gothic raid, yet knew nothing of the whereabouts of a line of wagons many miles long.
Moreover, a raiding party intending to cut the Roman supply lines would not likely debouch onto the plain in the army’s front. To gain the element of surprise, it must come down behind the Romans – meaning either that Valens has already reached or passed Adrianople (an idea that will be examined, and rejected, below), or that the raiders take a route through the hills somewhere farther east, above the Constantinople-Adrianople road.
Finally, Ammianus states that the passes were nearby (prope). The closest defensible position in the Tonzos valley, presuming there is one, must be one or two days beyond Adrianople, which may itself be a day’s march from the army when the units are dispatched. The passes in the hills immediately above the army’s line of march are more truly described as close by.
The raiders must intend to cut the road at some point far enough from Adrianople to avoid a sudden counterattack, say a day’s march. Probably they have timed the raid to take place just as the main horde is emerging from the gap, at a similar distance from the Roman army, which will have just reached Adrianople. It is very likely that they intend to cut the road in the vicinity of the Nike (Latin Nice) military post, no doubt also a supply depot, which Ammianus identifies later as the goal of the main horde. He has already mentioned this place at 31.11.2, where he has Valens marching to the post in connection with Sebastianus’s campaign earlier that summer. The name, “Victory,” has imperial associations – the goddess and the emperor appear together on many coins – and I think it safe to assume that this is an important military headquarters as well as a way station for travel and transport. It is surely well defended, but a raiding force could at least blockade it and shut down down all traffic on the road, while confiscating trains of provision-laden wagons. Meanwhile Fritigern would be menacing from the north and forcing Valens into negotiations. If successful, it would be a brilliant stroke.
I believe, then, that the angustiae are to be sought somewhere east of the Tonzos. Delbrück, as we have seen, describes a difficult pass in the vicinity of Beuyuk Dervent, and this does appear to be a rugged area, but it is still too far west. I would be inclined to look for an outlet from the hills above Nike itself. There is a modern road over the ridgetop, through the town of Cesmekoy, that follows a stream south right down to Havsa, the putative location of Nike (see Campaign map). This must have been at least a track at the time of our narrative — it is a natural shortcut to Deultum (Dibaltum) on the Black Sea, where a traveller can pick up the Marcianople road or visit the hot springs, and it is an obvious line of communication for the military post. This track passes through some rugged and perhaps forested country, where there are are likely to be defiles that can be blocked by a company of archers, backed up by cavalry to chase down any stragglers who might get through.
Wherever it is, the raid is thwarted. The cavalry make their way back toward the wagons, probably having sent fast riders ahead with the news. We can’t be certain where the horde is, but we know that it is moving too.
Paragraph 31.12.3 begins:
Triduoque proximo, cum barbari gradu incederent leni et metuentes eruptionem per devia, quindecim milibus passuum a civitate descreti, stationem peterent Nicen...
“During the next three days...” Next after what? The last incident Ammianus has described is the report of the planned raid, followed by Valens’s quick reaction. Now Valens, satisfied that his supply lines are secure, resumes his march toward Adrianople, keeping the troops in a defensive formation. Meanwhile, in Rolfe’s translation, “the barbarians, advancing at a slow pace and through unfrequented places, since they feared a sally, were fifteen miles distant from the city and were making for the station of Nice...”
But where are they, and why are they wary of an attack?
To answer the second question first, they go cautiously because they fear an eruptio, defined by the Oxford Latin Dictionary as “a sudden rush (of troops or sim.) from a position, sally, sortie, etc.” They are expecting a sortie directly from whatever city they are keeping away from, not an ambush (insidiae; cf. 31.11.5) from the surrounding countryside.
As for where they are, few historians have confronted this question; they either whisk the Goths to Nike without examining the chronology, or they decline to go into the matter of geography at all. MacDowall thinks that during these three days the Goths are on the Tonzos road. To this I would raise the following objections:
The text here deserves further examination, because it has been translated and interpreted in a variety of ways, most of them more or less misleading if not actually at odds with the Latin. A literal translation is:
In the next three days, while the barbarians were advancing at a slow pace and — fearing a sortie through out-of-the-way places, [themselves being] separated by 15 miles from the civitas — were making for the post at Nike...
In the Latin, “separated” can modify only “the barbarians.”
First of all, let us dispose of Rolfe’s gloss of civitas (a term that can be used for any organized settlement) as Constantinople. This is a guess that makes no sense, putting the horde somewhere on the fringes of the metropolitan area and offering no possible chronology for their getting as far as Nike, more than 100 miles away, let alone to a battleground beyond that. Nonetheless Burns (1973) accepts this interpretation, suggesting that the 10,000 are a splinter party, moving northwest from the capital to join the main horde. But since it appears from 31.11.5 that Fritigern has already finished consolidating his people and has departed from Kabyle, I can’t believe in the existence of such a subgroup in late July or early August. Besides, why would Valens march all the way to Adrianople if there were a strong enemy force lingering somewhere in the vicinity of the capital?
Delbrück, as rendered into English by Renfroe, has it thus:
In the next three days, while the barbarians, fearing an attack from the difficult terrain, advanced slowly in the direction of Nike, a way station fifteen miles from the city of Adrianople...
Now, the Barrington Atlas does happen to place Nike about 15 miles from Adrianople, but I hope not on the basis of this translation, which is unsupported by the text.
Delbrück also has the barbarians fearing an attack from difficult terrain, instead of fearing a sortie through out-of-the-way places. He has placed the horde on the main road, fearing an ambush from out of the hills! And if the Goths are along the Constantinople road, of course they must run into Valens, unless Valens has already passed by. Therefore, goes Delbrück’s argument, Valens must actually be somewhere on the road to Philippopolis; finding his communications threatened, he turns around and marches back to Adrianople. This scenario is also espoused by MacDowall and Barbero. Of course, it only works if we think Ammianus knew nothing about it, and if we imagine that Valens would have marched blithely past the junction of a direct road from the Goths’ base at Kabyle, trusting a couple of detached units to keep all safe while he made a rendezvous with Gratian somewhere to the west.
In reality, the problem Delbrück tries to solve with this contrivance does not exist. He takes it that the Goths have already arrived at Nike when Valens learns of their whereabouts and completes his march. But the Latin is explicit that the scouting report comes while the barbarians are moving slowly toward Nike; they have certainly not reached the place, and Ammianus doesn’t say they ever do. They may still be near the mouth of the Tonzos gap; indeed, it has probably taken several days for the long line of wagons to congregate in a defensible mass after leaving the valley, and their mustering ground would be a good place for the Roman scouts to compile first estimates of their numbers. In any case, since we are informed that the barbarians kept away from the road, there is no danger of their encountering Valens as he quickly advances to Adrianople from the southeast.
Moreover, if Delbrück’s view is taken, the three days have no particular significance: they are simply the interval between the episode of the raid and the horde’s arrival at Nike, after which some further days must be allowed for all the activity resulting from the scouting report – if we are to give Ammianus any credence at all. Surely we are meant to understand that the three days are those immediately preceding the battle: days full of urgency, with Valens rushing to Adrianople and there anxiously awaiting word from Gratian while the Goths edge to the southeast.
C.D. Yonge’s translation reads:
Three days afterwards, when the barbarians, who were advancing slowly, because they feared an attack in the unfavorable ground which they were traversing, arrived within fifteen miles from the station of Nice, which was the aim of their march...
There is no reason to equate the civitas with Nike, which is clearly an official post (statio), not a city, and in any case Yonge has defied the Latin by turning the journey into an arrival. He further suggests that the ground would be “unfavorable” to the Goths in an attack, which is not what devia means at all. In fact, every translator but Rolfe has made the mistake of turning these out-of-the-way places into difficult ground, perhaps persuaded that Ammianus is talking about the same broken terrain (viarum spatiis confragosis) over which the Romans hasten on their way to attack the camp. I believe rather that the sense of metuentes eruptionem per devia is “wary of a sortie across country.”
Walter Hamilton translates more accurately:
During the next three days the barbarians advanced slowly over difficult country expecting to be attacked, keeping fifteen miles from the city and making for the post at Nike.
I think this is almost right, except for the usual proviso about devia, and the suspicion that civitas may well not refer to a town at all (Adrianople is not mentioned by name until the next paragraph) but rather to the densely settled zone in general, much as we sometimes use the word “civilization.” Hamilton’s translation also fails to make it clear that what follows in the text – beginning with the erroneous scouting report – takes place while the barbarians are advancing, rather than at the end of the three days.
Finally, Barbero does not even attempt to get himself out of the geographical and logical muddle he creates with his paraphrase: “The barbarians, for their part, proceeded with caution, not wishing to expose themselves to a surprise attack in the mountains; they did not try to force the passes, choosing instead a more circuitous route.”
I have spent so much time on the various readings of this passage because the text is crucial to understanding where the Goths were during this time. Yet it seems that most of the interpreters have been rather free with what Ammianus says. If we take him at his word, the known facts are that during three days these things held true:
I believe this passage can only mean that the Goths are moving away from the Tonzos River, through the countryside roughly 15 miles northeast of the town, heading east to southeast (see Campaign map). They are progressing slowly, perhaps only five miles a day, because they must form the laager early in the day to avoid being caught in the open by a dawn march from the Roman camp.
At the end of the three days, the Goths stop moving at an advantageous position near a water source and send an embassy to Valens. He, meanwhile, has been very busy: spurred to action by the false report of the weakness of the horde, he has completed his march, built a fortified camp, made the necessary preparations for battle, established his court, waited impatiently for Gratian, and called his consistory together to receive Richomer and debate Gratian’s letter. The timeline here is worth considering.
Valens must receive the misleading report on the first of the three days (August 6th), as the horde is emerging from the valley, and by this time the Romans must be close to Adrianople, at Nike or already on the last leg of the march. The report reaches Valens early enough in the day to allow him to reach the city on that same evening. On the second and third days, as the Roman camp is being fortified and preparations are being made for battle, the scouts report the slow progress of the wagons toward the east (and must still be deceived about the barbarian numbers). On the evening of the third day, Valens receives the Gothic embassy, and on the following day he marches out.
