Ammianus' Res Gestae appeared sometime in the 390s a.d. and is the last work of Latin classicizing history to survive in bulk.Footnote 1 Its narrative, which originally picked up where Tacitus' Historiae left off, now runs from a.d. 353 to 378. But because so much is missing, fundamental questions remain about Ammianus' authorial practice and the structure of his narrative.Footnote 2 Indeed, the structural implications of Ammianus' extant books have worried scholars since the heroic age of Wilhelmine philology. Eighteen books survive, numbered 14 through 31 in the extremely tenuous manuscript tradition, and it is very likely that the lost books had already disappeared by the time of the sixth-century grammarian Priscian.Footnote 3 The extant books are quite long and even the most cursory of them treats only a couple of years of history. For that reason, it was and is difficult to see how thirty-one books on the model of those extant could plausibly have stretched from Nerva to Valens, as Ammianus himself tells us they did (31.16.9), unless the early books were very summary indeed.Footnote 4 There are other difficulties as well: a 31-book Res Gestae offends against every canon of classical symmetry, which ancient authors almost never ignore.Footnote 5 Finally, Book 31 is a monographic treatment of the Gothic rebellion and the Adrianople campaign of a.d. 378 that sits uncomfortably with the other books, which conclude with the events of a.d. 375.Footnote 6 Many scholars, more interested in other aspects of Ammianus' technique or in his value as a historical source, have ignored these interlocking problems altogether. Others have proposed various solutions, the most audacious of which posits composition by hexads in a manner that is detectable in the extant eighteen books.Footnote 7 Though inherently plausible, this hypothesis requires as its corollary the misnumbering of the extant books very early in their manuscript transmission and an original total of either thirty-six or twenty-four books.Footnote 8 Many will find that difficult to credit.Footnote 9 On the other hand, while one can make a case for pentadic structure joined to a stray thirty-first book, there is no evidence for the balanced decades into which such pentads should naturally have resolved.Footnote 10
In other words, none of the proposed solutions to the structural problems of the extant Res Gestae is fully satisfactory, but whether as the conclusion of a final hexad or as an obtrusive and asymmetrical conclusion to thirty prior books, Book 31 remains anomalous. As has long been recognized, it is monographic and tied only loosely to its predecessors, whether the extant books as a whole, or the five post-Julianic books that immediately precede it.Footnote 11 The full extent of the disjuncture between Book 31 and the rest of the Res Gestae will become clearer in what follows, and there have been various attempts to explain it. All start from the assumption that Book 31 was intended to form the conclusion of the Res Gestae. This means that the default analytical approach is to look for connections between it and the earlier books, however subtle, and at times to imagine them where they do not exist. If, however, one begins from the fact of disjuncture rather than the current position of Book 31 in the Res Gestae, another explanation becomes possible: separate composition. Composite construction has now allowed us to make sense of Procopius' tripartite Anekdota and has been thought to do the same for Augustine's anomalous 13-book Confessions.Footnote 12 A similar hypothesis might do the same for Ammianus' Book 31 and would explain many of the book's peculiarities. To be sure, there is good internal evidence both for and against separate composition of Book 31, evidence that warrants testing. That said, as one tests this evidence, the peculiarities of Book 31 become more rather than less striking. The present study reaches a conclusion that will explain these peculiarities: Book 31 is not a straightforward conclusion to the Res Gestae. Rather, it is an early work and perhaps Ammianus' first foray into history-writing, that is rooted in the eastern, indeed specifically Antiochene, controversies that followed in the wake of Adrianople. It was later translated from the original Greek, lightly revised, and soldered onto the end of a Res Gestae that might as easily have concluded with the events of a.d. 375. What survives is an opus imperfectum, in which the joins are still visible, making Book 31 palpably different from its predecessors. Before that argument can be developed, however, we need to consider questions of date.
I THE DATE OF THE RES GESTAE AND THE DATE OF BOOK 31
The date for the completion of the extant Res Gestae is, if not straightforward, at least not very problematic. There may or may not be external evidence in Libanius, Ep. 1063, a letter that refers to an unspecified Markellinos who made public recitations of his own work. If that means our historian Ammianus Marcellinus, then the Res Gestae had begun its publication by a.d. 391/392; if it does not — and the case against the identity of the two is very strong — then there is only internal evidence to go on.Footnote 13 All of that, however, converges on roughly the same date of a.d. 390.Footnote 14 Ammianus probably wrote, and certainly published, the bulk of the Res Gestae at Rome. He certainly lived in the city during the urban prefecture of Julianus Rusticus in a.d. 388 and had probably arrived there by a.d. 384 when severe food shortage resulted in the expulsion from the city of peregrini, among which unfortunates Ammianus should perhaps be counted.Footnote 15 If that is the case, then Julianus' prefecture gives a terminus post quem for composition. The mention of Aurelius Victor's prefecture (21.10.6) and the consulship of Neoterius (26.5.14; cos. a.d. 390, announced by late a.d. 389) push that date into a.d. 389, as does the almost certainly posthumous sketch of Petronius Probus (27.11.2).Footnote 16 Direct verbal reference to Pacatus' panegyric to Theodosius, delivered in a.d. 389, points to the same date.Footnote 17 Most of the evidence for date, then, gives a terminus post quem in the period after a.d. 388/389, though not necessarily long afterwards. The terminus ante quem is less certain, although the demolition of the Alexandrian Serapaeum in summer a.d. 391 should probably post-date the composition of the penultimate hexad (Books 20–25): the temple figures prominently in the Egyptian excursus of Book 22, but no mention is made of its destruction where one might reasonably have expected it (22.16.12). Negative evidence can never be wholly conclusive — particularly in a text as lacunose as that of Ammianus — but this is a very strong example of its kind.Footnote 18
We can thus establish the date of the entire Res Gestae fairly easily, but none of the evidence for it comes from Book 31 or bears upon the date of that book's composition.Footnote 19 Book 31 does offer some internal evidence of its own. Nothing in the historical material it covers explicitly refers to events after a.d. 378, although the general references to the decline in the emperor Gratian's morals may reflect a period closer to his death in a.d. 383 — perhaps the year of cold war between him and Theodosius in a.d. 382 — than to his early years as western emperor. A single passage, the description of the horrors of the battle at Ad Salices (31.7.16: ‘ut indicant nunc usque albentes ossibus campi’) may possibly refer to Ammianus' movements in the years after a.d. 378. Ammianus had certainly been to Thrace (27.4.2), but though the bleached bones of Ad Salices are usually thought to indicate autopsy, their description is not signalled with the first-person plural that Ammianus usually reserves for things he has personally observed.Footnote 20 That they are also a Vergilian and Tacitean allusion raises further suspicions.Footnote 21 It may therefore be doubted whether the reference belongs amongst the visa of the Res Gestae, but even if it does, it gives us only a terminus post quem of early a.d. 382, by which time Thrace and the Balkans had become readily traversable by civilians again.Footnote 22 In other words, the internal evidence of Book 31 provides at very most a terminus post quem of a.d. 382/383 for its composition. Further discussion of date must rest on Tendenzen or appeal to phantom sources rather than on hard data, so we can instead turn to the structure of the book, its relationship to the Res Gestae as a whole, and its specific argumentation, which between them constitute the evidence for separate composition.
