The traditional narrative of the late east Roman military recognizes two major phases. In the first phase, the emperor Diocletian (284–305) codified and expanded military innovations that had evolved during the series of military and political crises that afflicted the empire during the third century. The result was a new overall structure for the Roman military. In particular, Diocletian is credited with initiating the division of the Roman army into two broad groups: mobile field armies (the comitatus, whose soldiers were the comitatenses) and frontier soldiers (limitanei).Footnote ¹ The limitanei were intended to suppress small-scale raiding and generally defend and patrol the borders of the empire, while the mobile field armies represented the main battle forces of the empire and accompanied each emperor during his travels along the frontier as he took on major invaders and domestic usurpers.Footnote ²

The second phase, still according to the standard narrative, begins with the battle of Adrianople in 378, when the field armies of the eastern empire were crushed by the Goths. After the battle, Theodosius I (379–395) assumed the throne in the east and, allegedly, undertook a massive reorganization of the east Roman military, a snapshot of which is preserved in the Notitia dignitatum, a list of offices, provinces, and military units broadly agreed to date to the end of his reign or the beginning of that of his son Arcadius (395–408). The Notitia system reflects a division of the comitatus into five regional armies: Illyricum, Thrace, Oriens, and two praesentals (i.e., “armies in the presence of the emperor”) that were likely stationed around Constantinople, in eastern Thrace and Bithynia.Footnote ³ This system is believed to have continued to operate in the east until the reign of Herakleios (610–641) when the Avar and Slavic conquest of the Balkans and the Islamic conquest of Egypt, Palestine, and Syria forced the east Romans to radically reimagine their military system.

The reign of the emperor Theodosius I therefore plays an epochal role in the history of the east Roman military. However, this reconstruction is built on shaky ground. To begin with, our best guide to the Roman military of the fourth century, the history of Ammianus Marcellinus, a retired officer with a keen interest in military affairs, ends in 378 with the battle of Adrianople. Our sources for the reign of Theodosius itself are far less interested in, and detailed about, military affairs. We rely, for instance, on court panegyrics to reconstruct basic information about that emperor’s major Gothic war of 379–382.Footnote ⁴ Militarily, therefore, the last quarter of the fourth century is an obscure period.

Traditionally, scholars have looked to the Notitia dignitatum to illuminate the changes taking place in this period and fill in the gaps, in particular between Theodosius’ Gothic settlement of 382 and his departure for the west in 394 to suppress the usurper Eugenius. Scholars who argue for an early date for the Notitia and the military system it describes assign responsibility for the major relevant reforms to Theodosius I. This assignment is not driven by the testimony of the sources, whether explicit or inferred from changing patterns in appointments, but is instead the result of self-imposed necessity. The Notitia clearly does not reflect the military system operating in Ammianus’ narrative, which ends in 378, while the Notitia is dated by most scholars to 395 or shortly thereafter. Thus, scholars assign the reforms to Theodosius I. This assignment is enabled by our poor sources for the Theodosian army, which allows the Notitia to fill in the gaps left by our otherwise meager evidence. Moreover, historians link this alleged reform to a polemical passage in Zosimos, where they see a reference to the establishment of the Notitia system. However, a careful evaluation of the sources, one that does not assume an early date for the Notitia, reveals that there is no evidence to suggest that anything like the Notitia system was in place by 395, either in the offices of the Roman high command or in the organization and deployment of the armies.

This book argues that there is no indication that the system described by the eastern Notitia came into effect before the 440s and, moreover, that there is much positive evidence that it did come into effect in that later decade. This requires that the document be redated accordingly.Footnote ⁵ In turn, deprived of the evidence of the Notitia, the traditional understanding of the Theodosian military reforms becomes immediately untenable. In fact, pulling the Notitia out of the equation has profound consequences that extend back well before the reign of Theodosius itself because scholars have imposed aspects of the Notitia system even onto the army of Valens (364–378). However, the contemporary sources for all these reigns – straight through to the 440s – show that the Notitia structure was not in effect. This chapter will, for the first time, present this evidence without trying to force it teleologically into the mold of the Notitia.

As mentioned, the armies of the fourth-century Roman empire were divided into mobile field armies and limitanei. The field armies were led by the emperor or by top-tier generals, whose titles were magistri peditum (“masters of the infantry”), magistri equitum (“masters of the cavalry”), or the generic magistri militum (“masters of the soldiers,” which is first attested in 349 and abbreviated here as MM).Footnote ⁶ The field armies were not tied to (or named after) specific regions but operated on an ad hoc basis wherever there was need to repel an invader or suppress a usurper. They might be stationed for long periods in regular trouble spots, but they also expected to be moved around a lot, anywhere from Britain to Syria. Like their armies, the field army generals were less often posted on a permanent basis to specific regions than dispatched to deal with specific crises. They too were transferred across the empire during their careers.Footnote ⁷ To be sure, some regions, such as Illyricum and Thrace, were frequent trouble spots during this period, so the field armies were often operating there, but there is no evidence that specific commands or armies were formally or regularly designated for them, such as we find in the Notitia and in the command system that was in effect after the 440s.