The next uncertainty is where the Romans are camped, and how far they have to march to the wagon laager.
Ammianus tells us that Valens established a strong camp “near a suburb of Adrianople,” which in his overelaborated style apparently means somewhere in the suburbs of the town. One would expect to find it at the closest open ground that could be supplied with fresh water.
On the morning of the battle, the Romans leave their baggage “near the walls” under guard. On the face of it, this suggests that the camp has been abandoned; indeed, we never hear of it again, except perhaps obliquely at 31.15.4. The best sense I can make of it is that the main camp is not deemed defensible by the “suitable guard of legions” that is left behind, and the Romans have fortified a separate area under covering fire of the city’s parapets, perhaps consisting of a palisade around outbuildings that can be used for storage of the tents and other gear.
The continuation, in Rolfe’s edition, reads: Decursis itaque viarum spatiis confragosis, cum in medium torridus procederet dies, octava tandem hora hostium carpenta cernuntur. “So after hastening a long distance over rough ground, while the hot day was advancing toward noon, finally at the eighth hour they saw the wagons of the enemy.” This seems straightforward, with the caveats mentioned in the commentary below.
The biggest problem here is the doubtful state of the text. The manuscript reads: “finally at the eighth they saw,” with octavo being in the masculine or neuter. One editor (followed by Rolfe) has supplied hora and changed the gender of the ordinal to match it. Others take octavo to mean “at the eighth milestone” (lapide or miliario understood), pointing to a similar abbreviated usage in Tacitus. In consequence, many accounts of the battle have the army marching eight miles rather than till the eighth hour. But if Ammianus is talking about eight miles, is that the distance the army has marched when it spies the Gothic camp, perhaps still several miles off? This is Blockley’s reading, though it seems improbable that Ammianus would record a precise distance for this spot and not for the battlefield itself.
A later chronicle puts the distance of the battlefield from the town at 12 miles. This must be the authority for Wolfram’s confident assertion that Valens had to march “almost 11 miles,” allowing for different measurements of a mile. (The Roman mile is about 5,000 feet.)
In estimating the distance, we have to establish two things. Are the Romans in a hurry? And if so, how far can they have advanced by early afternoon?
Ammianus tells us that the Romans left abruptly, but contrary to Zosimus and the dramatic accounts of some modern writers, there is nothing in the text to suggest that they were unprepared or disorderly. In fact, we know that they had spent one or two days preparing for just such a march. One historian has gone so far as to say that Valens sent his troops into the field without food or water. This is absurd, of course. Yet the fact is inescapable that the army did outmarch its supplies. For some reason, Valens thought haste more important than the fact that, in this searing heat, men and animals quickly used up what water they carried.
Valens’s jealousy of Gratian might explain why the army set off on August 9th, but not why it was in such a hurry to reach the battlefield, which lay in the opposite direction from Gratian’s supposedly unwelcome reinforcements. We can also rule out any idea that Valens did not know how far a march lay ahead of him, as has been suggested by at least one scholar. He would surely have known the laager’s exact location: even if the Roman scouts had somehow failed to locate the encampment of a huge force whose movements they had been watching for three days, the Gothic ambassadors would have given the emperor this information along with Fritigern’s private letter (31.12.9). We must conclude, then, that when faced with a choice between getting to the camp as quickly as possible, and arriving later in the day with an army that had been fed, watered, and rested, Valens chose the first course. The explanation may be that he knew about the approach of Alatheus and Saphrax, and believed that only a forced march could win him the battle before they returned.
Without knowing the numbers of men who marched that day, how many columns they marched in, and what kind of obstacles they faced, it is impossible to make more than a rough estimate of the ground they covered. Supposing an average speed of two miles per hour across country, they might have marched 10 or 12 miles between dawn and noon; but of course that is only the van, and it would be several hours before the rest came up and were formed into line. Eight miles, however, does not seem far enough: a march of that distance could be completed in a morning without the sort of haste that left the troops without food and water. Furthermore, we have just been told that the barbarians were keeping 15 miles away from the civitas. The 12 miles cited by the chronicler is more credible, I think, than the “eighth milestone.”
At least two sites for the battlefield have attracted interest. MacDowall puts it just east of the Tonzos River, about 10 miles north of Adrianople, near the modern village of Muratçali. He cites another scholar, F. Runkel, who locates it a similar distance to the east of the city, near Demirhanli. Noel Lenski follows Runkel, while D.S. Potter insists that MacDowall has said the last word.
The barbarians have travelled toward Nike for three days. Muratçali is less than three miles from the Tonzos, which would mean very slow travelling indeed. Also, a camp in that spot would not present an immediate threat to the Roman supply line. Demirhanli is the more likely candidate, lying 10 miles closer to Nike and indeed only an easy day’s march from it. It is also near several streams that might have sustained the horde’s stock indefinitely.
We may wonder about the emperor’s willingness to negotiate with Fritigern, even to the point of risking the life of a great nobleman at the eleventh hour. After all, a treaty has been broken, Thrace has been ravaged, and Valens is at last in a position to put a finish to the whole matter. What is there to talk about?
This gets us into matters of grand strategy that are outside the scope of this article, but a few points should be made.
First, the Goths are sincere throughout the negotiations. They do not want to fight, with or without a tactical advantage. A victory gains them nothing. Ultimately, the only thing that can save them as a people is land. They are offering to settle on part of the land they have just pillaged, because it is now substantially empty. Never mind that it lies so close to Constantinople; this actually ensures that the Goths will remain peaceable as they spread out to till the soil and tend their herds. Yet their warriors will be there, on call, to defend against the increasing pressure from across the Danube.
What if Valens chooses to fight, and wins? He can massacre a substantial part of Fritigern’s people, eliminate them as a threat, and gain much in the way of treasure and slaves. But he loses many armed men, both through his own casualties and through the loss of warriors who could be turned to defending Thrace. The Danube frontier has become porous, more and more bands of adventurers are appearing on this side of the river, and the whole region is very unstable. One victory is not going to make much difference, and the army is badly needed in the east.
Valens refuses the initial proposal brought to the Roman camp: the Goths cannot have Thrace. In view of his later willingness to negotiate further on the battlefield, one expects him to wait for Fritigern’s response. Why then does he march out so abruptly? Besides believing that the numbers are in his favor, he must fear that Fritigern will move immediately, sending cavalry to take Nike while the main horde menaces any relief force from Adrianople. (Perhaps this is even the explanation for the whereabouts of Alatheus and Saphrax: they have been sent against Nike, but are recalled as soon as the Roman advance is spotted.) With Nike in his hands, the chieftain will be in an even stronger bargaining position.
As for the negotiations on the battlefield, Valens is perhaps spurred by two further considerations: the fact that the Gothic cavalry have been sighted and are close enough to join the battle, and his wish to see his army fed and watered before fighting commences.
Without further evidence, it is simply impossible to present any coherent account of the battle that is not full of guesswork. We know that the Roman cavalry on the left were slow to deploy, though we can suspect that the delays caused by Fritigern would have given them time to get in position, and indeed Ammianus hints (31.12.12) that their deployment went smoothly. We know that part of the Roman left reached the wagons, but what kind of resistance they encountered, other than missile fire, is not clear. Certainly at least some barbarian warriors had stationed themselves outside the camp, because Ammianus speaks of lines coming together. But as the Romans on the left pressed forward they found themselves deserted by “the rest of the cavalry” or “the remaining cavalry” (reliquo equitatu), who had perhaps been charged by Alatheus and Saphrax. The infantry were thus outflanked and pressed back one upon the other, while the Goths “poured forth in immense columns,” implying that strong reinforcements came from within the wagon circle.
A general massacre ensued, and only darkness enabled some few thousand Romans to escape toward Adrianople, while the rest of the survivors dispersed. The townspeople and the garrison refused to open the gates, neither trusting nor wanting to feed the indiscriminate mass of men and horses outside their walls. The Goths were close behind, entertaining notions of capturing the imperial treasury stored in the town. Another bitter fight ensued, broken up only by a thunderstorm and the exhaustion of the attackers. After several efforts in the coming days to gain entrance to the town, they moved off toward Constantinople.
It is difficult to judge exactly how or why the Romans lost so badly at Adrianople. They surely had superior armor and more abundant weaponry — the Goths had been disarmed at the Danube and cannot have completely re-equipped themselves in two years. Valens commanded a disciplined body of soldiers, many of them veterans of combat, against warrior migrants who had no time to drill and were more used to skirmishing than to pitched battle. (On the other hand, they had had successes and morale would have been high.) The Romans had knowledge of the ground, mostly good intelligence, and reliable staff work — this was an urgent action, but not a mad rush into battle.
So what counted against them?
The following is the account of the battle as translated by J.C. Rolfe in the Loeb edition (still in print, but the text is now in the public domain), with my commentary. I have omitted Rolfe’s notes, together with two or three historical parallels that shed no light on the battle.
In general, Rolfe is a more reliable guide to the Latin than Walter Hamilton in his abridged Penguin version. However, Hamilton is faithful to the sense, and I would not be without either translation. The nineteenth-century version by C.D. Yonge is of much less value.
Any interpretation of ancient writings must be offered with the caveat that our knowledge is only as good as the text as it has come down to us, but in Ammianus’s case the warning is particularly apt. Our source for most of the text is a single manuscript, and it is full of lacunae and accumulated scribal errors; indeed, it seems to have been copied by monks with no great knowledge of Latin. Scholarship has helped set the text to rights, but in places it is still guesswork, and of course we have no way of knowing if whole passages have disappeared without a trace.
