Hostname: page-component-cd9895bd7-7cvxr Total loading time: 0 Render date: 2024-12-25T07:26:45.752Z Has data issue: false hasContentIssue false
  • English
  • Français

Coded Polemic in Ammianus Book 31 and the Date and Place of its Composition*

Published online by Cambridge University Press:  07 June 2012

Michael Kulikowski
Michael Kulikowski*
Affiliation:
Penn State
*
[email protected]
Get access Rights & Permissions [Opens in a new window]

Abstract

The Res Gestae of Ammianus Marcellinus poses numerous structural puzzles for the historian, among them the anomalous final book, numbered 31 in the manuscript tradition. This book, which treats the Gothic rebellion of a.d. 376–378 and the campaign of Adrianople, is loosely connected to the other extant books, which conclude with events of a.d. 375. The present article argues that Book 31 was in origin a separate monograph, drafted in Greek at Antioch in the aftermath of the Roman defeat at Adrianople. Perhaps modelled on the Scythica of Dexippus, its contents reflect the Antiochene and Constantinopolitan polemic of its moment. For reasons that must remain speculative, Ammianus later translated his work into Latin and appended it to a finished draft of the Res Gestae.

Keywords

Ammianus MarcellinusEunapiusAdrianopleLater Roman EmpireValenslate antique historiography
Type
Articles
Information
The Journal of Roman Studies , Volume 102 , November 2012 , pp. 79 - 102
Copyright
Copyright © The Author(s) 2012. Published by The Society for the Promotion of Roman Studies

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

Footnotes

*

I first presented a version of this paper in December 2007 at the colloquium in honour of my doctoral supervisor Professor T. D. Barnes, on the occasion of his retirement from the University of Toronto. A substantially revised version was presented in the Society for Late Antiquity sessions at the 43rd International Medieval Conference, Kalamazoo, 2008. I received valuable comment, positive and otherwise, from those present at each event, and not least from the honorand of the original colloquium. I have also profited enormously from the repeated, acute readings of Gavin Kelly and Noel Lenski, and from the anonymous readers for the journal, for the most sceptical of whom I reserve especial thanks.

References

1 Fragments of works by Renatus Profuturus Frigeridus and Sulpicius Alexander, who wrote in the fifth century, are preserved in small excerpts in Gregory of Tours' Historiae. These fragments suggest but cannot prove that they wrote classicizing history in the same vein. For the much-discussed Annales of Nicomachus Flavianus, see now Cameron, Alan, The Last Pagans of Rome (2010), 627–90 and n. 130 belowCrossRefGoogle Scholar.

2 Kelly, G., Ammianus Marcellinus. The Allusive Historian (2008)Google Scholar. Kelly breaks new ground on the way Ammianus used allusion to earlier authors to signal historical judgements, but the way he parcels his material out between books has never been examined in the same way that Syme treats Tacitus' Annales. Syme, R., Tacitus (1958), 253–321Google Scholar.

3 Priscian normally quoted from a work's first book and his citation from Ammianus comes from Book 14. For a summary of the transmission, see Reynolds, L. D. (ed.), Texts and Transmission (1983)Google Scholar.

4 Michael, H., Die verlorenen Bücher des Ammianus Marcellinus (1880)Google Scholar, the most acute of the nineteenth-century writers on Ammianus, did not believe that the whole period from a.d. 96 to 378 could have been covered in thirty-one books and so postulated two separate works, by analogy with Tacitus' Historiae and Annales.

5 Asymmetric composition, such as that implied by the transmitted book numbers of the Res Gestae, is vanishingly rare in antiquity. Of the three possible comparisons — Pliny, Augustine's Confessions, and Apuleius — only that of Apuleius is superficially comparable to Ammianus. But whereas Book 11 of the Golden Ass provides the key which unlocks all that has come before, Ammianus' Book 31 plays no such structural rôle. It is an unhappy denouement to an unhappy history, but it does not unlock new meanings in the reign of Constantius in the way Lucius' initiation unlocks new meanings in the Cupid and Psyche story.