It is conventional in the scholarship to refer to the armies that accompanied the emperor as “praesental” or in praesenti (i.e., “in the emperor’s presence”). The term is unobjectionable if used in a descriptive, nontechnical sense.Footnote ⁸ The Latin participle praesens means “to be present,” or “here and now,” as did the Latin adverb praesto. Thus, Constantine in 326 referred to his elite soldiers as “always present” with him (praesto sunt semper).Footnote ⁹ In 398, Arcadius issued a law in Constantinople restricting the side-jobs that soldiers could undertake “who are present (praesentes) in our divine following (obsequium) … when our court (comitatus) is present (praesens) in this City.”Footnote ¹⁰ This use of the participle was not technical, as shown by the fact that the emperor uses it both for the soldiers (present at the court) and himself (present in the City). Likewise, Ammianus could refer to the praetorian prefects attached to the emperor’s court as praesens.Footnote ¹¹ This remains a descriptive usage and he never uses the term praesentalis.Footnote ¹² Before the 440s, the term praesentalis does not appear in these officers’ titles, for example, when those titles are attested in the laws.Footnote ¹³ However, the term suddenly appears in that way in the 440s as part of a military reorganization in the east, as we will show in Chapter 2.

Until the reign of Constantius II (337–361), the offices of magister equitum and magister peditum were separate and singular, that is there was never more than one magister equitum and one magister peditum for each emperor, though Constantius’ cousin Julian appears to have had only magistri equitum during his time as Caesar in Gaul in the 350s (and only one at a time).Footnote ¹⁴ Moreover, despite their titles, which distinguish between infantry and cavalry commands, it does not appear that, by the reign of Constantius II at the latest, the magistri commanded only cavalry or infantry units, at least not when detached from the emperor’s presence.Footnote ¹⁵ It is not clear if this was also the case under Constantine I.

The structure of the high command began to shift again during the reign of Julian (361–363). During his march eastward to confront Constantius II in 361, Julian appointed two magistri equitum, Nevitta and Iovinus.Footnote ¹⁶ As no magistri peditum are attested under Julian, it appears that the total number of magistri continued to be two.Footnote ¹⁷ Let us focus, then, on Nevitta and Iovinus. After the death of Constantius in November 361, Nevitta remained at Julian’s side through to the Persian campaign of 363, while Iovinus was given a sequence of two regional commands, first in Illyricum and then in Gaul, whose nomenclature (at least as conveyed in Ammianus’ literary narrative) anticipates, but is not identical to, that of the regional field armies found later in the Notitia. Specifically, Ammianus calls Iovinus magister equitum per Illyricum and then magister armorum per Gallias.Footnote ¹⁸ Ammianus does not explicitly say why Julian needed a more-or-less designated commander for those regions, but their strategic sensitivity for the emperor is apparent. Illyricum was the crucial node between the western and eastern empires, where the civil war between Julian and Constantius II had largely played out. Julian had massively overextended his forces in Illyricum, partly under Iovinus himself, in order to surprise his cousin. As soon as the war was over, Julian needed to lock that region down. As for Gaul, Julian, who intended to stay in the east without a western co-emperor or Caesar, needed a reliable general there. Unfortunately, we have no indication as to the size and nature of the forces that Iovinus commanded in those two posts.

Despite the relative continuity and stability of these offices, their precise nomenclature was in flux during the second half of the fourth century. In Ammianus, who often employs an informal military argot in place of technical titles, we find several generals with the title magister armorum, which is unattested in official sources such as the Theodosian Code but demonstrably equivalent, in Ammianus’ usage, to magister equitum.Footnote ¹⁹ During the same period, official sources begin to mention individual generals with the compound title magister equitum et peditum (“master of the cavalry and infantry”). Silvanus, for example, is addressed in nearly contemporary laws of about 350 as both magister equitum et peditum and magister militum, and he appears by implication in Ammianus to have been magister armorum, as that was the title given by the historian to his “successor” Ursicinus.Footnote ²⁰ These titles seem to have been used interchangeably. Iovinus is likewise called magister armorum in Ammianus, but addressed by two laws of 365 as magister equitum and magister equitum et peditum, and again in 367 as magister equitum.Footnote ²¹ It is noteworthy that in all of these cases, the new and enhanced titles are given only to generals whose base title was magister equitum and never to those whose base title was magister peditum.Footnote ²² It therefore seems likely that the new compound title, magister equitum et peditum (or, colloquially, magister armorum), was introduced in the mid-fourth century and was equivalent in rank to magister equitum. Still, as we will see, both magistri equitum and magistri peditum continue to be attested separately into the late fourth and early fifth century.