12.1 In those same days Valens was troubled for two reasons: first, by the news that the Lentienses had been defeated; secondly, because Sebastianus wrote from time to time exaggerating his exploits. He therefore marched forth from Melanthias, being eager to do some glorious deed to equal his young nephew [Gratian], whose valiant exploits consumed him with envy. He had under his command a force made up of varying elements, but one neither contemptible, nor unwarlike; for he had joined with them also a large number of veterans, among whom were other officers of high rank and Trajanus, shortly before a commander-in-chief [magister armorum], whom he had recalled to active service.
Sebastianus has been managing the campaign against the Goths while Valens prepares his army. He has shortly before replaced Trajanus as magister peditum, a sort of joint chief of staff with the magister equitum. Ammianus groups these high officers as magistri armorum.
12.2 And since it was learned from careful reconnoitering that the enemy were planning with strong guards to block the roads over which the necessary supplies were being brought, he tried competently to frustrate this attempt by quickly sending an infantry troop of bowmen and a squadron [turma] of cavalry, in order to secure the advantages of the narrow passes, which were near by.
Note that the Romans are not operating “in the dark,” as Barbero avers (p. 93).
Rolfe’s translation here is a bit misleading; for conatui competenter I would prefer “he competently undertook” to “he tried competently.” Hamilton goes further in his rendition: “This threat was effectively met.”
MacDowall says a turma is made up of only 30 men. That had been true 500 years before, but by this time the word has acquired a more general meaning, much like our “troop.” At 18.8.2, Ammianus has 700 men belong to two turmae. Orosius uses the term, in the passage quoted below, to describe the units of cavalry that abandoned the flank during the battle.
See the main text for further discussion of this passage and the following one.
12.3 During the next three days, when the barbarians, advancing at a slow pace and through unfrequented places, since they feared a sally, were fifteen miles distant from the city [civitas] and were making for the station of Nice, through some mistake or other the emperor was assured by his skirmishers that all that part of the enemy’s horde which they had seen consisted of only ten thousand men, and carried away by a kind of rash ardour, he determined to attack them at once.
The scouts affirm that this part of the horde (multitudo) numbers ten thousand; Rolfe correctly supplies “men.” If the figure is taken to include noncombatants, there could be no risk in attacking at once, hence no debate as in 31.12.6.
12.4 Accordingly, advancing in square formation, he came to the vicinity of a suburb of Hadrianopolis, where he made a strong rampart of stakes, surrounded by a moat, and impatiently waited for Gratian; there he received Richomeres, general of the household troops, sent in advance by Gratian with a letter, in which he said that he himself also would soon be there.
Historically the agmen quadratum was a hollow square of some type, meant to protect the baggage, but we need not believe that Valens “formed his army into a massive square,” in which case most of them would have been marching off the road. In the fourth century the term was applied conventionally to an army marching toward the enemy in defensive order, e.g. Ammianus 24.1.2, 25.3.1, 27.2.8, 27.10.6, 29.5.39, 31.16.4 (the horde moving on Constantinople); Historia Augusta “Maximus et Balbinus” 2.4, “Maximini Duo” 21.1. See also Hugh Elton, Warfare in Roman Europe AD 350-425 (Oxford, 1996) p. 244.
Vegetius, writing probably around 400 CE, says that the practice of fortifying camps with ditches and a stockade had not been followed for a long time (1.21). As usual with that author, it is difficult to know what weight to put on the statement, but it at least suggests the possibility that Ammianus is being purely conventional in his account, and that the camp was not fortified, which would explain why it does not figure in the next day's siege.
Delbrück, Barbero, and MacDowall are in agreement that Valens must already have passed Adrianople on his way to meet Gratian somewhere to the west, and that he had to turn back to meet the advance of the horde. But this reading supposes that Ammianus conflated two different visits to the town and totally overlooked Valens’s supposed intention to continue westward. See the further discussion in the main text.
12.5 Since the contents besought him to wait a while for the partner in his dangers, and not rashly to expose himself alone to serious perils, Valens called a council of various of his higher officers and considered what ought to be done.
12.6 And while some, influenced by Sebastianus, urged him to give battle at once, the man called Victor, a commander of cavalry [magister equitum], a Sarmatian by birth, but foresighted and careful, with the support of many others recommended that his imperial colleague be awaited, so that, strengthened by the addition of the Gallic army, he might the more easily crush the fiery over-confidence of the barbarians
Victor survived the battle, Sebastianus did not, so take this account for what it is worth. Burns (1973), following Hodgkin (p. 145), takes Zosimus as contradicting Ammianus when he speaks of Sebastianus as the cautious one, but Zosimus seems to be referring to an earlier occasion when Sebastianus argued for the continuation of his containment strategy rather than the mounting of a full expedition.
12.7 However, the fatal insistence of the emperor prevailed, supported by the flattering opinion of some of his courtiers, who urged him to make all haste in order that Gratian might not have a share in the victory which (as they represented) was already all but won.
This is not just a war council but a meeting of the consistory, the emperor’s supreme court. The fiery debate, the bluff advice of the generals contrasted with the cajolery of the sycophants working on the emperor’s blind flaw — this is the stuff of drama, and Ammianus loves it. But Valens’s fatal insistence may have been soundly based, even though supported by flattery. I am persuaded that he had good strategic reasons for setting out immediately.
12.8 While the necessary preparations for the decisive battle were going on, a Christian presbyter (to use their own term), who had been sent by Fritigern as an envoy, in company with some humble folk came to the emperor's camp. He was courteously received and presented a letter from the same chieftain, openly requesting that to him and his people, whom the rapid forays of savage races had made exiles from their native lands, Thrace only should be granted as a habitation, with all its flocks and crops; and they promised lasting peace if this request were granted.
Ammianus’s apology for using a Christian term is typical of secular writing at that time.
The Goths are not demanding the entire region, only the province of Thrace, which lies northwest of Adrianople and includes Philippopolis. On the question of what status they hope to achieve within the empire, see Heather, Goths and Romans pp. 175 ff.
12.9 Besides this the aforesaid Christian, apparently a confidant and trusted friend of Fritigern, presented also a private letter of the same king, who, all too skilled in craft and in various forms of deception, informed Valens, pretending that he hoped soon to be his friend and ally, that he could not tame the savagery of his people, or entice them to adopt conditions favourable to the Roman state, unless the emperor should from time to time show them near at hand his army ready for battle, and through the fear aroused by the imperial name check their destructive eagerness for war. But as to the envoys, their sincerity was doubted, and they left without accomplishing their purpose.
It’s difficult to evaluate Fritigern’s private letter. The idea seems to be that he sends a shabby delegation, led by a priest to give them some superficial credibility, to present his public offer in a way that can easily be dismissed, meanwhile privately luring Valens into battle. But as I stated above, Fritigern and his people have little to gain by fighting, no matter how glorious their victory.
Wolfram (pp. 125-6) chides Ammianus for contradicting himself by calling the embassy “unauthorized” even though the chief envoy is a confidant of the king. But I cannot get this sense from legati ut ambigui habiti, which I take to mean simply that the ambassadors were regarded as in some way doubtful. (As Hamilton translates, “they did not inspire confidence.”) They are, after all, mere humiles, not men of rank and therefore not men of honor. But their embassy cannot be considered unauthorized when they bear two letters from Fritigern himself. Their credentials are accepted, they are received courteously, their proposals are read, and they are dismissed because they are not suitable delegates for further discussions.
12.10 But on the dawn of that day which is numbered in the calendar as the fifth before the Ides of August [the 9th] the army began its march with extreme haste [praepropere], leaving all its baggage and packs near the walls of Hadrianopolis with a suitable guard of legions; for the treasury, and the insignia of imperial dignity besides, with the prefect and the emperor's council, were kept within the circuit of the walls.
The departure of the army is abrupt, for reasons already discussed, but they are not necessarily in disorder, as Zosimus states.
12.11 So after hastening a long distance over rough ground, while the hot day was advancing towards noon, finally at the eighth hour they saw the wagons of the enemy, which, as the report of the scouts had declared, were arranged in the form of a perfect circle. And while the barbarian soldiers, according to their custom, uttered savage and dismal howls, the Roman leaders so drew up their line of battle that the cavalry on the right wing were first pushed forward, while the greater part of the infantry waited in reserve.
Rolfe’s first comma is confusing. The eighth hour is an hour after noon. Ammianus is not contradicting himself but saying that the Romans marched all morning in the hot sun, before spying the wagons in the early afternoon. But see the discussion above about the reading of the text.
The translation may also be misleading about the nature of the march. Decursis viarum spatiis confragosis does not necessarily imply reckless haste: the verb decurro can mean “maneuver” or “drill” (much as we would “run through” an exercise), so may in fact suggest an orderly march, despite the difficulties encountered by some units in keeping their position. Nor does the Latin more than hint that the distances are long. Viarum spatiis confragosis means something like “over rugged stretches.” (Claudian In Rufinum 2.137 uses spatium viarum in the sense of “distance to be covered” in a context where the distance is perceived as short.) I think this just means that the Romans are marching cross-country, perhaps taking advantage of the irregular country lanes, perhaps having to climb or demolish walls and trample crops as they go. The ground rises slightly as they move toward the hills, but in cooler weather this would not be a terribly difficult march for men who are used to it, especially since they are marching without the equipment that has been left in camp.
The wagon circle is round, “as if turned” (by a lathe or wheel); but the text is corrupt here. See also the note on 31.15.5.
Delbrück seems to think there must have been a hollow ring of wagons, old-west style. Since it would require many miles of open ground to construct this, he argues, the Gothic numbers must have been small. The smallness of the barbarian hordes is a hobby-horse of Delbrück’s, and in this case his reasoning is absurd.
A more reasonable reconstruction is this. As the wagons arrive at a previously scouted campsite, the scouts direct the leaders to form a defensible, curved barrier in the direction of the enemy, perhaps two lines deep. Then the following wagons gather behind this, in an orderly fashion that leaves lanes for traffic and a space for the warriors to muster behind the barricade. It is difficult to imagine that the camp was enclosed on all sides: the practical difficulties of making an enclosure to accommodate more than 40,000 people are enormous, especially as there would inevitably be latecomers, the taking of animals to pasture and water, etc.