6 As they had to: to move beyond that date would have meant writing about living emperors, Theodosius I and Valentinian II, and about the execution of a reigning emperor's father in mysterious circumstances.

7 Barnes, T. D., Ammianus Marcellinus and the Representation of Historical Reality (1998), 20–31Google Scholar, which draws radical implications for the number of books in the original Res Gestae from the hexadic structure, viz. the loss of eighteen rather than thirteen books, and the subsequent misnumbering of those extant. The content of the extant eighteen books is clearly grouped into three hexads (14–19; 20–25; 26–31 using the traditional numbering) which resolve fairly readily into paired triads. The one real question lies with the contents of Book 14, which pick up the prosecution of Magnentius' followers and the career of the Caesar Gallus in media res. It is hard to see why these should stand at the start of a new hexad, but the long opening sentence is certainly stylistically consonant with the way Ammianus opens other hexads at 20.1.1 and 26.1.1 and, if the last of the lost books ended with the death of Magnentius himself, that episode would have formed a fitting climax to a hexad. Barnes, op. cit., 28, does not fully address that problematical transition when postulating a hexad running from (renumbered) Books 13–18 covering ‘Constantius’ rise to sole rule over the Roman empire' and 19–24 (extant 14–19) covering ‘Gallus and Julian as Caesars’. Sabbah, G., La Méthode d'Ammien Marcellin (1978), 305Google Scholar, long ago noted Ammianus' taste for symmetry and balance in his compositions.

8 That corollary is inescapable: if the extant books are hexadic, then symmetrical composition by hexad must also have been observable in the lost books, but if only thirteen books are lost, as the transmitted numbering requires, they could not have resolved into hexads. For symmetrical composition by hexad to have carried backwards into the lost books, the whole work must originally have totalled either twenty-four or thirty-six books, with either eighteen or six books lost, and the extant books misnumbered in the transmitted tradition. Given that the work began, however cursorily, with the reign of Nerva, a longer original is more plausible than a shorter, but that is not a decisive argument. Palaeography is more persuasive: it is easier to explain the corruption of ‘xxxvi’ than ‘xxiv’ (or ‘xxiiii’) to the transmitted ‘xxxi’, since the former merely requires the loss of a single letter of two minims, while the latter requires the loss of one or more letters and then the addition of a new one.

9 Most recent work on Ammianus accepts hexadic structure without belabouring it, although often without acknowledging the corollary of thirty-six books: e.g., many of the essays in Boeft, J. den et al. (eds), Ammianus after Julian: The Reign of Valentinian and Valens (2007)Google Scholar; and Boeft, J. den et al. (eds), Philological and Historical Commentary on Ammianus Marcellinus XXVI (2008)Google Scholar. See Sabbah, G., Ammien Marcellin: Histoires. Livres XXIX–XXXI (1999), xlvGoogle Scholar. The only option that allows for both hexadic structure and the transmitted book numbers is to conjecture that Ammianus' Book 1 was a sort of index or table of contents for the work that was soon given its own book number in transmission, as happened with Pliny's Historia Naturalis. I owe this suggestion to Gavin Kelly; it clarifies the otherwise opaque statement at Kelly, op. cit. (n. 2), 3.

10 Books 16–20, 21–25, and 26–30 all work as pentads, but not as segments of three decades, and to argue for a pentadic structure is immediately to run into the problem of asymmetric composition, discussed above.

11 This separateness has often been noted: Sabbah, G., ‘Ammien Marcellin, Libanius, Antioche et la date des derniers livres des Res Gestae’, Cassiodorus 3 (1997), 89–116, at 113Google Scholar, who believes that ‘le livre 31 possède une certaine indépendence dans le bloc des derniers livres et qu'il a pu être publié séparément, après la mort de Théodose’; in Sabbah, op. cit. (n. 9), vii, even while asserting the essential unity of the last six books, he admits that the last three books are ‘sinon une triade, du moins un ensemble très cohérent’; similarly Bitter, N., Kampfschilderungen bei Ammianus Marcellinus (1976), 104Google Scholar; Matthews, J. F., The Roman Empire of Ammianus (1989), 31, 481 n. 34Google Scholar; Blockley, R. C., ‘Ammianus Marcellinus’ use of exempla', Florilegium 13 (1994), 53–64, at 60 n. 27Google Scholar.