Although the magistri sat atop the military hierarchy of the later fourth century, there was, just below them, a corps of senior officers known by the title comites rei militaris or “counts of military affairs.”Footnote ²³ Although their responsibilities could vary widely, many of these men were senior military commanders whose remit placed them just below the magistri.Footnote ²⁴ For example, Sebastianus and the future usurper Procopius were both comites when they were given charge of an army of 30,000 men during Julian’s Persian expedition in 363.Footnote ²⁵ Comites also occupied regional commands along the frontier, a role that anticipated the MMs who oversaw regional field armies in the Notitia, and this was sometimes reflected in their titles (or, at least, Ammianus’ descriptions of their responsibilities).Footnote ²⁶ The comites rei militaris introduced flexibility into a military system dominated by the two magistri, and allowed emperors to place multiple independent armies in the field simultaneously, when necessary.

The picture so far presented accords generally with the scholarly consensus. Our reconstruction begins to diverge, however, in the aftermath of Julian’s Persian expedition of 363. Our sources report that Julian invaded Persia with an army of roughly 65,000 soldiers, a number that is credible, though just barely, for the mobile forces at the empire’s disposal. This was quite possibly the largest army ever fielded by the late Roman state.Footnote ²⁷ Despite this massive army, the expedition ended in disaster with the death of Julian in battle and a hasty peace arranged by his short-lived successor Jovian (363–364). After Jovian’s death, the army chose Valentinian as emperor (364–375) and he subsequently elevated his brother Valens (364–378). The brothers then divided the empire and the army. Unknown to them, they were laying the foundations for the distinct military establishments that would develop in significantly different directions in the eastern and western halves of the empire over the coming century:

Ammianus is here somewhat imprecise about the offices of the high command. Iovinus is called magister armorum, that is a magister equitum et peditum, but Dagalaifus’ office is more obscure. He is called a rector militiae (“overseer of the soldiery”), which is not an official title, but it does parallel the title that Ammianus gives him earlier in the same book, rector equestris militiae (“overseer of the cavalry soldiery”), which is apparently synonymous with magister equitum.Footnote ³⁰ Thus, there appear to have been two magistri equitum or magistri equitum et peditum under the emperor Jovian (363–364), just as there had been during Julian’s revolt, though it is possible, given Ammianus’ imprecision with titles, that Dagalaifus was, in fact, a magister peditum.Footnote ³¹ Both of these men went west with Valentinian as his two magistri. But Jovian had apparently introduced an innovation: in addition to those two generals, he had appointed Lupicinius as magister equitum with a command that Ammianus marks out as exceptional, for he was specifically responsible for the “districts of Oriens” (partes Eoae). The historian here uses a transliteration of ἐῴα, the Greek term for Oriens, making Lupicinus a forerunner of the magister militum per Orientem, which, as we will see, was the first of the regional field army commands listed by the Notitia to be implemented in the eastern empire (though this happened a generation after Lupicinus’ command).

The ranks of Victor and Arinthaeus are not explicitly mentioned in the passage quoted above, but for the year 366/7 we receive explicit confirmation from Ammianus that Victor was magister equitum and Arinthaeus magister peditum.Footnote ³² It is thus likely that both men held these offices beginning in 364, when we have a law addressing Victor as magister militum.Footnote ³³ Thus Valens, like his brother, also had two top-level magistri with him, but he also inherited Lupicinus’ third command “in the east.”

The high command of the brother emperors in 364 anticipates the lines along which the Roman military would evolve in the coming century. Each of the two emperors continued to be served by two magistri – Iovinus and Dagalaifus in the west, Victor and Arinthaeus in the east. At the same time, the central command was beginning to be supplemented by regional magistri, with Iovinus posted to Illyricum and Gaul under Julian, and Lupicinus posted to Oriens under Jovian and Valens. We should not, of course, understand Ammianus’ geographic descriptions as accurate reflections of official titles or strictly delineated areas of responsibility. That is, we cannot reconstruct the precise remit of Lupicinus’ command or the extent to which Iovinus’ command was geographically delimited (if it was at all). But Ammianus’ testimony does offer us insight into the perceptions of this system held by an informed and interested contemporary observer, and it is clear that he understood certain magisterial commands to be linked, at least de facto, to specific regions.

Victor continued in his position as magister equitum until at least 378, but the final secure attestation of Arinthaeus is during his consulship in 372, when papyri show that he was still magister peditum.Footnote ³⁴ Lupicinus, meanwhile, is last attested in office during the revolt of Procopius in 365–366 and was certainly replaced by 371, when one Iulius is explicitly attested as magister equitum et peditum by a firmly dated inscription in the east (Palestine).Footnote ³⁵ Iulius continued to serve until the battle of Adrianople in 378 and beyond, assuming that he is to be identified with the Iulius whom Ammianus, in another nontechnical descriptor, calls magister militiae trans Taurum (“master of the soldiery beyond the Taurus”). In the aftermath of the battle, Iulius coordinated the ambush and slaughter of many Goths serving in Roman armies throughout Oriens (orientales provinciae).Footnote ³⁶ Iulius was thus the successor to Lupicinus’ ad hoc eastern command. This, then, was the shape of Valens’ high command when the Goths arrived on the Danube in 376: the highest-ranking general was the magister equitum Victor, a long-serving veteran who had occupied the post for at least twelve years; the post of magister peditum appears to have been left vacant following the death of the equally experienced Arinthaeus; and the eastern frontier, though under Valens’ direct supervision in 376, retained a distinct military apparatus under Iulius, a regional magister equitum et peditum.