II STRUCTURAL EVIDENCE FOR AND AGAINST SEPARATE COMPOSITION
The place to start is the evidence against separate composition. This is quite real. For one thing, and most obviously, Book 31 includes six explicit cross-references to the rest of the Res Gestae. These include the words inter haec which open the book (31.1.1); the closing envoi to the writing of history, ‘Haec ut miles quondam et Graecus, a principatu Caesaris Nervae exorsus, ad usque Valentis interitum, pro virium explicavi mensura’ (31.16.9); and four references back to events treated earlier — one on Valens and Athanaric (31.3.4), one on the guardsman Maurus who had crowned Julian (31.10.21), one on the oracle of the tripod (31.14.8), and one on the Saracens (31.16.5). None of these references is, strictly speaking, probative, but neither are they trivial obstacles and so should be taken one at a time. Inter haec (or interea, which is sometimes printed by editors) is as imprecise a transitional phrase as Ammianus' Latin permits.Footnote 23 While such phrases as ‘Haec per orbis varias partes’, ‘dum haec in diversa parte terrarum’, or ‘haec per Illyricum perque Orientem rerum series fuit’ represent normal transitions between books, the simple and vague inter haec is occasionally used to link chapters, but never the books of the Res Gestae.Footnote 24 Book 30, which some have argued is very closely connected to Book 31, has a much more typical opening sentence: ‘Inter has turbarum difficultates, quas perfidia ducis rege Quadorum excitauit per scelus, dirum in Oriente committitur facinus Papa Armeniorum rege clandestinis insidiis truncato’ (30.1.1). This carefully lays out the contrasting difficulties besetting West and East respectively, and looks forward to the next disaster. By contrast the simple, indeed vague and sloppy, inter haec is only ever found at such internal transitions as 17.4.1 and 19.51. Its otherwise unparalled use at the start of a book may plausibly argue against Book 31's having been an integral part of the original composition.Footnote 25 Our second explicit cross-reference, the sphragis, has recently been the subject of an elegant exegesis by Gavin Kelly and requires little further comment. One may note that the alteration of the single word Valentis to Valentiniani would allow 31.16.9 to stand as well at the end of Book 30 as it now does at the end of Book 31.Footnote 26 That is not to suggest that either Ammianus himself did, or that we should, make such an emendation, but rather that a couple of displaced letters would change decisively the relation of the sphragis to the book it concludes — which in turn demonstrates its lack of integration specifically into Book 31.
By comparison with the opening Inter haec and the sphragis, the cross-references to points of historical content are intrinsically more likely to be substantive. Three of the four that appear in Book 31 are intrusive and in no way integral to the book's narrative content. Thus the reference back to the earlier wars of Valens against the Tervingian iudex Athanaric comes in the narrative of the Hunnic impact on the various Gothic peoples (31.3.4).Footnote 27 It is entirely parenthetical (punctuated accordingly in both Clark and the Loeb text) and does not bear on the meaning of the material that surrounds it; if one deleted the parenthetical comment, Athanaric would simply be another new name introduced into the list of Gothic leaders who needed to face the onslaught of the Huns. The explicit cross-reference to Maurus is similarly perfunctory (31.10.21).Footnote 28 He has a key rôle in Book 31 as the unworthy successor to Frigeridus, a commander of whom Ammianus approved. Maurus is subjected to Ammianus' full range of adjectival abuse, and his crowning of Julian with his torque is cited as evidence of his rashness and arrogance. But as in the case of the Athanaric reference, all that the reader might need to understand Maurus' character and to make sense of the passage remains in place if the phrase ‘prateritorum textu retullimus’ is deleted. A third explicit cross-reference concerns the troop of Saracens (31.16.5) that repelled Fritigern's Goths from Constantinople.Footnote 29 This conforms to the pattern of the two we have just considered: the characterization of the Saracens as crafty raiders draws the ethnographic contrast with the more warlike Goths needed in the immediate context and would lose nothing of its meaning in the absence of the cross-reference.Footnote 30 The fourth historical cross-reference is the most significant. It concerns an ex eventu prophecy of Valens' death (31.14.8) and looks back to the oracle of the tripod that is treated at 29.1.6–14 and 29.1.28–35 in the very long account of those persecuted by Valens for the crime of maiestas.Footnote 31 It comes immediately after the necrology of Valens, which ends with the first sentence of 31.14.8: ‘Haec super Valente dixisse sufficiet, etc.’. The rest of the section is an elaborate explanation of how a verse prophecy in three lines, foretelling Valens' death when Ares raged on the plains of Mimas, came true. Ammianus explains how Valens for a long time feared Asia, where he had heard from scholars (‘doctis referentibus audiebat’) that Homer and Cicero located a mountain by the name of Mimas.Footnote 32 But, as is the way with prophecies, this one came true unexpectedly: after Valens was dead, there was discovered on the field of Adrianople a gravestone commemorating ‘an ancient nobleman named Mimas who was buried there’. Ammianus puts this into a dicitur construction, with which he regularly indicates his disavowal of an opinion, and then intensifies this expression of doubt by using ‘cecidisse existimatus est’ of the place where Valens ‘was believed to have fallen’ and the inscription ‘is said to have been found’. None of this elaborate explanation is present, or even alluded to in 29.1, save for one key point: the three verses to which Ammianus refers in Book 31 are quoted in full at 29.1.33, while only the last line is quoted again in 31.14.8.Footnote 33 Without the first of the three lines, however, the prophecy of vengeance against Valens makes no sense at all. Here, then, we have a genuinely substantive cross-reference from Book 31 to material treated in an earlier book and the best piece of evidence for continuous composition that exists. That said, the story of the prophecy is also a supplement and signalled as such (‘illud autem praeteriri non conuenit’), coming at the end of a necrology and before a new chapter, introduces a new phase of the narrative.Footnote 34 And, as was the case with the other three specific cross-references considered above, if one were to omit everything after the first line of 31.14.8 and pick up again with 31.15.1, no meaning internal to the narrative of Book 31 would be lost. We would go from the end of the necrology of Valens, of whom Ammianus tells us he has said enough (‘dixisse sufficiet’), to the narrative of the battle's aftermath (‘Post exitialem pugnam’). That the rest of 31.14.8, along with 31.14.9, is indeed a reference back to 29.1 is absolutely certain; that it demonstrates continuous composition with the other books of the final hexad is by no means so clear.