12 For Procopius, Adshead, L., ‘The secret history of Procopius and its genesis’, Byzantion 63 (1993), 5–28Google Scholar; contra, Kaldellis, A., Procopius of Caesarea: Tyranny, History and Philosophy at the End of Antiquity (2004), 142–50Google Scholar. The bibliography on the Confessions is vast, but although its Book 10 certainly does treat many of the themes found elsewhere in the work, it is sufficiently different in scope and tone to best be explained as an initially separate composition.

13 The case against Ammianus as recipient does not rest, as Matthews, op. cit. (n. 11), 478–9 n. 1 and Matthews, J. F., ‘The origin of Ammianus’, CQ n.s. 44 (1994), 252–69CrossRefGoogle Scholar, suggest, primarily on the connotations of the word συνγραϕῆ, but rather in the tone of the Greek. Pace Sabbah, op. cit. (n. 11), 89–97 and op. cit. (n. 9), xxxiii, Fornara, C. W., ‘Studies in Ammianus Marcellinus I: the letter of Libanius and Ammianus’ connection with Antioch', Historia 41 (1992), 329–44Google Scholar, is decisive: Libanius cannot have written to Ammianus in the words he uses in Ep. 1063 unless deliberately seeking to give insult. Note, with P. Barceló, ‘Überlegungen zur Herkunft des Ammianus Marcellinus’, in Vogel-Weidemann, U. and Scholtemijer, J. (eds), Charistion C.P.T. Naudé (1993), 17–23Google Scholar; Sabbah, op. cit. (n. 11), 97–107; and Kelly, op. cit. (n. 2), 109–18, that rejecting the identification of Libanius' Markellinos with our historian Ammianus is no reason to reject the latter's Antiochene origin, whether in favour of the Thessalonica of Fornara, op. cit. above; the Alexandria of Bowersock, G. W., ‘Review of Matthews, The Roman Empire of Ammianus’, JRS 80 (1990), 244–50Google Scholar; or the Phoenicia of Barnes, op. cit. (n. 7), 54–64.

14 Matthews, op. cit. (n. 11), 17–32, lays out all the dispositive evidence for date and what follows here is treated at length there. Sabbah, op. cit. (n. 11), attempts to restate older arguments for a late date of composition for, at least, Books 26–31 and suggests that Book 31 is different enough from what precedes it that it may represent a still later stage of composition — perhaps Stilichonian.

15 Julianus' prefecture: Amm. Marc. 27.6.2. For the food shortage: 14.6.19, with Cameron, A., ‘The Roman friends of Ammianus’, JRS 54 (1964), 15–28Google Scholar, on the date; Kelly, op. cit. (n. 2), 133–41 is bracingly sceptical on Ammianus' having shared in this displacement. As Matthews, op. cit. (n. 11), 23, shows, it is tempting but not necessary to make Praetextatus one of Ammianus' informants for the activities of Julian in Constantinople in a.d. 361 — and if that is the case, to place Ammianus' arrival in Rome before the death of Praetextatus in a.d. 384. A close relationship between Ammianus and Praetextatus is assumed in the Francophone literature, but this has more to do with the ‘pagan revival’ of Praetextatus and Flavianus than with the evidence. See Cameron, op. cit. (n. 1), 627–90 and passim.

16 Amm. Marc. 27.11.2: ‘Et licet potuit, quoad uixit, ingentia largiendo …’

17 cf. Amm. Marc. 17.12.17 and Pan. Lat. 2.27.3.

18 Barnes, op. cit. (n. 7), 201–8, on the impact of lacunae.

19 Maenchen-Helfen, O. J., ‘The date of Ammianus Marcellinus’ last books', AJPh 76 (1955), 384–99Google Scholar, recognized that Book 31 could in theory have been composed before the rest of the work, while also advocating a much later date.