In 376, a group of Goths, under pressure in their homeland from the recently arrived Huns, famously approached the Roman limitanei on the Danube and petitioned to cross into Roman territory. These Goths sought lands and safety in the empire in return for providing military service to the emperor, a trade that Germanic tribes had been making with Rome for centuries. The emperor Valens, preoccupied with a war against Persia and an uprising in Isauria,Footnote ³⁷ granted them entry, whereupon mismanagement and exploitation by imperial officials drove the Goths to armed resistance. More Goths crossed the destabilized Danube defenses, and disaffected members of Roman society (such as slaves) made common cause with the invaders. The result was widespread destruction in Thrace.

By 377, Valens had sent his magister equitum Victor to sue for peace with Persia in order to lead his forces west and deal with the Gothic threat.Footnote ³⁸ In the meantime, Valens sent on ahead two commanders, Profuturus and Traianus, and he wrote to the emperor of the west, his nephew Gratian (367–383), asking for reinforcements.Footnote ³⁹ Those reinforcements came under the leadership of Richomer, Gratian’s comes domesticorum (i.e., the captain of the soldiers assigned to the imperial household). After joining forces with the east Roman taskforce under Traianus and Profuturus, Richomer fought a bloody battle with the Goths near the town of Ad Salices. The result was indecisive. Richomer returned west to seek further reinforcements, while Profuturus and Traianus withdrew to Marcianopolis, where they were subsequently reinforced by the comes Saturninus, whom Valens had sent ahead with a force of cavalry.Footnote ⁴⁰ Despite these reinforcements, the Romans felt that they could not hold the passes through the Haemus range and so withdrew to the south of Thrace to await the arrival of Valens with the main Roman battle force, effectively abandoning most of the Balkans to Gothic raids.

In the meantime, Valens, who marched in person to Thrace, continued to appeal to Gratian for reinforcements, while his generals, during the campaigning season in 378, organized ambuscades with picked forces in preparation for a major confrontation with the Goths. That spring, Gratian set out with an army to assist his uncle, only to be delayed by an ad hoc campaign against a group of Alamanni. Gratian’s delay prompted Valens to seek battle without western reinforcements and, on August 9, 378, a hot and dusty summer day, the east Roman field army marched out to meet the Goths near the city of Adrianople. The Romans were defeated and took massive casualties.

The size of Valens’ army and that of the Goths at Adrianople is unknown.Footnote ⁴¹ Ammianus reports only that Valens “was leading many units that were neither contemptible nor lazy, especially because he had added to these many veterans.”Footnote ⁴² This is, unfortunately, our most detailed report.Footnote ⁴³ Ammianus’ vagueness about the size of the Roman army at Adrianople, in turn, makes it difficult to quantify his claim that “it is agreed that scarcely one-third of the army escaped.”Footnote ⁴⁴ However, as we will see, the damage to Roman military manpower must have been severe, because for a generation afterward the east Roman army looked very different from its pre-378 form. Extensive also was the damage to the officer corps: two senior generals in Valens’ entourage were killed in action, as were Traianus and Sebastianus; thirty-five tribunes, both many without specific assignments (vacantes) and many commanding units (numeri); the officials in charge of the imperial palace and stables; and most famously the emperor himself.Footnote ⁴⁵ Despite these losses, the Roman high command survived Adrianople: Victor escaped the battle, while Iulius continued to serve in the east, where he responded to news of the disaster by organizing the massacre of Gothic soldiers by their Roman counterparts.Footnote ⁴⁶ It was Victor who brought news of the battle to Gratian, who was at Sirmium.Footnote ⁴⁷ This is Victor’s final attested act in office and marks the end of a career spent in the shadow of two epochal failures of Roman arms, Julian’s Persian expedition and the battle of Adrianople. He was still alive, and in retirement, in 380/1.Footnote ⁴⁸

Only those two men, Victor and Iulius, are explicitly attested as holding magisterial commands at the battle of Adrianople. Scholars have tried to add to their number, producing in some cases a whole college of magistri up to seven strong, most of whom fought at Adrianople. These other officers, who are not explicitly attested as magistri in the sources, are studied separately in Appendix 1, where we find that none of them were likely to have held that rank. Historians have attributed it to them in order to create a bridge between the two-magistri system that prevailed before 378 and the five-magistri system of the Notitia, which they believe was introduced by the next emperor, Theodosius I. They are enabled in this by the literary vagueness with which Ammianus sometimes describes their ranks, but this vagueness does not authorize the highly specific attributions that have been postulated, far less the assumption that a proto-Notitia system was in effect.