It is, after all, the only such integral cross-reference in Book 31 and it can, like all the others, be read as an intrusive supplement. That possibility is made more likely because of the care with which Ammianus builds cross-references into his narrative elsewhere in the history, a care that is particularly pronounced in the closely linked Books 26–30. Whereas the numerous, specific cross-references in Books 26–30 are almost entirely to events within those same books, only the verse about the plains of Mimas links Book 31 back to the rest of the hexad.Footnote 35 Similarly, the constant, tense comparisons of Valentinian and Valens with Julian, so marked a feature of Books 26–30, are missing from Book 31. That is, although the whole battle of Adrianople can be read as a disastrous parallel to Julian's victory at Strasbourg, the comparison must be inferred, whereas in the previous five books of the last hexad, the comparisons are frequent and explicit.Footnote 36 Thus, although the cross-references in Book 31 might argue in favour of continuous composition between it and earlier books, they might equally be explained as the type of relatively easy revision which could serve to superficially link together two existing texts that were not previously related to one another.Footnote 37
Other parallels with, or echoes of, earlier books are sometimes cited. These are of varying plausibility, but all start from the assumption that Book 31 was composed along with the rest of the larger work. That assumption makes it perfectly natural to look for parallels, however slight, without fully taking account of how anomalously monographic Book 31 is. Of various larger parallels that have been suggested, two merit particular attention. First, there are the two excursus, on the Huns and on the Alans, with which Book 31 begins (31.2–3). There are, of course, excursus throughout the Res Gestae, and so when one finds two near the beginning of Book 31, the question must arise of whether they might not help to demonstrate its continuous composition with the other books. In the larger work, Ammianus' excursus are quite various in theme, although both geographical and ethnographic excursus are common and cross-references demonstrate that both types also existed in the lost books.Footnote 38 Overall, the geographical excursus are far more numerous and comprehensive than the ethnographic, but what is striking about the ethnographic excursus in Book 31 is how much they differ from those in the rest of the work. Although, as they must, they catalogue the lands beyond the imperial frontiers in which the Huns and Alans dwell, they do not follow Ammianus' general pattern of geographical and ethnographic excursus, which are rigidly stereotypical elsewhere in the Res Gestae — e.g. those on Gaul (15.9–12), Thrace (27.4) or Egypt (22.15–16).Footnote 39 This is particularly noticeable when one compares the excursus on the Huns with that on the Saracens in 14.4.1–7. The Hunnic excursus, it has been suggested, is meant to deliberately balance the earlier one on the Saracens, but that is hard to demonstrate if only because the Saracenic excursus of Book 14 is probably a short reminiscence of an earlier, longer one, which Ammianus had placed in his now-lost account of Marcus Aurelius (14.4.1; 31.16.5).Footnote 40 In Book 14, the ethnographic excursus on the Saracens balances the geographical excursus surveying the eastern provinces in 14.8, but there is no such internal balance in Book 31. Rather, the Hunnic and Alanic excursus are something of a cross between Ammianus' ethnographic and geographical styles. What is more, the two ethnographic excursus in Book 31 are less thoroughly worked up than those of earlier books, which draw on various genres and on many official and Latin sources, not merely Greek ones. The Hunnic and Alanic excursus, by contrast, are modelled very heavily on Herodotus, to the point that where Ammianus had little meaningful knowledge, he introduces Nervi and Vidini from the Neuroi and Boudinoi of Herodotus.Footnote 41 That is to say, Ammianus' fondness for the formal excursus is clearly visible in Book 31, as it is throughout the Res Gestae, but the excursus of that book are different enough from those in earlier books, both in terms of typology and sources, that they cannot be used to establish continuous composition with preceding books.
A second parallel between Book 31 and the other books of the Res Gestae is also worth longer consideration, and that is the question of whether the account of Adrianople is meant to balance, parallel or comment on that of Strasbourg in Book 16 or of Amida in Book 19. It is hard to find a typology of battle in Ammianus along the stereotyped lines of the geographical excursus, but that the three major set-piece battle descriptions in the extant Res Gestae might provide intertextual insight into one another is not unlikely.Footnote 42 The comparison of Amida and Adrianople is hardly sustainable, for they are fundamentally differently conceived, one as a personal narration, the other as a set-piece battle such as one finds elsewhere in the ancient historiographical tradition, and there are no real linguistic echoes in the two accounts.Footnote 43 By contrast, a strong case can be made that the rhythm of the account of Adrianople in Book 31 is meant to provide a counterpoint to that of Strasbourg, with Julian's caution and Valens' recklessness sharply contrasted.Footnote 44 Julian allowed his soldiers to rest rather than keeping them out in the heat of the day, so they could enter battle fresh; Valens notoriously weakened his forces by having them march in full armour at noonday and then wait, growing hungry and thirsty, before battle was joined. The speech that Ammianus puts in Julian's mouth before Strasbourg (16.12.9–12) is a précis of how the disaster at Adrianople might have been avoided, the lesson being that ignorance of the past leads to disaster, while a knowledge of the past, such as Julian's (16.5.16), leads to success.Footnote 45 These parallels are significant enough that Ammianus almost certainly manipulated one or another of the two accounts to point exactly this contrast. That said, however, there is no way to determine the priority of the paired accounts. The exact same contrasted patterning of the battle accounts would be visible whichever description was written first. In other words, the whole of the early Julianic books, with their climax and entire point the battle of Strasbourg, might be a retrospective construction to balance an account of Adrianople and explain the later disaster. Other forms of thematic unity between Book 31 and its predecessors that are sometimes cited generally identify large things like incertitude, horror, pessimism, and so on.Footnote 46 These depend so much on the subjective selection of evidence by the individual reader that they cannot really be probative.
Similarities of vocabulary and habits of composition between Book 31 and the rest of the Res Gestae might also seem significant, but they do not carry great intrinsic weight. As noted, there are excursus in Book 31, as there are elsewhere in the Res Gestae. There is a necrology of Valens, as elsewhere there are necrologies of other emperors, although Valens' is shorter, more cursory, and lacks the historical exempla that decorate the others. Exempla are present in Book 31 as elsewhere, but they are entirely analytical rather than epideictic. Ammianus obtrudes his own voice into the narration of Book 31 as he does elsewhere. And similarities of expression can be paralleled throughout the Res Gestae as a whole.Footnote 47 But that is only to say that Ammianus was a fairly consistent writer of Latin and that Book 31 was indeed written by him, which is not in doubt. If Book 31 were strikingly different in vocabulary or idiolect from other books it would certainly help indicate a separate composition. The absence of such differences, however, cannot demonstrate continuous composition, as Ammianus might first have developed compositional patterns while writing Book 31 that were later replicated in other books of the history. As we shall see in the next section, the contents and argument of Book 31 represent a thoroughly eastern perspective and imply its original composition in Greek. The similarity of the Latin style in Book 31 and other books of Res Gestae therefore ceases to be at all surprising, for Ammianus will have translated his final book from a Greek original after his mature Latin style had developed during the composition of the work as a whole.