20 As at Amm. Marc. 26.10.19, for the decaying ship cast up in Mothone by the tidal wave of a.d. 365: ‘nos transeundo conspeximus’. Matthews, op. cit. (n. 11), 17, would like to accept Amm. Marc. 31.7.16 as evidence of autopsy, while recognizing the fragility of the attestation. Bitter, op. cit. (n. 11), 7 and passim, requires Ammianean autopsy here in order to sustain his tripartite schema of Ammianus' battle descriptions as participant (Amida), reader (Strasbourg), and researcher of battle sites (Adrianople). There is a strong stylistic parallel to this line at Amm. Marc. 15.11.12: ‘ut aedificia semiruta nunc quoque demonstrant.’

21 Verg., Aen. 5.864 and 12.36; Tacit., Ann. 1.61.2–3 deploys the same Vergilian echo for the aftermath of Varus' disaster. Kelly, op. cit. (n. 2), 13–19, explores the full range of the allusions and comes down against autopsy. Note additionally that Libanius, Or. 24.4, with which Ammianus is in dialogue (see below), applies an identical image to Adrianople and see Sabbah, op. cit. (n. 7), 282.

22 This leaves aside as improbable the suggestion of F. Trombley, ‘Ammianus Marcellinus and fourth-century warfare: a protector's approach to historical narrative’, in Drijers, J. W. and Hunt, D. (eds), The Late Roman World and its Historian (1999), 17–28Google Scholar, that Ammianus might have continued in service as a protector as late as the a.d. 380s.

23 Sabbah, op. cit. (n. 9), 97, but Clark and Seyfarth's reading of inter haec is to be preferred.

24 The examples cited are: Amm. Marc. 18.1.1; 22.1.1; 20.1.1. Other comparably informative transitions occur at Amm. Marc. 16.1.1; 23.1.1; 27.1.1; 28.1.1. On the programmatic goal of these opening sentences, Demandt, A., Zeitkritik und Geschichtsgebild im Werk Ammians (1965), 107Google Scholar; Brodka, D., Ammianus Marcellinus. Studien zum Geschichtsdenken im vierten Jahrhundert n. Chr (2009), 55Google Scholar.

25 If inter haec was instead deployed to link Books 30 and 31 very tightly together, as if they were mere chapters rather than books, then it can only have been meant to disguise the dissimilarity of the books' structure and content.

26 For the literary merits of the sphragis, G. Kelly, ‘The sphragis and closure of the Res Gestae’, in den Boeft et al., op. cit. (n. 9), 219–41.

27 Amm. Marc. 31.3.4: ‘Haec ita praeter spem accidisse doctus Athanarichus, Theruingorum iudex, in quem, ut ante relatum est, ob auxilia missa Procopio dudum Valens commoverat signa, stare gradu fixo temptabat, surrecturus in uires, si ipse quoque lacesseretur, ut ceteri.’ Valens' earlier Gothic war is treated at Amm. Marc. 17.5, and at Amm. Marc. 17.5.7 Valens' eventual death at Adrianople is alluded to, although nowhere is the rôle of Athanaric in these later events mentioned, merely the fact of his death and honourable burial at Constantinople in a.d. 381 (Amm. Marc. 17.5.10).

28 Amm. Marc. 31.10.21: ‘Frigerido … successor Maurus nomine mittitur comes, uenalis ferociae specie et ad cuncta mobilis et incertus; is est, quem praeteritorum textu rettulimus ambigenti super corona capita imponenda Iuliano Caesari, dum inter eius armigeros militaret, arroganti astu fidenter torquem obtulisse collo abstractam’, an episode treated fully in Amm. Marc. 20.4.18.

29 Amm. Marc. 31.16.5: ‘Saracenorum cuneus, super quorum origine moribusque diuersis in locis rettulimus plura, ad furta magis expeditionalium rerum quam ad concursatorias habiles pugnas recens illuc accersitus congressurus barbarorum globo repente conspecto a ciuitate fidenter erupit diuque extento certamine pertinaci aequis partes discessere momentis.’