Even on its face, the proposed link between Adrianople and the Notitia system does not hold up to scrutiny. The armies described by the Notitia were formally established field armies with specific units and regional remits. By contrast, the forces Valens poured into the Balkans during the crisis were ad hoc deployments of units drawn from other theaters, such as Armenia, and the comitatus, which followed the emperor. These are not analogous structures and there is no reason to believe, even if we were inclined to accept the proliferation of magistri, that the command structure of 378 in any way anticipated or inspired the later system. There is one more problem. If we accept the inflated number of magistri whom scholars have imagined operating during this period, at least four of them are explicitly attested with Valens at Adrianople: Traianus, Sebastianus, Victor, and Saturninus. But there is no parallel in Roman history, before or after Adrianople, for four magistri militum being active at the same time in the presence of the emperor, much less at the same battle.

We find that the high command of the eastern empire was not characterized by radical innovation in the lead-up to Adrianople. Instead, it was marked by continuity with established practice. This conclusion extends the recent scholarly rehabilitation of Valens, whose reign has too often been judged in the shadow of the disaster at Adrianople.Footnote ⁴⁹ Whatever his other mistakes, we do not find Valens scrambling in the years 376–378, multiplying generals without regard to their effect on the chain of command. Instead, we see an emperor responding to a major crisis as promptly as the logistics of ancient warfare and diplomacy allowed.

After Adrianople, the Romans ceded the field to the Goths and it would be years before they were able to successfully challenge them again in open battle. Instead, the Roman military played to its strengths and confronted the Goths from behind formidable walls and with the aid of advanced artillery, seeing as the Goths were incapable of conducting effective siege warfare. In this way, the Romans repulsed the Gothic attempt to take the city of Adrianople after the battle, while a later assault on Constantinople was defeated with the memorable participation of a recently arrived detachment of Saracen auxiliaries.Footnote ⁵⁰

We cannot track in detail the political aftermath of Adrianople or how the Roman government responded to the disaster in the fall of 378. The narrative of Ammianus ends just after the battle and our subsequent sources do not give a clear account of what followed. All we can say for certain is that there followed a series of intrigues at Gratian’s court that resulted in the return from Spanish exile of one Theodosius, the son of a general who had been executed for treason under murky circumstances in 375/6. Theodosius was sent to the Balkans to restore order and was made co-emperor in January 379, perhaps after a minor initial victory.Footnote ⁵¹

Ammianus’ account of Roman casualties at Adrianople suggests that Theodosius’ immediate problem in 379 was manpower, and our other sources confirm this picture. The Theodosian Code contains a flurry of laws dated between 379 and 382, in which Theodosius addressed issues of recruitment, desertion, and the failure of the sons of veterans to enroll in the army, as was legally required.Footnote ⁵² The penalties involved could be harsh – estate managers were to be executed if deserters were found working under their supervision – but they were matched by the desperate measures that men took to avoid service, such as cutting off their own fingers to render themselves ineligible; these self-mutilations are even attested in the papyri.Footnote ⁵³ Theodosius responded by drafting these men anyway.

These new recruits were poor replacements for the experienced comitatenses lost at Adrianople, and contemporaries expressed pessimism about the state of Roman arms: “the Goths have beaten us every time they have fought us. But we die well, as is fitting for noble men.”Footnote ⁵⁴ Recruits had always come overwhelmingly from the agricultural labor force, but now their numbers aroused comment, and court propagandists attempted to spin this in Theodosius’ favor: “you [Theodosius] make farmers fearsome to barbarians.”Footnote ⁵⁵ The empire did not depend entirely on raw recruits, however, as Theodosius transferred existing units from the east, from the frontiers of Syria and from Egypt, to the Balkans.Footnote ⁵⁶ He made good these transfers by sending back to those provinces new recruits from among Gothic deserters. Ironically, the events of 376–378 had only increased Roman dependence on barbarian manpower, but that dependence coincided with a surge in anti-Gothic sentiment. During this period, Roman citizens lynched a Gothic commander, while Roman soldiers repeatedly and violently clashed with Gothic recruits.Footnote ⁵⁷

Despite these reinforcements, the east Romans were unable to openly confront the Goths on the field during this period. When Theodosius tried to do so in 380, using units brought from Egypt, his army was crushed and he himself barely escaped.Footnote ⁵⁸ The emperor subsequently left the war to his subordinates and retired to Constantinople, where he sought another kind of victory by meddling in Church politics.