In the end, it remains true that evidence against the separate composition of Book 31 is quite real — real but not decisive. Enough doubt subsists to justify looking beyond the evidence of cross-references and compositional habit for positive traces of separate composition. As the next section will argue, the polemic that we find in Book 31 strongly suggests an originally separate composition, in Greek, and a later translation into Latin for inclusion in the Res Gestae. Before turning to that argument, however, we need to understand just how many differences of structure and content there are between Book 31 and the others. That a major structural break in the Res Gestae lies between Books 25 and 26 has never been in doubt, as Ammianus takes up the post-Julianic narrative at a new, faster pace and systematically treats East and West as separate narrative continuums. The Valentinianic books, from 26 to 30, form a tightly-knit compositional unit.Footnote 48 Several signs suggest that Book 31 was not composed continuously with the five that precede it. First, Book 31 lacks all the references to haste and the apologies for cursoriness of description that form a constant leitmotiv in the Valentinianic books.Footnote 49 When Ammianus addresses his readers at 31.9.10, it is to explain why details of the numbers of dead could not be discovered, rather than why his narrative races past the details. More significantly, the attempt to maintain an annalistic structure while simultaneously narrating multi-year blocks of events in different partes imperii is abandoned in Book 31.Footnote 50 Likewise, while the forward and backward cross-referencing within Books 26–30 is frequent and closely connected, those five books contain no precise references to the contents of Book 31 and only two references to the events that it treats, both of them to Adrianople.Footnote 51 One of these two references is wholly generic.Footnote 52 The other suggests a very different account of that battle — more cursory and in the manner of the campaign narratives of Books 26–30 — than the monographic account which actually appears in Book 31.Footnote 53
In the same way, the contents of Book 31 show no special knowledge of the immediately preceding books. For instance, 31.1.3 refers to the ghost of the Armenian King Pap and the victims pursued after Theodorus' conspiracy as victims of Valens. Theodorus' supposed conspiracy is treated in Book 29 and the Armenian narrative occupies very substantial parts of Book 30, but Book 31 makes no real reference to those accounts.Footnote 54 More troublingly, several of the military figures who fought in the Adrianople campaign are described in Book 31 as if introduced for the first time, even when they have appeared in significant rôles earlier. Thus Barzimeres, a minor figure and tribune of the Scutarii, was mentioned at 30.1.11, in quite uncomplimentary terms. But he is introduced at 31.8.9 merely as ‘eruditum pulvere militari rectorem’, without mention of his implication in the Armenian affairs of Book 30, and praised in a way that might suggest that Ammianus did not yet know the rôle Barzimeres played in the Pap affair when he wrote Book 31.Footnote 55 When Traianus is introduced at 31.7.1, we are reminded neither of his post as comes rei militaris in the East nor of his implication in Armenian intrigue (29.1.2; 30.1.18–21).Footnote 56 Book 31 also misses any number of chances to cross-reference events in Books 14–25 of the Res Gestae, but that might reflect no more than distance from the time of composition.Footnote 57 However, three such instances are significant. First, the barritus is originally described in the extant Res Gestae at 16.12.43; when it reappears at 26.7.17 the reader is assumed to know what it is, presumably because he has been told in Book 16. In Book 31, however, the barritus is described at the battle of Ad Salices as if for the first time (31.7.11).Footnote 58 Second, in the enumeration of the noble dead of Adrianople, Potentius, son of Ursicinus, is praised, but his father is merely magister quondam (31.13.18). As was demonstrated long ago, not only did Ammianus at some point choose to make Ursicinus the hero of his pre-Julianic books, but his loyalty to his former general was such that he systematically exaggerated his qualities as a commander and suppressed the mutual contempt between his two heroes, Ursicinus and Julian.Footnote 59 In that light, a mere magister quondam is peculiar. Third, there is the way Ammianus handles the shift from the eastern narrative to the discussion of the western preoccupations that prevented Gratian's immediate intervention in Thrace. That transition introduces western affairs not as a return to an ongoing topic, but as something altogether new: ‘The madness of the times, as if the Furies were stirring up the whole world, snaked likewise into distant regions, spreading widely’ (31.10.1).Footnote 60 This is significant, for Gaul has been a persistent source of violence throughout the Res Gestae, not least in Books 26–30. In Book 31 it is described as if only just being introduced as the scene of turmoil incited by the furies. Whereas in Books 26–30, eastern and western affairs have alternated, here the West makes a very brief appearance solely because Gratian's delay in responding to Balkan events would otherwise be inexplicable.
If some of the foregoing might be explained by the hurried or unfinished revision of an early draft, there are also more serious discrepancies of tone between Book 31 and those that come before, for instance its portraits of Gratian and Valens, which differ from those in the preceding five books. The praise for Gratian at 31.10.18 is fuller than the descriptions which introduce him in earlier books. Even if Book 31's praise does then move on to a damning comparison with Commodus, Ammianus' assessment of Gratian in earlier books is chillier throughout; it may suggest that Ammianus, as he wrote his larger history, came to see the whole of Gratian's reign in a dark light, where once he had seen the decline of an initially strong character. That same type of discrepant portrayal is much more visible in the figure of Valens, whose necrology in Book 31 points up virtues that are entirely absent from Books 26–30, where the emperor is granted not one single redeeming characteristic. In the necrology of Book 31, Valens appears as fundamentally just, for all that he was rude and uncultured, too gullible, and too ready to listen to bad advice. His merits, we are told, are ‘cognita multis’. They include loyalty, justice to the provincials, leniency in respect of taxes, and being the scourge of wicked officials. Yet one is hard pressed to find so much as a glimmer of these virtues in the narrative account of Valens' reign.Footnote 61 The Valens of Books 26–30 is irredeemable, that of 31 merely another failed emperor. On their own, not one of these differences in tone and characterization demands a separate composition, any more than the cross-references in Book 31 to earlier parts of the Res Gestae rule it out. On the basis of structure and content, the argument must remain in suspension. However, the polemic of Book 31, to which we can now turn, not only explains the various differences at which we have been looking, but also makes a strong case for an originally separate composition.