30 For the other possible connections of the Saracen episode in Book 31 see below.

31 Amm. Marc. 31.14.8: ‘Haec super Valente dixisse sufficiet, quae uera esse aequalis nobis memoria plene testatur. Illud autem praeteriri non conuenit, quod, cum oraculo tripodis, quem mouisse Patricium docuimus et Hilarium, tres uersus illos fatidicos comperisset, quorum ultimum est ‘en pedíoisi Mímantos agaioménoio Áreos’, ut erat inconsummatus et rudis, inter initia contemnebat, processu uero luctuum maximorum abiecte etiam timidus eiusdem sortis recordatione Asiae nomen horrebat, ubi Erythraeo oppido superpositum montem Mimanta et Homerum scripsisse et Tullium doctis referentibus audiebat. 9. Denique post interitum eius, discessumque hostilem prope locum, in quo cecidisse existimatus est, inuentus dicitur saxeus monumenti suggestus, cui lapis affixus incisis litteris Graecis sepultum ibi nobilem quendam Mimanta ueterem indicabat.'

32 The references are to Cicero, Ad Att. 16.13.2, echoing Homer, Odyss. 3.169–72.

33 Amm. Marc. 29.1.33: Οὐ μὰν νεποινί γε σὸν ἔσσεται αἷμα καὶ αὐτοῖς/ Τισιϕόνε βαρύμηνις ἐϕοπλίσσει κακὸν οἶτον/ ἐν πεδίοισι Μίμαντος ἀγαοιμένοιο Ἄρεος.

34 Amm. Marc. 31.14.1: ‘Post exitialem pugnam cum iam tenebris nox terras implesset …’ What is more, the ex eventu prophecy is one of several places, all of them in the Valentinianic books, where Ammianus includes material shared by late Byzantine authors like Zonaras and Cedrenus (here Zon. 13.16.20–4; Cedr. 549.20–550.3), on which Bleckmann, B., Die Reichskrise des III. Jahrhunderts in der spätantiken und byzantinischen Geschichtsschreibung. Untersuchungen zu den nachdionischen Quellen der Chronik des Johannes Zonaras (1992)Google Scholar, though he is rather too willing to see a mechanical use of written sources by Ammianus. Note that where they do overlap, there are very significant divergences between the late Greek tradition and that in Ammianus, and that these divergences do not fit into a consistent pattern. At Amm. Marc. 31.4.8–9, we cannot rule out the use or correction of Eunapius, on whom see Section IV below.

35 Cross-references: Amm. Marc. 19.2.3; 19.13.1; 22.8.49; 25.4.12; 25.4.23; 25.6.4; 25.10.17; 26.8.5; 28.1.1; 28.1.47; 28.1.57; 28.6.30; 29.1.25; 30.1.1; 30.5.16. Kelly, op. cit. (n. 2), 306–13, unintentionally illustrates the contrast.

36 For Strasbourg and Adrianople contrasted, Kelly, op. cit. (n. 2), 313–16; and Brodka, op. cit. (n. 24), 54–66; 106–26.

37 The references to King Pap of Armenia at the start of Book 31 are considered below.

38 Mommsen, Th., ‘Ammians Geographica’, Hermes 16 (1881), 602–36= Ges. Schr. 7, 393–425Google Scholar; Kelly, G., ‘Ammianus Marcellinus: Tacitus' heir and Gibbon's guide’, in Feldherr, A. (ed.), The Cambridge Companion to the Roman Historians (2009), 348–61, at 353–3CrossRefGoogle Scholar. For the structural locations of excursus, Barnes, op. cit. (n. 7), 32–42, 222–4.

39 Mommsen, op. cit. (n. 38), 604. The excursus on Thrace, Amm. Marc. 27.4.1–14, fits his standard pattern; that on the lands beyond the Black Sea in Amm. Marc. 31.2–3 does not.

40 Matthews, op. cit. (n. 11), 334, 347 for the postulated link between the excursus.

41 Matthews, op. cit. (n. 11), 334. Ammianus clearly knows Herodotus well, for bits of the ethnographic material in Herodot. 4.46–120 appear throughout the Res Gestae, with a striking preponderance in 31.2–3.