We can reconstruct only the broad outlines of the ensuing Gothic war. The Romans relied on ambush tactics, diplomacy, and small-scale battles designed to reestablish the Danube frontier.Footnote ⁵⁹ That is, they were fighting a war of attrition that favored the resources and infrastructure of the empire. In 380/1, the western empire intervened by sending an army under the command of Bauto and Arbogast, which successfully drove the Goths out of Illyricum into Thrace and to the negotiating table. The result was a negotiated peace between an exhausted empire and an undefeated but harried force of Goths. The MM Saturninus, one of the officers who had survived Adrianople and had since been promoted, negotiated the treaty for the Romans. In return for this service, he received the consulship in 383 along with a panegyric that is one of our chief sources for the settlement of the war.Footnote ⁶⁰

The terms of the treaty of 382 can only be vaguely recovered, but their cumulative effect was to cede a portion of Roman territory to a semi-autonomous Gothic statelet.Footnote ⁶¹ This was a radical departure from established Roman policy and can only be understood as an admission by the emperors that the Roman military could not defeat the Goths after Adrianople. This explains some of the terms of the treaty, in particular the Gothic obligation to provide auxiliaries under their own commanders, and the ability of Goths to enroll in the Roman army as regular recruits. The Romans were playing for time to rebuild their forces. If they could metabolize Gothic military manpower and grant tax relief to their citizens in the meantime, all the better.Footnote ⁶²

We can discern little of the Roman high command or the organization of the Roman military in the era of Theodosius’ Gothic wars. There is little indication that the limitanei remained in control of the Danube border, while the comitatus of the east appears to have been too badly mauled to resist Gothic depredations. Yet in the second half of the 380s, the Roman order can be seen reestablishing itself along traditional lines. By 386, frontier defense had been reestablished on the Danube. In that year, the general Promotus, either a dux or comes rei militaris, led a force of limitanei and river patrol craft to a significant victory against a Gothic group, the Greuthungi, who were trying to force a crossing. Theodosius celebrated a triumph for this victory and settled some of the survivors in Phrygia.Footnote ⁶³ By 388, the eastern comitatus had likewise returned to its pre-Adrianople structure. In that year, Theodosius confronted the usurper Magnus Maximus (383–388), who had rebelled against Gratian in 383 and expelled Gratian’s half-brother Valentinian II (375–392) from the west in 387. In this war, Theodosius’ army was commanded by a magister equitum, the same Promotus who had proven himself on the Danube in 386, and a magister peditum, Timasius. The army was accompanied by the two magistri of the western empire, Arbogast and Richomer, the latter of whom had been among the survivors of Adrianople; obviously, these two western generals were not commanding western armies, which fought for the “usurper” Maximus.Footnote ⁶⁴ Overall, then, this was a traditional fourth-century structure of command. Before setting out, however, Theodosius was forced to suppress a conspiracy among the Goths in his army, some of whom had allegedly been suborned by Maximus.Footnote ⁶⁵ The Romans thus continued to rely on, and yet be suspicious of, Gothic manpower.

Theodosius spent the next three years in the west helping to establish Valentinian II on the throne. By the time that he returned to the east in 391, the situation in the Balkans had once again deteriorated. The origins of the troubles are unclear – Zosimos blames it on the Goths who had survived the purge of the army prior to the war with Maximus – and Theodosius’ returning army suffered a defeat. Once again, the emperor yielded command, this time to Promotus, and returned to Constantinople. It was in this context that Promotus died in battle and Stilicho rose to prominence.Footnote ⁶⁶

In the interval between the Gothic treaty of 382 and the civil war of 388, Theodosius had managed to negotiate a peace with the Persians, in 387. Tensions with Persia had been running high since the end of the reign of Constantine I in 337, and it is unclear what arrangements were negotiated by Victor prior to Valens’ departure for Adrianople in 377. However, we know of no major military conflicts with Persia in the decade that followed Valens’ departure, and, moreover, it appears that Theodosius stopped assigning magistri to the region after Iulius left office.Footnote ⁶⁷ The treaty of 387 was built around the partition of Armenia and would initiate over a century of almost uninterrupted peace with Persia.Footnote ⁶⁸ It is thus interesting that in the run-up to Theodosius’ final civil war (against Eugenius in 394), he created the office of magister utriusque militiae per Orientem which is first attested, along with its first known occupant Addaeus, in 393.Footnote ⁶⁹ Although not precisely the title found in the later Notitia, the post was consistently filled after Addaeus and its title quickly took on the standard form of magister militum per Orientem (MMO). There is even a law of 400 or 405 that discusses the command of the MMO as distinct.Footnote ⁷⁰

The MMO was the first of the regional commands attested by the Notitia dignitatum to come into being, and it is interesting that it did so in a region that was largely quiet by 393. This suggests that the office was formalized not out of pressing military need, but largely due to logistical realities. In 393, Theodosius was already preparing for his war with Eugenius, for which he would depart the following year. So far during his reign, he does not appear to have traveled east of Constantinople,Footnote ⁷¹ and yet, even with Persia quiescent, the region had perennial, low-level military needs, such as defense against raids. Oriens needed a ranking officer able to respond more quickly than the government in Constantinople. The wisdom of this approach was confirmed in 395/6, when Addaeus faced a Hunnic raid from his base in Edessa, though his response appears to have left much to be desired.Footnote ⁷² Unfortunately, we know nothing about the internal structure of this new office or the number and nature of the forces under its command.