III THE ARGUMENT OF BOOK 31
Adrianople was a disaster by any standard and created an immediate need for explanation. Regardless of when he wrote Book 31, Ammianus was part of the contemporary debate over the meaning of Adrianople, the sources for which have been comprehensively examined.Footnote 62 If one accepts that Book 31 was written at more or less the same time as the other books of the Res Gestae, then Ammianus becomes a late entrant into the debate, one who offered a synthetic response to the many different paths the argument had already taken.Footnote 63 Yet all the opinions which Ammianus seeks to attack or correct had been voiced very early, in the first couple of years after the battle. More to the point, they had been voiced by eastern authors, Antiochene and Constantinopolitan, from precisely the world in which Ammianus moved.Footnote 64 The Latin, western approaches of Ambrose, Jerome, Pacatus and Rufinus are not held up to challenge in Book 31 in the way that eastern arguments are.Footnote 65 That might be unsurprising in the case of Christian authors, who are on the whole ignored by Ammianus, but we know for certain that he read and used Pacatus, so the omission is significant.Footnote 66 Taken as a whole, the narrative thrust of Book 31 counters, first, the Antiochene interpretation of Libanius and, second, the shifting and tendentious palatine voice of Themistius, each of whom was long used to countering the other's arguments.Footnote 67 The polemic of Book 31 seems to reflect that Antiochene background, in a place where Libanius was as busy explaining Adrianople to his pagan contemporaries as Chrysostom was to Christians.
We cannot be certain of Ammianus' movements after he left the army, but he was certainly in Antioch in the earlier 370s and need not have left there until a.d. 383 or so.Footnote 68 Earlier studies have shown the extent to which Ammianus was embedded in an Antiochene context. He gives disproportionate space to historical figures with a connection to Antioch and quite a few of these same men also appear in the works of Libanius. Nevertheless, because both authors shared a similar urban background, one gains little from simply comparing what each one says about this or that person.Footnote 69 Each might independently reflect rumours or opinions current at Antioch, and each knew enough about his city to make up his own mind without reference to the other.Footnote 70 Despite these reservations, it has proved possible to demonstrate that Ammianus was very definitely in dialogue with his older contemporary, so much so that the occasional verbal echo survives across the linguistic divide between Libanius' Greek and Ammianus' Latin.Footnote 71 What is more, Ammianus' use of Libanius can be demonstrated textually, without assuming their relationship a priori on the basis of Libanius, Ep. 1063 to Markellinos.Footnote 72 For instance, Ammianus' overall portrait of Julian as a philosophical soldier-emperor repeatedly counters Libanius' sophistic Julian. The way Ammianus develops his portrait turns that of Libanius on its head, and a pointed comparison of Julian with Epaminondas echoes Libanius' language to make a diametrically opposing point.Footnote 73 Something else follows from this close reading of Libanius by Ammianus: Libanius' eighteenth oration, the Epitaphios for Julian, was not delivered publicly. It was, rather, a long and embittered pamphlet meant for private circulation, although it eventually went on to rank with his most popular works. We cannot know whether Ammianus was close enough to Libanius to have seen the Epitaphios in the immediate aftermath of Julian's death, but we can be certain that his Antiochene network was well-enough connected to make this private masterpiece accessible to him. He knew other works of Libanius, too, for instance the twelfth oration, on Julian's consulate, and the fourteenth, on the Egyptian Aristophanes. Echoes of this latter oration make their way into Ammianus' famous diatribe on the Egyptian character, and it is not impossible that his other famous diatribe, against lawyers, also refers to Libanius.Footnote 74 In other words, Ammianus was an attentive reader of Antioch's star intellectual, and we need not hesitate to look for a dialogue with Libanius over Adrianople.
As Libanius himself tells us, as soon as he got news of the battle, he ‘contemplated the causes of the disaster and the day afterwards explained them to others’ (Or. 2.53). We do not know exactly what he said at the time, but a year later, in his widely-diffused twenty-fourth oration, Libanius argued that the death of Valens had been divine punishment for the failure to avenge the death of Julian.Footnote 75 The motif of divine punishment is present in Christian writers of the time as well, Ambrose in the West to be sure, but also the Antiochene John Chrysostom.Footnote 76 Like Libanius and John, Ammianus was a religious man. Beneath the seemingly reasonable surface of the Res Gestae, there lies a quantity of anti-Christian polemic, and some have tried to uncover a coherent theology in Ammianus' multifarious references to the supernatural.Footnote 77 Throughout his history, Ammianus is perfectly capable of attributing events to direct divine action, or to human incapacity to act usefully in the face of maleficent fortuna.Footnote 78 In Books 29 and 30, it is Valens' fated death in Thrace that preserves him, inevitably, through the many other possible ends that constantly threaten.Footnote 79 Yet in Book 31, Ammianus sticks to the merest intimations of Valens' fated destruction, instead expatiating on the catalogue of human error which led inexorably to the disaster at Adrianople.Footnote 80 Flatterers, not fate, convinced Valens to accept the Gothic petition (31.4.4). Though it might have been a malign god that impelled Valens to appoint ‘men of ill repute’ to manage the Danube crossing, it was the fault of Maximus and Lupicinus that it went so badly awry.Footnote 81 The Goths Sueridus and Colias, themselves ‘lacking in arrogance (sine tumore)’, are provoked by the ill-considered action of a local magistrate (31.6.1–3). Rather than fight a guerrilla war, the generals Profuturus and Traianus ‘incautiously (intempestive)’ choose to fight a pitched battle (31.6.7). At Adrianople, Valens, ‘struck by a sort of rash ardour, hurried into action’, human error, not divine necessity.Footnote 82 It was the headlong (immature) charge of Bacurius and Cassius that scuppered the last minute negotiations that might have averted disaster (31.12.16–17). And it is surely significant that, in describing the aftermath of Ad Salices, Ammianus decries Fortuna not in his own narrator's voice, but through an imaginary ingenuus taken prisoner by the Goths.Footnote 83
In a similar vein, while Libanius (Or. 24.3 and 5) had explicitly rejected any attempt to blame the soldiers or their generals, Ammianus is at pains to show the failures of both.Footnote 84 The clearest statement comes at 31.4.6: ‘And so with the stormy zeal of eager men the ruin of the Roman world was led in.’Footnote 85 Human rashness — his own, his generals', his soldiers' — and not divine punishment destroyed Valens. This overwhelming focus on the human factor is atypical of the Res Gestae as a whole, which generally places human decisions within the context of divine activity, whether the divine is described as fortuna or numen or deus. It is hard to see that reflecting anything other than an immediate riposte to arguments about Valens current at Antioch around a.d. 379 and 380, particularly in the circle of Libanius, for though Christian authors like Chrysostom also saw divine vengeance at work, it is most unlikely that Ammianus paid them any heed.Footnote 86 Be that as it may, the consistent refusal to blame Adrianople on fate forms another real point of distinction between Book 31 and those that precede it.