42 Bitter, op. cit. (n. 11), introduces a very rigid typology of battles reported, battles experienced, and battles researched by the author.

43 Sabbah, op. cit. (n. 7), 470–1, 586–8.

44 Kelly, op. cit. (n. 2), 313–16, including a verbal parallel; Sabbah, op. cit. (n. 7), 502–6.

45 At Amm. Marc. 23.5.21, Ammianus makes Julian himself the mouthpiece of this sentiment. In general, Brodka, op. cit. (n. 24), 29–31.

46 Sabbah, op. cit. (n. 9), xix–xxxii is a sensitive reading of Ammianus' worldview in the last three books that is equally applicable to the Res Gestae as a whole.

47 Personal intrusion: Amm. Marc. 31.2.1 (‘hanc comperimus causam’); 31.5.10 (‘et quoniam ad has partes post multiplices ventum est actus’) a reference to the Gothic narrative, not necessarily the Res Gestae as a whole. Necrologies: Brandt, A., Moralische Werte in den Res Gestae des Ammianus Marcellinus (1999), 55–60CrossRefGoogle Scholar; Kelly, op. cit. (n. 2), 265. Exempla: Blockley, op. cit. (n. 11). Habits of expression, e.g. the formula for ending an excursus: Amm. Marc. 31.2.25; 31.5.17 (cf. 21.1.14; 23.4.14); or rettulimus for cross-references, twice in Book 30 (2.10; 7.11) and more than once in almost every other book save 17–18, 26–27, and 29. See Sabbah, op. cit. (n. 9), xxvi for Ammianus' use in these books of words expressing horror (horror, funus, luctus and their cognates) but note that they are commonplace throughout the Res Gestae, not merely here.

48 Barnes, op. cit. (n. 7), 28, 39–42; Sabbah, op. cit. (n. 9), viii–xi. The ties among Books 26–30, and particularly those between 29 and 30, are very strong. Note that to bracket these books together as a single compositional phase is not to suggest that they are a sort of afterthought to the earlier books, nor that they were composed much later than them.

49 See, e.g., Amm. Marc. 27.2.11; 28.1.2; 28.2.12; 29.2.9; 29.3.1.

50 cf. Sabbah, op. cit. (n. 9), ix, xvii, where the structural analysis of Book 31 actually reveals its dissimilarity to the preceding two books, rather than the parallelism and the maintenance of a ‘principe de l'alternance’ between East and West that it is claimed to show.

51 Cross-references, both forward and backward, within Books 26–30 are quite specific and closely connected (see n. 35 above), with the famous exception of 28.1.57, the one time Ammianus promises to give an account of something (viz., Maximinus' execution under Gratian) and fails to do so.

52 Amm. Marc. 29.1.15: ‘fato reflectente depulsum [viz. ferrum], quem lacrimosis in Thracia discriminibus destinarat.’ Nothing here suggests the account in the extant Book 31, pace Sabbah, op. cit. (n. 9), xliv.

53 Amm. Marc. 30.2.7–8.

54 One might perhaps have considered Amm. Marc. 31.1.3 above, among the cross-references to earlier books. Unlike the genuine cross-references, however, Ammianus gives the reader of Book 31 no indication that he has treated these events elsewhere.

55 Sabbah, op. cit. (n. 9), 269 n. 470 (by Laurent Angliviel de la Beaumelle) falls back on Ammianus' use of a different source for Armenia than for Adrianople to explain this — but Ammianus is no Zosimus, changing his opinions when he changes source.

56 Note, too, that the introduction of these figures, at Amm. Marc. 30.1.11 and 30.1.18 respectively, offers no forward reference to the Adrianople campaign in which both fell.

57 Thus, although the Lentienses are introduced as if for the first time at Amm. Marc. 31.10.2 as a populus Alamannicus, their earlier mention in 15.4 is sufficiently distant in terms of narrative space that the absence of a cross-reference is not significant. Again, when Saturninus appears at Amm. Marc. 31.8.3, there is no indication of his appearance as ‘ex cura palatii’ in 22.3.7.