I would not venture a guess as to the size of the barricade, but it may be useful to suggest a general range. If you allow 500 sq. ft. per wagon, including all common spaces, a camp large enough to accommodate 5,000 wagons would have a diameter (presuming circularity) of about 1,800 ft. or 540 m, and a circumference of 5,600 ft. or 1,700 m.
The “savage and dismal howls” of the barbarians may be imagined as war chants, with a strong element of the old Germanic religion despite the Goths’ recent conversion to Arian Christianity. Hodgkin says they were “probably meant for melody.” See also 31.7.11, where the war cries of the Romans and barbarians are contrasted, and the Goths are said to have “sounded the glories of their forefathers.”
The end of the section suggests that the right-wing cavalry are the first to reach the field, and they screen the deployment of the infantry. The infantry are meant to be in a supporting role (subsidebat, which implies remaining in place, but not necessarily to the rear; at 31.13.9 the Batavi are described as being in subsidiis, which is more clearly a reserve position). We hear nothing more of them until they are outflanked by the Gothic cavalry; perhaps their intended role is to attack the camp once the flanks are secure. In the meantime they may be close enough to the camp to be contributing to the terror described in the following paragraphs, by shooting into it.
12.12 But the left wing of the horsemen (which was formed with the greatest difficulty, since very many of them were still scattered along the roads) was hastening to the spot at swift pace. And while that same wing was being extended, still without interruption, the barbarians were terrified by the awful din, the hiss of whirring arrows and the menacing clash of shields; and since a part of their forces under Alatheus and Saphrax was far away and, though sent for, had not yet returned, they sent envoys to beg for peace
See the main text for discussion of the negotiations. Note that arrows are already flying. Although their columns have become separated on the winding country roads, the left wing is continuing to deploy nullo interturbante, with nothing to interrupt them. At this point the Romans have the upper hand.
Alatheus and Saphrax are leaders of the Greuthungi, the eastern Goths who fled from the Huns after their conquest of the Alans (31.3.1-3). Modern scholarship no longer identifies the Greuthungi and Tervingi with the fifth-century groups known as the Ostrogoths and Visigoths; see for example Peter Heather, The Goths, p. 52.
Most summaries of the battle state that the missing cavalry was off foraging, but that is an assumption. Fritigern was certainly expecting a battle or at least a Roman demonstration on this day, and would not have allowed the best part of his cavalry to range too far afield, but since they did return on time, it is not impossible that they were grazing within sight of a smoke signal or even within earshot of the “dismal howls” and trumpet calls from the camp. Although Ammianus does speak of their “return,” I think it possible that they are actually late arrivals from the north, and it is conceivable that they are the “strong guards” earlier sent on the raid; in either case they would almost certainly be arriving from the northwest, on the Goths’ right wing. See Open Question.
12.13 The emperor scorned these because of their low origin, demanding for the execution of a lasting treaty that suitable chieftains be sent; meanwhile the enemy purposely delayed, in order that during the pretended truce their cavalry might return, who, they hoped, would soon make their appearance; also that our soldiers might be exposed to the fiery summer heat and exhausted by their dry throats, while the broad plains gleamed with fires, which the enemy were feeding with wood and dry fuel, for this same purpose. To that evil was added another deadly one, namely, that men and beasts were tormented by severe hunger.
Once again Fritigern has sent an embassy that is likely to be spurned. Is he simply delaying? No doubt he still wants peace, but reckons he’d better have the cavalry on hand anyway.
The smoke may be intended to summon Alatheus and Saphrax, or possibly to make their cloud of dust less apparent. But fuel must be limited, and it is unlikely that there is grass left to burn anywhere near the hungry Gothic stock, so the fires are possibly only large enough to cause discomfort as their smoke drifts down the slope toward the Roman line.
The heat further torments those in full battle armor (31.13.7), which they are perhaps not accustomed to wearing.
12.14 Meanwhile Fritigern, shrewd to foresee the future and fearing the uncertainty of war, on his own initiative sent one of his common soldiers as a herald, requesting that picked men of noble rank be sent to him at once as hostages and saying that he himself would fearlessly meet the threats of his soldiers and do what was necessary.
Again Fritigern portrays himself as the peacemaker acting in defiance of his bellicose warriors. However, the text is uncertain here. Surely he is not in a position to demand hostages of the Romans without reciprocating, but it is not clear whether he intends to send “suitable chieftains” to the Roman lines, or come himself.
Rolfe’s “common soldier” is literally “one of the plebs.” For a third time Fritigern sends a lowly messenger to Valens.
12.15 The proposal of the dreaded leader was welcome and approved, and the tribune Aequitius, then marshal of the court and a relative of Valens, with the general consent was chosen to go speedily as a surety. When he objected, on the ground that he had once been captured by the enemy but had escaped from Dibaltum, and therefore feared their unreasonable anger, Richomeres voluntarily offered his own services and gladly promised to go, thinking this also to be a fine act and worthy of a brave man. And soon he was on his way [bringing] proofs of his rank and birth. . . .
There is a short lacuna here.
Aequitius is otherwise unknown, except for his death in the battle (31.13.18). Although he has a court post, he is a mere tribune, not so exalted in rank as Richomeres. Evidently his connection with Valens makes him a worthy hostage. Perhaps he is a son of the Flavius Aequitius or Equitius who had a long career fighting Germans, served as consul alongside Gratian in 374, and after Valentinian’s death was involved in the elevation of Valentinian II. As a Pannonian, he might have had kinship with Valentinian and Valens.
12.16 As he was on his way to the enemy’s rampart, the archers and the targeteers, then under the command of one Bacurius of Hiberia and Cassio, had rushed forward too eagerly in hot attack, and were already engaged with their adversaries; and as their charge had been untimely, so their retreat [discessus] was cowardly; and thus they gave an unfavourable omen to the beginning of the battle.
Note it is the Romans who start the fight. The Goths have held back even after the initial shower of arrows.
Bacurius was of Iberian (Georgian) royal blood, and no fool; he eventually rose to high rank in the Roman service, and distinguished himself at the battle of the Frigidus (394 CE) before meeting his death there. Cassio is otherwise unknown.
The offending units are evidently cavalry. If they are to be identified with the deserting cavalry of 31.13.2 (and discessus implies an outright abandonment of the field, not simply a withdrawal to their lines), they are probably part of the left wing. Some writers assume that they are part of the advanced right wing, but the Roman line is probably fully deployed at this time, so they could be anywhere. And although the Roman attack precipitates the charge of the Gothic and Alanic cavalry, these events are not necessarily associated with the eventual collapse of the Roman left wing. See Blockley’s remarks on the following paragraph.
This is one of only two appearances of the Gothic entrenchments after they are first sighted. We can infer from this passage that Fritigern is within the camp, and at least part of the curved rampart lies at the front of the Gothic lines. At some point a line of men must extend in both directions from the curve to prevent encirclement; and since it would seem that the greater part of the cavalry is absent, the existence of this line places some Gothic infantry outside the barricade. Also, the hasty retreat of the Romans suggests that they have a mobile opponent and are not just shooting at the camp.
12.17 This unseasonable proceeding not only thwarted the prompt action of Richomeres, who was not allowed to go at all, but also the Gothic cavalry, returning [reversus] with Alatheus and Saphrax, combined with a band of the Halani, dashed out as a thunderbolt does near high mountains, and threw into confusion all those whom they could find in the way of their swift onslaught, and quickly slew them.
Contrary to popular accounts, the cavalry cannot make a surprise appearance out of nowhere. On a hot August day, their cloud of dust has long been visible.
Much was made of these barbarian “shock tactics” by writers like Charles Oman and J.F.C. Fuller. In fact they were nothing new. Philip Sidnell in his excellent book Warhorse shows that cavalry had been used in just this way for centuries.
MacDowall (pp. 73-77, with map) places the onslaught on the Roman left, though he rather confusingly has placed the “bulk of the Roman right-wing cavalry” on that side as well, so that he can have both forces driven off the field by Alatheus and Saphrax. Wolfram, on the other hand, states (p. 127) that “the Greutungian and Alanic horsemen fell on the right flank of the Romans” and then “one detachment of the Gothic cavalry withdrew, went round the Romans and attacked the left flank.” This is an equally speculative attempt to account for the fact that the Roman infantry were ultimately pressed in on both flanks. We could with equal justification argue that Alatheus and Saphrax divided their forces as they approached the camp, to take up positions on both wings.
Burns (1973 and 1994) argues that the number of Gothic cavalry has been greatly exaggerated, pointing out the difficulties of keeping horses alive when even the people are starving. However, the same argument could be used about their draft animals — oxen, most likely — which had to at least outnumber the wagons.
13.1 On every side armour and weapons clashed, and Bellona, raging with more than usual madness for the destruction of the Romans, blew her lamentable war-trumpets; our soldiers who were giving way rallied, exchanging many encouraging shouts, but the battle, spreading like flames, filled their hearts with terror, as numbers of them were pierced by strokes of whirling spears and arrows.
This is a generalized picture, and “our soldiers who were giving way” are likely not the archers and targeteers of 31.12.16.
13.2 Then the lines dashed together like beaked ships, pushing each other back and forth in turn, and tossed about by alternate movements, like waves at sea. And because the left wing, which had made its way as far as the very wagons, and would have gone farther if it had had any support, being deserted by the rest of the cavalry, was hard pressed by the enemy’s numbers, it was crushed, and overwhelmed, as if by the downfall of a mighty rampart. The foot-soldiers thus stood unprotected, and their companies were so crowded together that hardly anyone could pull out his sword or draw back his arm. Because of clouds of dust the heavens could no longer be seen, and echoed with frightful cries. Hence the arrows whirling death from every side always found their mark with fatal effect, since they could not be seen beforehand nor guarded against.