Similarly, the unusual treatment of Valens in Book 31 makes sense in light of the immediate Theodosian propaganda response to Adrianople.Footnote 87 As we have noted, Valens is given far more credit in his necrology, which stresses in particular his capacity for justice, than anywhere in the five other books of the Res Gestae that treat his reign. Nicene Christians, of course, were swift to condemn Valens' injustice as an Arian and a persecutor of the orthodox, but Ammianus is not interested in that perspective at all. It is the fact of Valens' just treatment of his subjects under the law which exercises the historian, and it seems likely that Theodosian portrayals of Valens hold the key to Ammianus' treatment of him here. Like the Antiochene writers, the court orator Themistius affected to believe that Adrianople was condign punishment of Valens. In Oration 14, of early spring a.d. 379, Themistius is sanguine; he does not dwell on the battlefield disaster, choosing instead to focus on the damage Theodosius will soon do to the Goths and the dawning of the new age that has given everyone a renewed strength.Footnote 88 The next of Themistius' speeches, Oration 15, was delivered in January a.d. 381, almost certainly on Theodosius' dies imperii, the 19th.Footnote 89 Its tone is dramatically different to that of its predecessor, both in the scale of disaster which he attributes to Adrianople and in the attitude he takes towards Valens, who is now condemned specifically for his injustices.Footnote 90 This charge represented a new phase in Themistius' Theodosian propaganda, marking the shift to a new period in military fortunes, when swift victory in the Gothic war had been recognized as impossible, so that the emperor's other virtues required more emphasis.Footnote 91 If the justice of Theodosius was to be emphasized, so Valens' injustice had to be deplored, and Themistius continued to harp on this theme thereafter, as in his Oration 16 of 1 January a.d. 383.Footnote 92 When Ammianus makes special and extended reference to Valens' capacity for justice, despite all his other faults, we may posit a response to the new twist in the official story at court.
If these echoes of the contemporary blame game seem to situate Book 31 in the immediate aftermath of the battle, there are also signs that the Ammianus of Book 31 is responding to, and rejecting the logic of, Theodosius' a.d. 382 treaty with the Goths.Footnote 93 In the first place, it seems as if Ammianus is rejecting the excuses for the treaty which were supplied at the time by the orations of Themistius. Themistius' Orations 15 and 16 in particular are concerned to explain why it really was a good thing that the Goths had not been expunged from the face of the earth, a theme to which the orator returned in his 34th oration of a.d. 384 or 385.Footnote 94 Those same Orations 15 and 16 are also the first major public speeches known to us which emphasize the significance of Adrianople as an utter and complete disaster, in fact, the ‘Iliad of disasters on the Danube’.Footnote 95 Ammianus wholeheartedly rejects this notion, admitting that Adrianople might have been as bad as Cannae (31.13.19), but that it was a reversible disaster, from which the Empire would recover as it had recovered in the past. He cites historical precedents to make his case (31.5.11–17): the Cimbric invasions destroyed by Marius, the Marcomannic destroyed by Marcus. Now, there are other parts of the Res Gestae in which Ammianus sets recent events against possible historical precedents, as for instance in Jovian's surrender of Nisibis (25.9.8). Nowhere else, however, is Ammianus arguing against the opinions of others: ‘Those who are ignorant of past times say that the state was never before blanketed with such dark evils, but they are led astray by the horror of the recent evils which have overwhelmed them.’Footnote 96 Elsewhere, Ammianus sets up historical comparisons which he himself has invented as an exercise. In the discussion of Adrianople, he is in dialogue with another tradition, to which he is hostile.
One could doubt whether Themistius was the immediate target of Ammianus' argument, save for an interesting collocation. Ammianus unobtrusively prefaces his list of historical precedents for Adrianople with one of his most programmatic and self-reflexive statements about writing. Readers ought not to demand of him a precise numbering of those slain at Adrianople, for there was no way of finding that out.Footnote 97 Instead, ‘it will suffice to describe the high points of what happened, without hiding the truth through any lie, because faithful honesty is always necessary in explaining past events’.Footnote 98 He will not lie (‘ueritate nullo uelata mendacio’) and, by implication, those who suggest that Adrianople was the worst of all Rome's disasters — those like the Themistius of a.d. 380 and 381, in other words — are liars. That Themistius is here Ammianus' main target is further suggested by what is probably an oblique reference to him as one of the main forces pushing Valens into his all too hasty action at Adrianople: the eruditi adulatores who, at 31.4.4, urge Valens to fight before the arrival of Gratian may well be singular not plural, may in fact be the singularly sycophantic Themistius — it would not be the only place in which Ammianus uses a plural when only one specific person is in question.Footnote 99 It was Themistius, after all, who had talked up the advantages of letting the Goths into the Empire in a.d. 376 and 377, and who in his second oration to Theodosius had felt it necessary to disclaim any intention to flatter.Footnote 100 Themistius lived until c. a.d. 389/390. By the time the last books of the Res Gestae were published he was dead and it would have been unnecessary to refer to him periphrastically. But if Ammianus wrote Book 31 earlier, while the great orator was still alive and making speeches with which Ammianus disagreed, there was reason to be circumspect, for Themistius was part of the érudit world that linked Antioch, Constantinople and the imperial court, within which network Ammianus might hope to find an audience.Footnote 101
The significance of the Themistian connection is actually greater still, for the historical exempla which seem so openly to debunk Themistius are also relevant to the criticism of the Gothic treaty, and linked explicitly to it through Ammianus' choice of words. The massacre of Gothic soldiers in Asia Minor on the orders of the magister militum per Orientem Julius is a source of scholarly controversy because Ammianus' account seems to be contradicted by Zosimus, who was here drawing on Eunapius.Footnote 102 Many scholars simply reject Zosimus and accept Ammianus, but it has also been argued that Ammianus deliberately transposed to the aftermath of Adrianople a decision made only in the following year.Footnote 103 It may be, however, that Ammianus focuses on one moment in what was actually a fairly widespread pogrom against Goths in the eastern provinces, both soldiers and young males more generally, an interpretation suggested by two sermons of Gregory of Nyssa.Footnote 104 There may, in fact, be no contradiction. Ammianus' most recent editor prints an irregular clausulation at 31.16.8: ‘His diebus efficacia Iulii/ magistri militiae trans Taurum/ enituit salutaris et uelox.’ But if, instead, one reads a regular clausula here — ‘his diebus efficacia Iulii/ magistri militiae/ trans Taurum enituit/ salutaris et uelox’ — then it is the massacre which takes place trans Taurum, in Asia Minor, rather than Julius who is serving trans Taurum.Footnote 105 This second reading is to be preferred and, if it is, then the Antiochene perspective of the writer is clearer still: Asia Minor was indeed trans Taurum for someone resident in Antioch.