58 Amm. Marc. 31.7.11: ‘Et Romani quidem uoce undique Martia concinentes a minore solita ad maiorem protolli, quam gentilitate appellant barritum, uires ualidas erigebant.’ A similar argument might apply, though with less force, to the introduction of the scorpio at 31.15.12 (‘scorpio, genus tormenti, quem onagrum sermo uulgaris appellat’), which had been part of the long excursus on artillery in 23.4.4 (‘scorpionis autem, quem appellant nunc onagrum’).

59 See Thompson, E. A., The Historical Work of Ammianus Marcellinus (1947), 42–55Google Scholar; Barnes, op. cit. (n. 7), 9–10, 117.

60 Amm. Marc. 31.10.1: ‘Quae temporum rabies uelut cuncta cientibus Furiis ad regiones quoque longinquas progrediens late serpebat.’

61 What is more, the necrology of Valens is the only one introduced by the firm announcement that a special section on imperial character is being essayed: ‘Cuius bona multis cognita dicemus et uitia’ (Amm. Marc. 31.14.2). As G. Zecchini, ‘Greek and Roman parallel history in Ammianus’, in den Boeft et al., op. cit. (n. 26), 201–18, at 214, notes, the necrology of Valens is also unusual in not including comparisons to great Greek and Roman figures from the past. What is more, the more subtly introduced necrologies of Constantius, Julian, Jovian and Valentinian all suggest a stylistic development beyond the abrupt transition found in Book 31.

62 Straub, J., ‘Die Wirkung der Niederlage bei Adrianopel auf die Diskussion über das Germanenproblem in der spätrömischen Literatur’, Regeneratio Imperii I (1972)Google Scholar, reprinted from Philologus 95 (1943), 255–86; Lenski, N., ‘Initium mali Romano imperio: contemporary reactions to the Battle of Adrianople’, TAPA 127 (1997), 129–68Google Scholar.

63 This is the position of Lenski, op. cit. (n. 62), 160–3.

64 Dagron, G., ‘L'Empire romain d'Orient au IVe siècle et les traditions politiques de l'hellénisme. Le témoignage de Thémistios’, Travaux et Mémoires 3 (1968), 1–242Google Scholar, at 110, places Ammianus within the eastern, not the western, debate on Adrianople and the ‘barbarian crisis’, though he does not draw out the consequences of this prescient observation.

65 viz., Ambr., De excessu fratris 1.30; De fide 16.136–40; Rufinus, HE 11.13, though the latter may have been written too late to bear on the question at hand (that is certainly true of Oros., Hist. 7.33.15 which shares Rufinus' explanation of Valens' defeat on account of his Arianism and persecution of Nicenes). One should note that the stereotyped themes of Pacatus had all been aired independently by Themistius between a.d. 379 and 382. Jerome's developing views are treated in Lenski, op. cit. (n. 62), 157–9, and show no connections with Ammianus. In general, it is quite rare to find a direct echo of contemporary western writings in the Res Gestae as extant, even where one might expect it, though see Gutzwiller, H., Die Neujahrsrede des Konsuls Claudius Mamertinus vor dem Kaiser Julian. Text, Übersetzung und Kommentar (1942)Google Scholar, 194 on Pan. Lat. 3.20.4 for its use at Amm. Marc. 18.4.3 and, ibid., 190 for a less likely parallel between 3.19.4 and 18.4.2.

66 For the use of Pacatus see n. 118 below.

67 See, e.g., Vanderspoel, J., Themistius and the Imperial Court. Oratory, Civic Duty and Paideia from Constantius to Theodosius (1995), 145Google Scholar; Petit, P., ‘Recherches sur la publication et la diffusion des discours de Libanius’, Historia 5 (1956), 479–509Google Scholar, for the two authors in dialogue with one another.

68 PLRE 1: 547–8 has the essentials.

69 Sabbah, op. cit. (n. 7), 251 with a helpful list of such cases.

70 For the impact of a ‘fama antiochéene’, very felicitously put, see Sabbah, op. cit. (n. 7), 266. By contrast, the notes to Lib., Or. 18 (the Epitaphios) in Förster's edition (2: 222–371) find far too many parallels with Ammianus. Most, if not all, are coincidental.