This passage is critical to understanding the battle, but enigmatic. It is generally taken that some part of the left-wing cavalry has reached the wagons unsupported by the other part (reliquo equitatu desertum). The remnants of the wing (cornu) are then overwhelmed, leaving the infantry in the center exposed. If this is the correct interpretation, probably the deserters are the forces under Bacurius and Cassio, who have fled the field after being repulsed (31.12.16).
Sidnell takes reliquo equitatu as referring to the right wing; that is, the cavalry on that wing takes the brunt of the charge by Alatheus and Saphrax and retreats first. However, one would expect the two cornu to be deployed on opposite sides of the infantry, so that one wing can hardly be said to be deserted by the other — unless MacDowall (see the comment on 31.12.17) is correct that most of the right-wing cavalry has never actually moved to the Roman right. It seems unlikely that the entire front line is made up of cavalry, with the infantry playing their supporting role (31.12.11) behind them.
It is difficult to comprehend why the wing would be said to have reached the wagons and to be on the point of going farther (ultra processurum). I don’t think cornu can encompass any infantry, as it clearly does not at 31.12.11 and 12, where cornu equitum is twice used in the sense of cavalry forming the wing. Yet surely storming a barricade is a job for infantry? And surely the cavalry, late getting to the field and no doubt exhausted, is not going to dash forward in a daring assault?
Blockley infers that the cavalry are about to wrap around the curved barricade, which may be the correct view; perhaps they hope to win the battle by getting behind the camp. If we further argue from Ammianus’s silence on the matter that the Roman infantry were not truly involved in the assault, other than perhaps by hurling missiles into the camp, then it has to be allowed that the battle of Adrianople was largely fought, and lost, by the left-wing cavalry. They had been ordered not just to guard the flank but to drive off whatever resistance there was in front of the barricade as it curved back to the north. Only when the camp was threatened from the rear would Valens feel confident in sending the legions forward to storm the palisade. Such a scenario would make Valens culpable of sending his most exhausted troops on what was essentially a suicide mission, and perhaps Ammianus means us to understand that, without actually accusing the emperor of a disastrous blunder.
The image of the beaked ships, alternately penetrating into the enemy line and retreating, seems more apt for cavalry movements than for the steady push of heavy infantry; but not too much can be read into Ammianus’s “turgid metaphors,” as Gibbon described them.
Note the emphasis on missile weaponry even after the opposing lines have come together, though most likely the Gothic cavalry are pressing with the lance.
13.3 But when the barbarians, pouring forth in huge hordes, trampled down horse and man, and in the press of ranks no room for retreat could be gained anywhere, and the increased crowding left no opportunity for escape, our soldiers also, showing extreme contempt of falling in the fight, received their death-blows, yet struck down their assailants; and on both sides the strokes of axes split helmet and breastplate.
The pouring forth (effusi) suggests that the bulk of the Gothic warriors are now streaming from the wagon camp, probably through prepared sally ports.
13.4 Here one might see a barbarian filled with lofty courage, his cheeks contracted in a hiss, hamstrung or with right hand severed, or pierced through the side, on the very verge of death threateningly casting about his fierce glance; and by the fall of the combatants on both sides the plains were covered with the bodies of the slain strewn over the ground, while the groans of the dying and of those who had suffered deep wounds caused immense fear when they were heard.
Although Hamilton transforms the hissing cheeks into grinding teeth, Ammianus’s image is quite vivid as it stands: the cheeks are drawn back in a grimace as the warrior expels his breath.
13.5 In this great tumult and confusion the infantry, exhausted by their efforts and the danger, when in turn strength and mind for planning anything were lacking, their lances for the most part broken by constant clashing, content to fight with drawn swords, plunged into the dense masses of the foe, regardless of their lives, seeing all around that every loophole of escape was lost.
As long as they can maintain their shield wall, the infantry at the front of the line keep the enemy at bay with their spears while those behind shower javelins and other missiles into the mass. Now the wall is broken and it is every man for himself.
13.6 The ground covered with streams of blood whirled their slippery foothold from under them, so they could only strain every nerve to sell their lives dearly; and they opposed the onrushing foe with such great resolution that some fell by the weapons of their own comrades. Finally, when the whole scene was discoloured with the hue of dark blood, and wherever men turned their eyes heaps of slain met them, they trod upon the bodies of the dead without mercy.
13.7 Now the sun had risen higher, and when it had finished its course through Leo, and was passing into the house of the heavenly Virgo, scorched the Romans, who were more and more exhausted by hunger and worn out by thirst, as well as distressed by the heavy burden of their armour. Finally our line was broken by the onrushing weight of the barbarians, and since that was the only resort in their last extremity, they took to their heels in disorder as best they could.
Rolfe misunderstands the astrological reference, which tells us nothing about the time of day. A better translation is: “And the sun, being rather high, since it had made its course through Leo and was passing into the house of the heavenly Virgin, scorched the Romans...” The sun is high above the southern horizon because of the season. By Ammianus’s reckoning, it is about to pass into Virgo, where it stays for a month. (In the modern tropical astrological calendar, the transit of Leo does not end until late August. In the sidereal calendar, it ends in September.) Even Blockley, who understands this much, reads the passage as meaning that it is still midday. In fact, the wagons are sighted after noon and probably the battle lines are not drawn up till several hours later. Then the negotiations eat up more time. The battle probably gets underway in the late afternoon. Sunset at that latitude in August is at about 7 p.m.
13.8 While all scattered in flight over unknown paths, the emperor, hedged about by dire terrors, and slowly treading over heaps of corpses, took refuge with the lancers and the mattiarii, who, so long as the vast numbers of the enemy could be sustained, had stood unshaken with bodies firmly planted. On seeing him Trajanus cried that all hope was gone, unless the emperor, abandoned by his body-guard, should at least be protected by his foreign auxiliaries.
The lancers and mattiarii are also paired at 21.13.16, where Constantius despatches them as part of an advance force against Julian. The lancers (lancearii) are not cavalry, but spear-carrying infantry. According to A Glossary of Later Latin, the mattiarii wield the mattiobarbulus, which Vegetius (1.17) describes as a lead-weighted missile or bullet (perhaps something like a heavy lawn dart); but the word used by Vegetius is spelled variously in the manuscripts. The name mattiarii might equally derive from unattested *mattea, a club (cognate to the English “mace”). There is some evidence that the Romans used percussive weapons: see 31.13.3 above, where both sides are wielding axes, and the insignia of the Master of Offices in the Notitia Dignitatum, which clearly show battle-axes among the products of the arms factories. The Notitia Dignitatum lists matiarii and lanciarii (sic) among the “Palatine legions.” Very little is known about the composition of units, and whether their names had more than historical significance.
Libanius (Oration 24) gives Valens the credit of having refused the chance to flee, but as Gibbon says, “The truth of history may disclaim some parts of this panegyric, which cannot strictly be reconciled with the character of Valens or the circumstances of the battle.”
13.9 On hearing this the general called Victor hastened to bring quickly to the emperor’s aid the Batavi, who had been posted not far off as a reserve force; but when he could find none of them, he retired and went away. And in the same way Richomeres and Saturninus made their escape from danger.
Hodgkin is mistaken in saying (p. 148) that this comes Victor is not the same as the magister equitum of 31.12.6. See Prosopography of the Later Roman Empire vol. 1 (at Victor 4) for the well documented and extensive career of this man.
The Batavi are evidently the foreign auxiliaries (adventicio auxilio) referred to in the previous section, although units by that name had been part of the Roman army since at least the early empire. The unit (or its western equivalent) intervened at a crucial time in the Battle of Strasbourg in 357 (16.12.45). Zosimus (4.9) describes a battle in which the Batavi under Valentinian were held responsible for starting the rout of Roman forces; after he threatened to turn them all into slaves, they promised to do better.
Gibbon, apparently through mistranslation, has the Batavi coming up but failing to find the emperor.
13.10 And so the barbarians, their eyes blazing with frenzy, were pursuing our men, in whose veins the blood was chilled with numb horror: some fell without knowing who struck them down, others were buried beneath the mere weight of their assailants; some were slain by the sword of a comrade; for though they often rallied, there was no ground given, nor did anyone spare those who retreated.
The broad picture is one of a fighting retreat in the gathering darkness. They can neither gain ground by rallying nor escape by falling back.
13.11 Besides all this, the roads were blocked by many who lay mortally wounded, lamenting the torment of their wounds; and with them also mounds of fallen horses filled the plains with corpses. To these ever irreparable losses, so costly to the Roman state, a night without the bright light of the moon put an end.
There was a new moon on August 8th by the Julian calendar.
13.12 At the first coming of darkness the emperor, amid the common soldiers [gregarios] as was supposed (for no one asserted that he had seen him or been with him), fell mortally wounded by an arrow, and presently breathed his last breath; and he was never afterwards found anywhere. For since a few of the foe were active for long in the neighbourhood for the purpose of robbing the dead, no one of the fugitives or of the natives ventured to approach the spot.
The church historian Socrates (c. 38; translated by Zenos) says: “Others affirm that having put off his imperial robe he ran into the midst of the main body of infantry, and that when the cavalry revolted and refused to engage, the infantry were surrounded by the barbarians, and completely destroyed in a body.” The revolting cavalry may be those on the left wing who were driven off in the beginning, Socrates having simply mixed up the order of events. They cannot be the Batavi, who are infantry.
If no one claims to have been with the emperor, where did the report come from that he had been killed by an arrow? This may be taken as further evidence of the importance of missiles in combat.
The battlefield must have been a treasure trove for the Goths, who had been disarmed two years earlier when they immigrated.
13.14 Others say that Valens did not give up the ghost at once, but with his bodyguard and a few eunuchs was taken to a peasant’s cottage near by, well fortified in its second storey; and while he was being treated by unskilful hands, he was surrounded by the enemy, who did not know who he was, but was saved from the shame of captivity.