Regardless, the figure of Julius himself merits a moment's attention, for as magister militum per Orientem until late a.d. 378 or 379, he was also well-regarded by the Antiochene circle of Libanius.Footnote 106 He is explicitly attested as attending upon Libanius when the latter was feeling unwell.Footnote 107 More interestingly, he is probably to be identified with the anonymous strategos who reprimanded and obtained the dismissal of the Christian consularis Syriae (possibly named Protasius) who was an enemy of Libanius.Footnote 108 Julius, in other words, attached himself to, if he did not actually belong to, the Antiochene pagan circles of Libanius. Shortly after a.d. 378, however, Julius was superseded by Sapores, an orthodox Nicene.Footnote 109 That Ammianus singles out Julius and no other for such high praise at the very end of Book 31 demands comment on several fronts.Footnote 110 First of all, turning Julius into the prime mover of the Gothic massacre may or may not reflect the historical truth; certainly Zosimus shows that the massacre covered a very wide area and it may be that we have in Ammianus a finely-turned compliment, giving credit to Julius for a plan of action that was in fact widely contemplated and broadly instigated. And it is a highly pointed compliment too, for to praise Julius' massacre is to reprove Theodosius' treaty of a.d. 382.Footnote 111 This, rather than the peace which Theodosius concluded, was the way Goths should be treated: the massacre was salutaris et velox, whereas, by implication, the Theodosian treaty was dangerous and had come slowly, nearly four years after Adrianople. Secondly, there is the way Ammianus treats Julius. Whether or not Theodosius came to power as a convinced Nicene Christian, he had certainly become a vocal supporter of orthodoxy by a.d. 382. Ammianus affected to know little of Christians, and certainly had no interest in their internecine struggles except as a stick with which to beat them, but in Book 31 he twice engages in his characteristic, quiet Christian-baiting, implying that Christians are treacherous supporters of the Goths. To praise a pagan magister militum — or at least a magister militum who had favoured Antiochene pagans over Christians and was replaced by a Christian — was a further implicit attack on Theodosius, and a further link between Christianity, treachery and Roman weakness.Footnote 112 That is important for, thirdly, Ammianus clearly signals his contempt for Theodosius' reconciliation with the Goths by means of verbal echoes.
We have already looked once at the passage earlier in Book 31 wherein Ammianus rejected Themistian arguments for the unprecedented nature of the disaster at Adrianople — arguments which Themistius began to make when swift victory in the Gothic war was not forthcoming and public opinion had to be prepared for the possibility of a messy compromise peace along the lines of that which emerged in a.d. 382. In his historical argument, Ammianus did more than provide exempla that proved Themistius and those like him wrong. He also set out the virtues which allowed the great generals of the past to revive Roman fortunes after defeats far worse than Adrianople.Footnote 113 The ‘great generals’ (‘duces amplissimi’) who fought the Teutones and Cimbri triumphed thanks to ‘warlike might combined with prudent discipline’ (‘potestas Martia adhibita prudentia’); Julius, according to Ammianus, acted ‘with prudent counsel’ (‘consilio prudenti’) — prudentia being a word Ammianus favours to signal approbation.Footnote 114 The Cimbri and Teutones learned, ‘to their uttermost peril’ (‘discriminibus supremis’), what Roman potestas could inflict on them; Julius saves the oriental provinces from ‘great perils’ (‘discriminibus magnis’).Footnote 115 The other exempla point the same lesson. Marcus all but destroys the Marcomanni; the seaborne invaders of the mid-third century retreat after most of them are killed; the Decii fall in battle but rout their enemies; Claudius, the glorious commander (‘ductor gloriosus’), routs the Goths and his work is carried on by Aurelian, the severest avenger of crimes (‘severissimus noxarum ultor’). Verbal echoes recall the historical precedents for Julius' actions and then hold them up to comparison. Barbarian invaders should, like the Cimbri, be rooted out entirely (‘radicitus extirpati’). Theodosius, the Christian, had done just the opposite, and his spokesman Themistius rejected precisely those historical lessons which proved how wrong-headed Theodosian policy was. On the contrary, for Themistius, if the Goths ‘have not been altogether wiped out, no murmur should be raised, for reason and philanthropy's triumphs are not to destroy but rather to improve those who have caused suffering’.Footnote 116 Even Libanius might argue that the troops of his own day were no less brave than those of prior ages.Footnote 117 Ammianus shows that this cannot be true: if it were, the Goths would not still exist, and there would be no enemy left with whom to conclude a shameful peace. This particular argument against Theodosius is likely to be contemporary and uninfluenced by later developments in Ammianus' thought and reading. We know that Ammianus did, in time, come to know and use Pacatus' panegyric to Theodosius, for there is a direct verbal echo of it in the Res Gestae.Footnote 118 In Book 16, Ammianus uses the argument and shape of Pacatus' panegyric to comment on Theodosius, subtly comparing his then-recent Roman adventus after the victory over Maximus with Constantius II's visit earlier in the century — thereby suggesting that Theodosius, like Constantius before him, was a poor general, with no victories of his own to his credit.Footnote 119 This is a classically Ammianean technique, using a work in praise of Theodosius against him. But there are no echoes of Pacatus in Book 31, even though the final chapters are, as we have just seen, a substantial critique of Theodosius and his policies. That fact would be surprising if Book 31 was written in the West around a.d. 390, but much less so if its original context is early and eastern.
Taken together with the structural discrepancies discussed above, the polemical arguments of Book 31 are the key to explaining its strange position within the Res Gestae. Book 31 manifests enough structural differences from the rest of the Res Gestae that it sits uneasily within either a pentadic 31-book history, or a hexadic 36-book one. But if it was, in origin, a monographic response to Adrianople, later lightly revised for inclusion at the end of the larger work, these discrepancies are explained. The polemic of Book 31 certainly supports a monographic origin in the immediate wake of the battle. Ammianus' focus in the whole of the book is eastern, as is his perspective. More than eastern, in fact, his perspective is Antiochene, reflecting both the traffic in ideas and explanations in that city, and also the special politics that obtained there. Throughout, the arguments Ammianus engages are those of the Antiochene literati and their connections at the eastern court. The specific arguments, the praise of Julius, the rejection of both Themistius and Libanius, all this was very old news by the start of the a.d. 390s; at the same time, Book 31 shows no awareness of the western, Latin, debate over Adrianople and over Theodosius himself that had developed in the decade between the Gothic peace and the publication of the Res Gestae, although Ammianus knew and elsewhere used one of its key texts, that of Pacatus. Book 31, then, might best be understood as a volley in the intellectual pamphlet war of a.d. 379–382, one that was launched from an Antiochene base shortly before Ammianus took the momentous decision to move to the West and there compose a monumental history of the Empire's decline.