Addaeus’ title reflects a broader shift in military nomenclature that occurred during the late 380s. During this period, official use of the titles magister equitum, magister peditum, and magister equitum et peditum declined and the compound title magistri utriusque militiae (MVM) and the simplified form magistri militum (MM) gradually took their place. The office of MVM is first attested in the western empire by an inscription firmly dated to 370 and was interchangeable with magister equitum et peditum.Footnote ⁷³ Senior commanders in the east are addressed as magistri utriusque militiae beginning with Hellebicus in 383.Footnote ⁷⁴ This title reflected the joint composition of the forces that these generals commanded and suggests the increasing obsolescence of the distinction between the equitum and peditum. The final eastern magister equitum attested in the Theodosian Code is Timasius in 386,Footnote ⁷⁵ after which all magisterial generals are addressed as MVM or MM. This was, however, a long process and it is possible that a pair of generals, Varanes and Arsacius, were serving as magister equitum and magister peditum as late as 409.Footnote ⁷⁶ Given the limitations of our sources, it is impossible to know to what extent this overall titular shift corresponded to changes in military practice. Magistri, especially those not serving in praesenti, had led combined forces since the time of Constantius II, so the new title may simply have been a belated acknowledgment of the fact that infantry and cavalry commands were not practically separable, especially when detached from the comitatus. The acceleration of this process under Theodosius may also have been linked to his absence from all non-civil wars during his reign, requiring his field generals to command both types of forces.

When Theodosius marched west again in 394 to defeat the new usurper Eugenius and his military backer Arbogast, he took a large part of the eastern field army with him. Unfortunately, we do not know which armies, or units, these were or how large they were. The literary sources indicate that they contained large contingents of barbarian auxiliaries (or “federates”), a type of unit that would not have been part of the regular command structure to begin with (i.e., such units would not have been listed in the Notitia, regardless of when it was drawn up).Footnote ⁷⁷ Zosimos reports that Theodosius appointed two MMs, Timasius and Stilicho, to lead his “Roman armies” (στρατόπεδα ῥωμαϊκά), while command over the auxiliaries was given to the Goth Gaïnas, the Alan Saul, and Bacurius, a former Iberian king (i.e., from Georgia). These auxiliaries were mostly Goths, and a contingent of them fought under the command of the young Gothic warrior Alaric.Footnote ⁷⁸ This heavy reliance on barbarian units in the eastern field army reflects the losses incurred at Adrianople (in 378) and the subsequent treaty with the Goths (in 382), while Theodosius’ choice to place his barbarian allies in the front ranks at the battle of the Frigidus in 394, where they suffered appalling casualties, reflects the Romans’ ongoing discomfort with this status quo and their desire to shift the balance between Roman and Gothic forces in the army.Footnote ⁷⁹

Theodosius remained in the west until his death in 395. He was the last emperor to rule both halves of the empire in their entirety, so his death marks the last point of common history between the military systems of east and west. His two generals, Timasius and Stilicho, whose specific offices are not identified in the sources, followed separate paths. Timasius returned to the east, where he fell from favor and was exiled in 396 as a result of the power struggles gripping the eastern court. Stilicho, meanwhile, remained in the west, where he pioneered a system of unified military control over the imperial administration that would characterize the politics of the western empire for the remainder of its existence.

Looking back from 395, we find that the history of the Roman military in the later fourth century was a story of continuity rather than radical innovation. The changes that followed the defeat of Adrianople did not involve any large-scale reimagining of the Roman command structure proper, but rather the coupling of the surviving Roman forces with semi-autonomous Gothic auxiliaries. There is no evidence that the Notitia system of three regional and two praesental Roman field armies was introduced by Theodosius I or his immediate successors. Although our sources are poor, whenever they do shine a light on the armies and generals of the eastern empire, we find these operating in the same mode established by the Constantinian dynasty. Thus, when Theodosius led his comitatus west in 388 and 394, his army was commanded by two magistri and, at least in 388, we know that these included one magister equitum and one magister peditum. The distinction between the comitatenses and limitanei continued to be maintained, with local frontier forces responsible for river defense in 386, just as they had been in 376. We find no conclusive evidence for the supposed proliferation of magistri that is attributed by scholars to Valens and Theodosius (see p. 20 for the passage in Zosimos).