As Blockley points out, not all Valens’s bodyguard have abandoned him, though now they are described as candidati rather than armigeri.
13.15 For while the pursuers were trying to break open the bolted doors, they were assailed with arrows from a balcony of the house; and fearing through the inevitable delay to lose the opportunity for pillage, they piled bundles of straw and firewood about the house, set fire to them, and burned it men and all.
Sozomen (Ecclesiastical History 6.40) gives more circumstantial detail of the burning of the house, or tower. This episode is the one incident of the battle that made it into all subsequent histories. Catholic Christians made a moral tale of it; for example, Jordanes, writing in 551 CE: “Plainly it was a direct judgment of God that he should be burned with fire by the very men whom he had perfidiously led astray [into the Arian heresy] when they sought the true faith, turning them aside from the flame of love into the fire of hell” (Gothic History 138, trans. Charles C. Mierow). See Lenski pp. 340-1 for other examples and an assessment of the truth of the story.
13.16 From it one of the bodyguards leaped through a window, but was taken by the enemy; when he told them what had happened, he filled them with sorrow at being cheated of great glory, in not having taken the ruler of the Roman empire alive. This same young man, having later escaped and returned secretly to our army, gave this account of what had occurred.
13.18 Amid this manifold loss of distinguished men, the deaths of Trajanus and Sebastianus stood out. With them fell thirty-five tribunes, without special assignments [vacantes], and leaders of bodies of troops [numeri, units] as well as Valerianus and Aequitius, the one having charge of the stables, the other, of the Palace. Among these also Potentius lost his life in the first flower of his youth; he was tribune of the promoti, respected by all good men and honoured both for his own services and those of his father Ursicinus, formerly a commander-in-chief. Certain it is that barely a third part of our army escaped.
The account of Potentius’s death is a personal tribute: Ammianus had served under his father. The promoti are cavalry.
13.19 The annals record no such massacre of a battle except the one at Cannae, although the Romans more than once, deceived by trickery due to an adverse breeze of Fortune, yielded for a time to ill-success in their wars, and although the storied dirges of the Greeks have mourned over many a contest.
15.1 After the murderous battle, when night had already spread darkness over the earth, the survivors departed, some to the right, others to the left, or wherever their fear took them, each seeking his nearest associates, for none could see anything save himself, and everyone imagined that the enemy’s sword hung over his own head. Yet there were still heard, though from afar off, the pitiful cries of those who were left behind, the death-rattle of the dying, and the tortured wails of the wounded.
15.2 But at daybreak the victors, like wild beasts roused to cruel ferocity by the provocative tang of blood, driven by the lure of a vain hope, made for Hadrianopolis in dense throngs, intending to destroy the city even at the cost of the utmost dangers; for they had heard through traitors and deserters that the most distinguished officials, the insignia of imperial fortune, and the treasures of Valens were hidden there, as within an impregnable fortress.
Traitors and deserters indeed. The Goths have prisoners who could be interrogated, and everyone knows that the emperor goes nowhere without treasure. This whole paragraph is Ammianus at his deepest purple.
15.3 And in order that no delays meanwhile might cool their ardour, at the fourth hour of the day they had encircled the walls [muri] and were engaged in a most bitter struggle; for the besiegers with their natural ferocity rushed upon swift death, while on the other hand the defenders were encouraged to vigorous resistance with might and main.
On a moonless night, the Goths cannot have pursued far, and most of them probably returned to the camp. In the morning they cover the same distance as the Romans did on the previous day, but they do it in roughly half the time. (Of course, those on foot do not necessarily arrive until later.) They carry on an assault for five hours before the weather — and exhaustion, no doubt — causes them to fall back. They still have plenty of daylight to get back to the wagons.
Attackers are attacking and defenders are defending, but otherwise Ammianus tells us nothing about what is happening. Are the Goths actually hurling themselves against the walls, and are the defenders, at this point, those on the parapets?
15.4 And because a great number of soldiers and batmen [calones, military servants] had been prevented from entering the city with their beasts, they took their place close to the shelter of the walls and in the adjoining buildings, and made a brave fight considering their low position [humilitas]; and the mad rage of their assailants had lasted until the ninth hour of the day, when on a sudden three hundred of our infantry, of those who stood near the very breastworks [lorica], formed a wedge and went over [desciverunt] to the barbarians. They were eagerly seized by the Goths, and (it is not known why) were immediately butchered; and from that time on, it was noticed that not a man thought of any similar action, even when the outlook was most desperate.
The people of Adrianople are very particular about who they admit within the walls. Earlier in the year (31.11.3), they have barred the gates to a force under Sebastianus, thinking they might have gone over to the enemy. The townspeople’s suspicion is not unwarranted, as they have already suffered badly at the hands of the Goths. They also do not want to have mouths to feed during what might be a lengthy siege.
“They took their place close to the shelter of the walls and in the adjoining buildings [affixus parietibus moenium aedibusque continuis].” Parietes are properly the walls of buildings, not city walls, but Ammianus does not make this distinction; see for example 20.11.10, 21.12.6, and most famously 31.6.4, where Fritigern is “at peace with walls.” Here affixus parietibus moenium means something like “staying close to the walls of the fortification.” Some of the adjoining buildings are perhaps built up against this same stone face and might be defensible. But what has happened to the fortified camp? Even if it was abandoned, it is surely still of use to the fugitives.
The survivors put up a brave fight despite their subservient position; most of their officers are dead. Junior officers were in the first rank of the infantry and took disproportionate casualties.
The episode of the 300 infantry is not easy to understand. Probably they are a regiment of Germanic auxiliaries who hope to find a place in the horde. (Some candidati do succeed in deserting, as we learn from 31.15.8.) Perhaps they are the fugitives huddling closest to the city wall, in which case they form a wedge to break through their own lines. But lorica is an odd choice of words for a stone wall. At 24.5.2 the word is used for a large fence enclosing zoo animals, and at 31.3.7 it is an improvised palisade; Rolfe’s “breastworks” is exactly right. Perhaps the men stand at the palisade of the baggage dump or some part of the camp, treating with the enemy, before forcing their way out the gate and into a trap.
15.5 Now, while this accumulation of misfortunes was raging, suddenly with peals of thunder rain poured from the black clouds and scattered the hordes roaring around the city; but they returned to the circular rampart formed by their wagons, and carried their measureless arrogance so far as to send an envoy with a threatening letter, ordering our men to surrender the city on receiving a pledge that their lives would be spared.
The hordes are circumfrementes, defiantly shouting; circum lends the sense of “rounding on” someone. They go back to their “measured palisade of wagons in a perfectly rounded form.”
15.6 The messenger did not dare to enter the city, and the letter was delivered by a certain Christian and read: but it was scorned, as was fitting, and the rest of the day and the whole night were spent in preparing defensive works. For the gates were blocked from within with huge rocks, the unsafe parts of the walls were strengthened, artillery was placed in suitable places for hurling missiles or rocks in all directions, and a supply of water that was sufficient was stored nearby; for on the day before some of those who fought were tormented with thirst almost to the point of death.
Once more we have a Christian messenger, probably a priest, acting as a go-between.
It seems that till now there has been no direct assault on the town, suggesting once more that the survivors outside the walls have put up a good struggle. However, as the sequel shows, there’s not much the Goths can do against walled towns in any amount of time.
To sum up a great deal of detail, these are my tentative conclusions:
What is the connection, if any, between (1) the “strong guards” sent on the raid, (2) the returning party of cavalry under Alatheus and Saphrax, and (3) the Roman underestimate of the Gothic numbers? Ammianus suggests that three days elapsed between the dispatching of the two Roman units and the fateful day of August 9th. The raiding party, finding the pass blocked (not necessarily on the first of those days), would not have tarried long before lack of forage forced them to retrace their steps. But they might have faced a ride of 100 miles or more if they had to recross the hills before riding west to the Tonzos and then south and east after the horde, with horses in need of rest and fodder.
Can Valens have judged it impossible that the raiding force, of which we know he had intelligence, could rejoin the horde before August 10th? It is certainly tempting to identify the two strong cavalry forces as one, and to believe that their unexpected return was what threw out Valens’s calculations; but the neatness of the solution is not in itself proof. Ammianus, of course, does not make the connection; he tells us (31.12.12) only that the cavalry were far off but had been sent for. See further the article by N.J.E. Austin, who sheds much light on the prelude to the battle.
The only other near-contemporary description of the battle comes from the History against the Pagans, written by Orosius in the early fifth century. It would seem to be based on Ammianus’s account. The translation is by Roy J. Deferrari.
32.13. Itaque quinto decimo imperii sui anno lacrimabile illud bellum in Thracia cum Gothis iam tunc exercitatione virium rerumque abundantia instructissimis gessit. ubi primo statim impetu Gothorum perturbatae Romanorum equitum turmae nuda peditum deseruere praesidia.
“Thus, in the fifteenth year of his rule, Valens fought that lamentable war in Thrace with the Goths, who were then very well equipped with strong training and an abundance of resources. As soon as the squadrons [turmae] of Roman cavalry were thrown into confusion by the sudden attack of the Goths, they left the companies of infantrymen without protection.”
Orosius is definite that the entire cavalry on at least one wing was thrown back, leaving the infantry unprotected. See the commentary on 31.13.2.
14. Mox legiones peditum undique equitatu hostium cinctae ac primum nubibus sagittarum obrutae, deinde, cum amentes metu sparsim per devia cogerentur, funditus caesae gladiis insequentum contisque perierunt.
“Then the legions of infantry, becoming surrounded on all sides by the enemy’s cavalry and, when first overwhelmed by showers of arrows and then mad with fear they were driven over devious paths, being completely cut to pieces by the swords and lances of those who were pursuing them, perished.”
Note that again devia are not devious paths or rough ground but out-of-the-way places. The soldiers are fleeing over the ignotos tramites, unknown paths, of Ammianus. In other words, they have scattered.