IV BOOK 31 AS A MONOGRAPH IN GREEK
That conclusion brings with it a necessary corollary about the language in which Book 31 was originally composed. Ammianus must, at least in the first instance, have written in Greek; no one wishing to contest arguments of Libanius and Themistius would have written in Latin.Footnote 120 On the other hand, if composed in Greek, Ammianus' Bellum Gothicum poses problems of genre. Had it been composed in Latin, its monographic structure would have had an obvious model in Sallust, who has, with Livy, a far more visible imprint on Ammianus' Latinity than Tacitus, normally regarded as the inspiration for the Res Gestae.Footnote 121 However much Book 31 may, in its present form, resemble a ‘Sallustian monograph’, the Jugurtha cannot have been the model for an originally Greek composition.Footnote 122 We are therefore faced with two questions: was there a Greek analogue of the kind of historical monograph which Sallust exemplifies for us in Latin? And, if so, why would Ammianus choose a work of that type, rather than a more obviously rhetorical or polemical genre, in which to write?Footnote 123
The two questions are in fact interconnected, but the question of Greek models comes first; it is, after all, too much to imagine Ammianus inventing a new genre in Greek with which to argue about the relative significance of Adrianople with his eastern contemporaries. However, a third-century Greek analogue does indeed exist, in the shape of Dexippus' Scythica. The precise scope of Dexippus' writings is not altogether clear.Footnote 124 His first work, and sadly the only one extensively described by Photius, was a Τὰ μετὰ Ἀλέξανδρον along the lines of Arrian's work of the same title. His Chronica or Σύντομος Ἱστορικός comprised twelve books which, according to its continuator Eunapius, ran down to the reign of Claudius Gothicus.Footnote 125 A final work, the Scythica, seems to have dealt in monographic fashion with the third-century invasions, perhaps running from the reign of Philip to that of Aurelian, although that is mere inference from extant fragments, itself problematical because the attribution of fragments between the Chronica and the Scythica is so uncertain a business. A long passage involving Aurelian's negotiations with the Iuthungi is attributed to the Chronica in the Excerpta de Sententiis (ἐκ τῆς ἱστορίας Δεξίππου Ἀθενάιου. λόγος γ), yet it deals with Aurelian — not, according to Eunapius, treated in that work, and even were Eunapius mistaken, its book number (γ = 3) is too low for so late an event.Footnote 126 The passage, then, should belong to the Scythica, indeed to the very end of that work, but if so the numeration remains problematical, for what other Greek historical works appeared in three books? We cannot begin to judge the length of either the Chronica or the Scythica through the distorting effects of the Excerpta de Sententiis and de Stratagematibus, which preserve all we know of either work's contents. But what seems certain from those excerpts that can be securely attributed to the Scythica is that it was a monographic account in three or fewer books of wars with Scythian barbarians, consisting of many set-piece battles, sieges and much decoration on the Thucydidean model.Footnote 127 Given the length of even the fairly sparse excerpts still extant, the original will certainly have been longer than Ammianus' Book 31. Nevertheless, the Scythica of Dexippus is exactly the sort of monographic precedent that allows us to contemplate an original of Book 31 published not in Latin but in Greek.
The question of rationale follows naturally from that, for it is not just that Dexippus wrote a work like that which Ammianus produced. On the contrary, Dexippus also provides us with an explanation of why Ammianus chose to engage in contemporary debate in a historical genre not obviously suited to polemic. The point emerges from the consideration of Ammianus' Greek contemporary Eunapius of Sardis, the author of a series of lives of philosophers and sophists, and of a history that went through two editions, both of which Photius knew in the ninth century. The first of these editions was early, earlier than the Res Gestae, and published c. a.d. 380.Footnote 128 It was thus a sort of ‘instant history’, a response to Adrianople that described how the Romans ruined their empire. Eunapius, in other words, wrote a catastrophist interpretation of Rome's fourth century, no doubt blaming it, as a committed pagan, on Constantine's conversion — a historia adversus Christianos in the same way Orosius would later write a historia adversus paganos.Footnote 129 As an example of pagan apologetics, Eunapius both prefigured and served as a source for the Historia Augusta in its late and most tendentious sections.Footnote 130 In Book 31, as we have seen, Ammianus is relentlessly hostile to such catastrophism. He sets out explicit arguments against Adrianople's having been a larger disaster than any the Roman world had previously seen. It was not the fall of the Roman world — another Cannae perhaps, but one from which Rome would recover as she had from Cannae. Book 31, in other words, may well be attacking Eunapius' history.Footnote 131 Thus far, we have placed Ammianus' arguments up against those of Libanius and Themistius because we can be certain that they wrote and operated in overlapping circles and because the overlap of argumentation is precise. Too little of Eunapius survives for us to know how widely his work circulated or to state confidently that a man in Ammianus' position would surely have seen it. Nevertheless, given that the arguments of Eunapius and Ammianus are so clearly in opposition one to the other, it is at least possible that Ammianus was responding not just to Themistius and Libanius, but also to a Greek history that he knew had just appeared.Footnote 132 He had probably read Eunapius, though he need not necessarily have done so.Footnote 133 He will, however, have known that Eunapius' history continued Dexippus' Chronica and argued that Adrianople was the climax of Roman disasters.Footnote 134 What better way to combat that assertion than to write a similarly Dexippan response arguing the opposite point? Doing so in monographic form in fact trumped Eunapius in two ways. First, using the Scythica rather than the Chronica as a model was an act of homage that paid greater tribute, and certainly showed more originality of inspiration, than the continuation essayed by Eunapius.Footnote 135 Secondly, the Dexippan fragments display a much greater emphasis on human than divine responsibility for historical events. Given that, Ammianus' focus on human rather than divine explanations of the Adrianopolitan disaster was not just an answer to Libanius and Themistius, but also a truer reflection of the Dexippan model than was the catastrophism of Eunapius.Footnote 136 That is just the sort of literary one-upsmanship which Ammianus clearly and consistently relished.Footnote 137
V CONCLUSIONS
Taking Book 31 out of its present location at the end of the Res Gestae and examining it on its own terms suggests very strongly that it was originally a monographic response to the immediate controversy over the Battle of Adrianople and its meaning. The Antiochene and palatine responses of Libanius and Themistius gave Ammianus the impetus for writing, and presented many of the views against which he would argue. Eunapius (and, because of Eunapius, Dexippus) determined the generic form the writing would take. The hypothesis is not without real impediments, chiefly cross-references to earlier books of the Res Gestae. Those references are neither numerous nor integrated enough to rule out the case for separate composition. They do, however, demonstrate that as extant, Book 31 is not an unambiguous relic of a.d. 382. It was retouched in order to provide a conclusion to the Res Gestae some time in a.d. 389/390 or a little later. We might reasonably hypothesize that in composing the original of Book 31, Ammianus found himself inspired to write a larger work of history in which the heroic young Julian could appear as everything Valens, and after him Theodosius, were not. But we can only speculate about the circumstances in which the larger Res Gestae ultimately concluded with the partial reuse of older material.Footnote 138 The argument of the present piece can be summarized briefly. Regardless of the compositional structures underlying the larger work, Book 31 remains notably different from the other books of the Res Gestae, whether the integral unit made up by Books 26–30 or the rest of the extant books. Substantive differences of structure, tone, and content can best be explained if Book 31 was originally a separate monograph, written in Greek, to challenge the theories and arguments of other Greek authors about Adrianople. The ties that bind it to the earlier books are real but clumsy, the result of hasty surgery. We cannot hope to determine why such an expedient should have been necessary, but the evidence outlined above suggests that it was.