Our survey has focused on generals rather than armies because our sources preserve much more evidence about offices than they do about the size and nature of the imperial field armies. How many field armies did the eastern empire have in the fourth century? If we define a field army by the standards of the Notitia (and the attested practice of the later fifth century) as about 20,000 men,Footnote ⁸⁰ it is unlikely that the fourth-century eastern empire had more than two such armies, matching the two magistri who commanded them. Put differently, the military structure of the Tetrarchy still remained in force. Moreover, these two armies would probably not have operated at full strength, as units from them were likely dispatched to the provinces on special assignment or placed under the command of the magistri who were occasionally posted to Illyricum or Oriens (Iovinus, Lupicinus), where they were joined by upgraded limitanei forces (sometimes called pseudo-comitatenses). Thus, if Valens marched to Adrianople with the equivalent of one and a half field armies, the ensuing battle caused the destruction of one field army equivalent (two-thirds, as Ammianus says). These losses required Theodosius to supplement the remainder (possibly still divided formally between two armies) with barbarian auxiliaries. Of all eastern emperors, Theodosius was the least likely to have been responsible for creating the three regional and two praesental armies that are documented in the Notitia, requiring 100,000 Roman soldiers in total. He simply did not have the resources, and for years after his death, as we will see, the eastern empire relied on a mix of depleted Roman and unruly barbarian forces. Nor is there any evidence that regional magistri were created for Illyricum or Thrace during this period, whether on a permanent or temporary basis, though a regular command for Oriens was established in about 390.Footnote ⁸¹ In sum, there is no way to reconcile what we find in the sources for the reign of Theodosius with the military organization described by the Notitia.

This last point is important because scholars have often placed great weight on a passing polemical comment by the later pagan historian Zosimos about the high command under Theodosius, an emperor whom he hated on largely religious grounds:

Scholars have traditionally understood this passage as proving the establishment of the regional commands listed by the Notitia. But even on the face of it, Zosimos does not claim that Theodosius appointed five generals, as in the Notitia, but “more than five.” In fact, Zosimos is being both rhetorical and polemical.

Zosimos’ digression is not so much about the command structure of the Roman military as the proliferation of magisterial offices (ἀρχαί) and the ensuing financial cost. In other words, the issue is not that there were too many generals in the field, but that there were too many generals drawing a salary. Understood in this way, Zosimos’ comment can easily be reconciled with our available evidence, but at the cost of its offering any support for the implementation of the Notitia system under Theodosius. For example, in 388 we know of at least five active magistri in the eastern empire: Timasius and Promotus, who were in praesenti under Theodosius himself (we use the term descriptively, not in a technical sense); Richomer and Arbogast, who were in praesenti for the fugitive Valentinian II; and Hellebicus, who was returning from an investigation in Antioch. It is also possible that men such as Saturninus, who is reported to have been continually influential at court throughout Theodosius’ reign, continued to enjoy the benefits associated with their former rank as magistri.Footnote ⁸³ It would not be surprising if all of these men received salaries appropriate to their rank. Thus, we can easily explain Zosimos’ comments without recourse to the eastern Notitia. Moreover, a similar situation occurred about 393, when Richomer was once again in the east, Timasius and Stilicho were assigned to the army that Theodosius would take west, and Abundantius and Addaeus were also in office, the latter in Oriens.Footnote ⁸⁴ When added to the proliferation of honorary appointments attested during this period, Zosimos’ complaint makes perfect sense and accurately reflects the shifting nomenclature of generals, but offers no insight into the formal organization of the Roman high command.Footnote ⁸⁵ Zosimos is polemically comparing the minimum number of active generals under Theodosius’ predecessors with the maximum number of titleholders under that emperor.

Yet despite the broad continuity we observe during this period, our sources preserve evidence for three major innovations in the reign of Theodosius. The least important of these is the changing nomenclature of magisterial commands, which by 395 were generally called MVM or MM rather than by the branch-specific designations found in earlier periods. It is important to note that these offices were qualitatively different from those of the same name found in the west after 395. In the west, the MVM became a supreme commander-in-chief, as demonstrated by the career of Stilicho and the sequence of generalissimos who followed him as the effective heads of state, often acting behind weak emperors. In the east, however, these commands were more numerous and under civilian control.Footnote ⁸⁶

Theodosius’ second innovation was the creation of a formal command for Oriens. Although emperors had been appointing generals to ad hoc commands in the region since at least the reign of Jovian, it was Theodosius who gave this office its first official form and, in so doing, established the template for the regional commands that are characteristic of the (much later) Notitia system. While it is inaccurate to claim Theodosius as the father of the Notitia system, it would be fair to understand him as its grandfather.

The final innovation of Theodosius’ reign was a product of necessity rather than policy: his dependence on barbarian manpower. To a large extent, this was not Theodosius’ fault. He inherited a shattered army in an empire that lacked the infrastructure necessary for the rapid, mass mobilization that had characterized the Roman military in the Republican period. The eastern empire of the fourth century was never going to recover quickly from a defeat such as Adrianople. In response to this reality, Theodosius temporized, fighting a war of attrition, seeking political settlements, and finding opportunities to co-opt and erode the military capacity of the Goths. It was a sound strategy, one that played to the structural advantages of the Roman state. And it ultimately worked, as least for the east. It also, however, established the dynamics that would define the next phase of the eastern Roman military, which was marked by under-militarization on the Roman side and by violent, ethnically inflected tensions between the Roman civilian administration and the Gothic military forces that it ostensibly commanded.