Orosius then repeats Ammianus’s tale of Valens’s death in the burning house.
Only works directly relevant to the battle are listed. Quoted translations and histories in the public domain can be found online by searching for a phrase; this seems a more permanent solution than any links I might provide here.
 “The battle,” says Hodgkin (p. 147), “...is described with much minuteness but no great clearness by Ammianus. What the professional Roman soldier has failed to make clear, a modern and unprofessional writer may be excused from attempting to explain.” John Matthews (The Roman Empire of Ammianus, pp. 298-9) provides a balancing view: “In circumstances of such genuine confusion, and such variety of individual experience, it is reasonable to ask what a historian could be expected to do other than describe initial dispositions, indicate the terrain and therefore the type of battle that would be fought, give a general outline of its main phases, select notable individual events and exploits..., describe the result, with casualty figures if they were available, and give some impression of the atmosphere of the battle.” The reader can decide whether Ammianus has in fact done all these things. I for one could wish for the opportunity to ask him two or three questions.
 For instance, John Warry in Warfare in the Classical World (University of Oklahoma) p. 207. His detailed diagrams, complete with scale, are pure invention. He also states on the basis of Ammianus 31.13.13 – a historical parallel from the third century – that the “Caesar Decius” died at Adrianople!
 Other studies, mostly in German, are referenced by Lenski pp. 328 ff.
 For the strategic importance of the Adrianople plain, where at least 15 battles have been fought, see John Keegan, A History of Warfare (1993), pp. 70-1.
 Burns (1973), MacDowall (p. 7), and Barbero (p. 99) err in calling this the Via Egnatia. That road lies farther south, and is the principal route between Constantinople and the Adriatic ports.
 The Barrington Atlas shows only the northern portion; doubtless the rest is untraceable because it would have been paved with gravel.
 “The military importance of Adrianople,” says Charles Oman, “was unmistakable; it was a victory of cavalry over infantry.” (The Art of War in the Middle Ages, A.D. 378-1515, Blackwell 1885, p. 6)
 Ammianus (16.12.8) and Claudian (In Rufinum 2.353-65) vividly describe heavily armored cavalry among the imperial forces.
 Rance’s focus is on a somewhat later date, but I adopt his model for the late fourth century if only because it explains a great deal, for example Vegetius’s statement (1.20) that the infantry of Gratian’s time no longer bore armor.
 Ammianus 31.11.5: revocatis omnibus prope Cabylen oppidum cito discessit, ut agentes in regionibus patulis nec inedia nec occultis vexarentur insidiis. “Having recalled everyone to the vicinity of the town of Kabyle, he quickly departed, so that by operating in open country they might be troubled neither by starvation nor by secret ambushes” (my translation).
 According to Zosimus (4.25), in the following year the Roman general Modares, after ambushing some Goths “while they were immersed in voluptuousness,” captured 4,000 wagons. Peter Heather (The Goths, p. 171, citing an article by J.B. Hall) points out that in the Boer and Western American treks there was one wagon for about every six people, and a line of 460 wagons stretched three miles.
 Though Hodgkin is an excellent narrator, and well worth taking off the shelf, his view of this episode (p. 145 n.) is clouded by a careless reading of the text. He has the Goths gathering and marching to Kabyle in their search for security, rather than gathering at Kabyle and then departing for open country; he places them immediately afterward at Nike; and he makes Ammianus’s three days into the duration of Valens’s march to Adrianople, without saying from where — certainly not from Melanthias, which is six or seven days distant.
 Socrates, Church History 4.38. The date is confirmed by Consularia Constantinopolitana, cited below. Ammianus’s account is at 31.11.1 ff.
 History is, of course, full of campaigns where opposing armies separated by a few miles had no intelligence of one another. But the gathering of the Goths at Kabyle, probably over the course of weeks, must have been common knowledge for many miles around, and the Romans would have taken pains to learn what direction the horde took from there. In any case, the route south was one of their obvious choices.
 Lenski (p. 337) suggests that this episode may be a doublet, i.e. Ammianus has confused the earlier campaign with the later one and created another visit by Valens to Nike, prior to the one he must make on his march to Adrianople.
If we assume that Ammianus is correct, it is apparent that sometime after June 11 Valens takes a large force to Nike in preparation for Sebastianus’s offensive (31.11.2). Sebastianus then moves west with the pick of the men, taking fortified places (where? Paschoud p. 381 reduces this part of the campaign to “une nuit passée à Adrinople”), and presumably leaving a weaker but still considerable body of troops at Nike – unless these march back to Melanthias with the emperor. A month or so later, Valens, now inspired by jealousy, or more likely just realizing that Sebastianus’s tactics are not going to win the war, again marches forth from Melanthias.
Zosimus’s account of this period (4.23; cf. Eunapius frag. 8.4 [Blockley]) makes no mention of Nike, instead leaving the impression that Sebastianus is still at Constantinople when he chooses his men (2,000 of them in total; Ammianus says 300 from each unit). It does make sense that Melanthias, the emperor’s seat and more convenient to supplies from the capital, should be the place where Sebastianus assembles his force, especially because we are told (31.11.3) that he then approached Adrianople by “rapid marches” (itineribus celeratis; an iter or itiner is technically one day’s march), whereas if he started from Nike he could easily have been there on the same day. Set against this evidence is Ammianus’s unambiguous account (31.11.2) of the Goths’ reaction to the approach of Valens with a large army, which strongly suggests that he has indeed reached Nike.
There is no clear basis for Wolfram’s claim (p. 125) that Valens reached “the city” (by which he seems to mean Adrianople) in mid-July, unless we see here the pernicious influence of Delbrück: if Valens was west of Adrianople in early August, he must have reached the town in July. Ammianus could be mistaken in locating the emperor as far away as Melanthias immediately before his march to Adrianople (31.12.1), but there is no evidence to suggest that he has advanced farther than Nike at any time before August, and in Ammianus’s account he is still on the road to the city on August 5th or 6th.
 With Delbrück, I take per devia with eruptionem. If it is taken with incederent, as Rolfe and Hamilton have done, the sentence seems to be missing a conjunction: “In the next three days, while the barbarians were advancing — at a slow pace and fearing a sortie — through out-of-the-way places [and,] separated by 15 miles from the civitas, were making for the post at Nike...” The reading I have adopted also seems to be supported by the metrical structure of the sentence as indicated in the Teubner edition; but I am no skilled Latinist. It really comes to the same thing: they fear a sortie coming from Adrianople through the remote places where they themselves are. The important thing is that the sortie does not come “from” remote places, as in the Delbrück/Renfroe translation.
 We should examine in this context the testimony of Socrates (4.38) that when Valens left the capital on June 11, he “advanced against the barbarians, whom he routed with great slaughter, and pursued as far as Adrianople.” Sozomen (6.40), who largely follows Socrates, puts it this way: “When Valens marched out with his army, the Goths retreated while pursued. In his advances he passed by Thrace, and came to Adrianople.” Putting these notices beside the accounts of Ammianus and Zosimus, it seems probable that both the church historians are seeing the campaign of Sebastianus, who cleaned up the Adrianople plain in June and July, as part of the emperor’s advance.
 Hodgkin puts the post 15 miles farther east, more convenient to the Marcianople road. Wolfram puts it 14 miles north of the town. See also Lenski p. 337 n. 103.
 MacDowall pp. 59-60 and the map on p. 62; Barbero p. 94. The idea is also swallowed whole by John Julius Norwich in Byzantium: The Early Centuries, c. 5. But I do not see how Delbrück’s view of the events leading up to the battle can possibly be squared either with the text of Ammianus or with common sense: “But when Valens now [after the gathering of the barbarians at Kabyle] marched on from Adrianople through the Maritza [Hebros] valley toward Philippopolis, he received the astonishing report that the Goths had appeared behind him near Adrianople and were threatening the road to Constantinople. It even appears that Goth cavalry were seen behind the Roman army on the Maritza road, so that it was easy to believe that the Goths intended to cut the emperor’s communications with Adrianople.” (p. 270)
MacDowall has less reason to place the army so far west, since he does not put the Goths anywhere near Nike. This leaves only the supposed route of the raiders down the Tonzos to account for it: they must be east of Valens, therefore Valens must be farther west than Ammianus tells us.
 “Three days later, the barbarians marched to Nike: Valens learns that they are only 10,000 men strong, and he marches to Adrianople to join battle with them. If the barbarians were already at Nike...how is Valens, who was on the march from Melanthias, supposed to have come to Adrianople?” (p. 281)
 It was a mansio (halting-place) for the compiler of the fourth-century Itinerarium Burdigalense. Socrates, however, calls it a city, and tells how a synod was held there solely for the purpose of promulgating a “Nicene creed” that would be confused with the one issued at Nicaea (2.37).
 It’s said that wagon trains migrating to the American west averaged two miles an hour, or ten miles a day.
 “Secure,” says Sozomen (6.40).
 The so-called Consularia Constantinopolitana, formerly ascribed to Idatius, or Hydatius. The entry for 378 reads:
His conss. ingressus est Valens Aug. ab oriente Constantinopolim die III kal. Jun. Et ipso anno profectus est Valens Aug. ex urbe ad fossatum die III idus Junias: et pugna magna fuit cum Romanis et Gothis a milliaro XII ab Hadrianopoli, die V idus Augusti. Ex ea die Valens Augustus nusquam apparuit; et toto anno per dioecesim Thraciarum, et Scythiae, et Moesiae, Gothi habitaverunt simul, et eas praedaverunt; deinde usque ad portas urbis Constantinopolitanae venerunt.
“In this year, Valens came from the east and entered Constantinople on May 30th. On June 11th he advanced from the city to his encampment, and on August 9th there was a great fight between the Romans and the Goths at the twelfth milestone from Adrianople. From that day forward Valens was nowhere to be seen. During the whole year, the Goths dwelt throughout the dioceses of Thrace, Scythia, and Moesia, and depredated them, and then came right up to the gates of the city of Constantinople.”
It is not clear whether fossatum signifies the mustering-ground at the imperial villa of Melanthias (31.11.1), or an intermediate camp at Nike (if Valens did indeed spend some time there as suggested by 31.11.2). A Glossary of Later Latin defines the word as “royal quarters” (perhaps on the basis of this passage, with the assumption that the villa is meant?), but in Chronica Theodericiana (see the Loeb edition of Ammianus, v. 3, p. 540, first and last lines), a fossatum is simply an entrenched camp.
 Burns (1994) p. 32. His statement that “whatever the Romans knew about the whereabouts of the Goths was obsolete” is contradicted by 31.12.3 and 31.12.7, which cite Roman scouting reports, current up to the latest encampment.
 Austin convincingly elaborates the theory that the Romans thought they were marching only against the Gothic infantry. “By only half a day,” he concludes, “Valens failed to achieve what he had set out to do.”
 A normal day’s march was 20 Roman miles (30 km) in five summertime hours, or 24 miles at the full step (plenus gradus). That would be on roads, of course, but it is a measure of the Roman soldier’s endurance. See Michael Whitby in The Cambridge History of Greek and Roman Warfare, p. 329, and Vegetius 1.9.
A parallel from the account of the battle of Strasbourg (16.12.7 ff.) provides some insight into how much ground an army of a similar size could cover in a day. The distance from the Roman camp to the barbarian entrenchment, says Ammianus, is 21 miles — exactly the distance between Julian’s last stated location, Tres Tabernae (Saverne), and Argentorate (Strasbourg). At dawn the army sets out at a moderate pace, and at midday Julian calls a halt and advises that it would be wiser to make camp and continue on the following morning. His arguments about the difficulties of the march ahead, including the darkness of the coming night, suggest that the army has covered less than half the distance. However, we cannot believe that Julian literally gathered his miles-long column of 13,000 soldiers for a noonday discussion; the passage is rhetorical and intended to highlight his prudent generalship, perhaps even to contrast it with that of Valens on a later occasion. In any case, the eager soldiers are then reported as stating that the enemy is already within sight. We hear no more of the German entrenchment, and it is impossible to know whether the battlefield itself is 21 miles from the army’s starting-point. But since the day ends with the barbarians being pursued into the Rhine, the Romans must have covered at least the stated distance by nightfall, besides fighting a major battle.
 Bearing in mind, of course, Gibbon’s warning: “The difference of the eight miles of Ammianus, and the twelve of Idatius, can only embarrass those critics...who suppose a great army to be a mathematical point, without space or dimensions.” (Chapter 26, p. 116, n. 94 in the Bury edition)
 F. Runkel, Die Schlacht bei Adrianopel (Diss. Rostock, 1903). Runkel’s view is said to be modified by Ulrich Wanke in Die Gotenkriege des Valens (1990). I have not been able to consult either work.
 Lenski p. 338; Potter p. 531 n. 27. François Paschoud in his commentary on Zosimus (vol. 2 p. 382), following Otto Seeck, places the final Gothic encampment northwest of Adrianople, and John Curran (Cambridge Ancient History, vol. 13 p. 100) also has Valens marching northwest from the city. I have been unable to consult Seeck to learn the basis for this interpretation.
 Streambeds that are now dry in August might have held water in the days before the hills were deforested.
 One of the few to tackle the issue directly is Roger Tomlin (in Peter Connolly, Greece and Rome at War, Greenhill/Stackpole, pp. 257-8): “The prime cause of disaster at Adrianople would seem to be the decision to assault a field fortification...while the enemy’s powerful cavalry was uncommitted. This ‘decision’ was forced upon Valens by the undisciplined and incompetent cavalry on his right wing.” However, there is no basis for assigning the rash attack by the “archers and targeteers” to the right wing.
 For sources and a contrary view see Peter Heather, The Fall of the Roman Empire (Oxford, 2006) p. 508 n. 29.
 E.g. Peter Heather, Goths and Romans 332-489 (Oxford, 1991) p. 146.
 For details of Valens’s command structure at this time, see Lenski pp. 362-3.
 So John Curran in The Cambridge Ancient History, vol. 13 p. 99.
 At 4.23: “Sebastianus told him to stay in the capital [lit. ‘where he was’] and not advance any further, saying it was not easy to conduct an open war with such a numerous enemy. It was better to prolong it with manoeuvres and ambushes until, weakened by lack of provisions, they either surrendered or retreated from Roman territory, preferring to give themselves up to the Huns rather than suffer the pitiful destruction usually resulting from a famine. Although Sebastianus advised this course of action, his opponents, eager for the emperor to advance with his whole force, encouraged the last to go to war, saying that the barbarians were almost all destroyed and he was close to an easy victory. This worse counsel prevailed, since Fate was guiding events to disaster, and the emperor led his whole army to battle in complete disarray.” (trans. R.T. Ridley)
Zosimus, who is epitomizing, has likely compressed two incidents into one. Sebastianus was an ambitious general who had met with some success in encounters with barbarian splinter parties. It is believable that in late June or July he made a strong case to the emperor for his methods: why risk the army when with his 2,000 men he has forced the enemy into one mass that can do nothing but keep moving or starve? It is equally believable that, as Ammianus tells us, he advised the emperor on August 8th to give battle immediately, while some part of the horde had not yet come up with Fritigern. Both things can have happened without inconsistency. Paschoud (p. 382) comes to the same conclusion: “Sébastien peut avoir été honnêtement induit en erreur par le faux rapport des éclaireurs sur les effectifs des ennemis, et donc avoir eu un motif raisonnable et loyal de conseiller un changement de tactique.”
 Because Ammianus is in general so reliable a guide, it is easy to forget that he is also a melodramatist, and his explanations of motives and causes are often simplistic. His portrait of Valens in the last weeks of his life – jealous, impetuous, easily flattered, last glimpsed as a ghastly figure stalking over corpses in the blinding dust – is as close to the real man as, say, Shakespeare’s Richard III.
 Burns (1973), following Runkel, says the area is characterized by rugged hills that would have been barren and rock-strewn. This is true only of the area to the north of any possible battlefield. Satellite images and elevations do not show any significant hills between Edirne and Demirhanli; on the contrary, a line of march from the eastern outskirts of the city does not vary in altitude by more than 100 feet and for long stretches is almost completely flat. Ammianus similarly exaggerates the difficulty of the march to Strasbourg, when he has Julian warn his troops of the “steep and blind paths” that await them (16.12.11). Yet the march from Tres Tabernae was down into the valley of the Rhine along a straight Roman road.
 Claudian describes a large, enclosed wagon camp some decades later: “[The Goths] gathered together in the plain and enclosed their pasture lands within a defensive ring. They then built an impregnable fortification with a double moat, planted stakes two deep at intervals along its summit and set wagons rigged with ox-hide all round like a wall” (In Rufinum 2.166 ff., trans. Maurice Platnauer). But this was a more permanent arrangement than the camps of Fritigern and his people in early August 378.
 Rutilius Namatianus (De Reditu Suo 1.201-4) claims to have heard the shouting in the Circus Maximus from his ship’s deck at Ostia, 15 miles distant – though he admits that the familiarity of some of the voices was the work of his imagination. On Gothic signalling, see also Ammianus 31.7.7, with Rolfe’s note.
 Pace Hodgkin pp. 146-7.
 The Cambridge History of Greek and Roman Warfare, pp. 351-2. Note that the illustration (better reproduced in E.A. Thompson, A Roman Reformer and Inventor, Oxford, 1952, fig. IX) does not show equipment as it was, but as the anonymous author of that strange and wonderful pamphlet De Rebus Bellicis wished it could be. The vestments on the right are a felt tunic or thoracomachus, invented (he says) by the ancients to be worn under armor, and a fleece overcoat for wet weather, which seems to be his own idea. What appears to be an ax oddly propped against the model’s hip was surely, in the original art, a sword hanging from the baldric on the opposite side, and indeed it is so rendered in another manuscript copied from the same archetype.
Nowhere is body armor shown, and the text of the manuscript does not actually state that body armor was normally worn in the writer’s time: “So when...the soldier has donned this Thoracomachus (which has adopted this name from the Greek because it protects the body), and has put on socci too (that is, boots), and iron greaves, with a helmet on his head and a shield and a sword fitted to his side and has caught up his lances in his hand he will be fully armed to enter an infantry battle.” (trans. Thompson p. 118)
It’s hard to believe that the pamphleteer is proposing a (thick?) felt tunic as armor in itself (unless, like Vegetius, he thinks the average Roman soldier unwilling to bear anything heavier), and he is surely not recommending it for its warmth, for which purpose he supplies the fleece overcoat. On the other hand, if the author’s intention is to revive the use of the thoracomachus in its original role, to protect against the chafing and heat of metal armor, it is very odd that such armor is neither mentioned in the text nor shown in the illustration.
 See PLRE vol. 1, at Equitius.
 PLRE vol. 1.
 Hugh Elton, Warfare in Roman Europe, p. 106; MacDowall p. 73.
 Sidnell p. 289. Note also that the Latin (reversus) can mean that the Gothic cavalry had already returned, rather than that they made their attack as they arrived.
 On weapons of this type, see Pat Southern and Karen Ramsey Dixon, The Late Roman Army (Yale), pp. 113-15.
 Shown by the Oxford English Dictionary as *mat(t)ea. Rolfe says mattium.
 The two versions in the manuscript at Munich can be seen here (elaborated by the artist) and here (a poor but more faithful rendering from the lost archetype, the Codex Spirensis). Two others from the Bodleian Library are reproduced in MacDowell, p. 35; the one on the left appears to be one of the surviving watercolors based on the Codex, which are fairly free renditions. I cannot identify the two objects under the rightmost shield in the manuscript versions of the picture (nor, it seems, could the illuminators), but perhaps they were originally mace-like weapons.