71 e.g. Sabbah, op. cit. (n. 7), 271.

72 Our Marcellinus cannot be the addressee of the letter to Markellinos (n. 13 above), but cf. the other demonstrations of the literary relationship between Libanius and Ammianus at Sabbah, op. cit. (n. 7), 244–5.

73 cf. Lib., Or. 18.297 and Amm. Marc. 25.3.8. For the ordering of events in each work as a demonstration of Ammianus' use of Libanius, Sabbah, op. cit. (n. 7), 274.

74 Sabbah, op. cit. (n. 7), 278–80 for Or. 12 and 14 and Amm. Marc. 22; ibid., 286–7 for Or. 51 and 52 and Amm. Marc. 30.4 on lawyers.

75 On the diffusion of Or. 24, Sabbah, op. cit. (n. 7), 281–4.

76 For Chrysostom and other Greek patristic writers on Adrianople, see Lenski, op. cit. (n. 62), 149–53. Ammianus, as one would expect, ignores the explicitly Christian debate on the battle altogether.

77 Barnes, op. cit. (n. 7), 79–94, for the polemic; Rike, R. L., Apex Omnium: Religion in the Res Gestae of Ammianus (1987)Google Scholar and Brodka, op. cit. (n. 24), for two different reconstructions of a coherent theology, both of which overlook the frequency with which the divine and the supernatural are merely narrative devices.

78 On the intersection of divine fortuna and human action see inter alia Demandt, op. cit. (n. 24), 99–111; Brodka, op. cit. (n. 24), 32–40, with 76–87 on Julian's Persian campaign as an extended illustration of the motif.

79 Especially Amm. Marc. 29.1.15–16.

80 Rosen, K., ‘Wege und Irrwege der römischen Gothenpolitik in Ammians 31. Buch’, in Boeft, J. denet al. (eds), Cognitio Gestorum. The Historiographic Art of Ammianus Marcellinus (1992), 85–90Google Scholar, has already noted Ammianus' rejection of a supernatural explanation for Adrianople and the theme is developed at considerable length in Brodka, op. cit. (n. 24), 106–26, contra, Marié, M.-A., ‘Virtus et Fortuna chez Ammien Marcellin. La responsabilité des dieux et des hommes dans l'abandon de Nisibe et la défaite d'Adrianople’, REL 67 (1989), 179–90Google Scholar, which reads a great deal of significance into the purely narratological references to the Furies in Amm. Marc. 31.1.1 and 31.10.1 and at various places in Book 29.

81 Amm. Marc. 31.4.9: ‘homines maculosi’. See also 31.5.1–3 and 5–6; 31.5.9: ‘with more haste than discretion’ (‘temere magis quam consulte’); contrast Zosimus 4.20.7, from Eunapius, blaming barbarian faithlessness.

82 Amm. Marc. 31.12.3: ‘procaci quodam calore perculsus eisdem occurrere festinabat.’ On the rôle of temeritas as the cause of Roman errors during the Gothic war, Brodka, op. cit. (n. 24), 115.

83 Amm. Marc. 31.8.8: ‘Inter quae cum beluae ritu traheretur ingenuus paulo ante diues et liber, de te, Fortuna, ut inclementi querebatur et caeca …’

84 Paschoud, F., Roma Aeterna. Études sur le patriotisme romain dans l'Occident latin à l'époque des grandes invasions (1967), 40–2Google Scholar demonstrates the consistency with which Ammianus is willing to contemplate failure on the part of the military, and see Dagron, op. cit. (n. 64), 92–3 for differences in emphasis among his contemporaries.

85 Amm. Marc. 31.4.6: ‘Ita turbido instantium studio orbis Romani pernicies ducebatur.’

86 The diametrically opposed arguments here meet the criteria of Sabbah, op. cit. (n. 7), 243–5, 268, for establishing a relationship between Ammianus and Libanius. John Chrysostom, Ad vid. iun. 4–5 contains a somewhat parallel attack on Valens and his soldiers, but we have no evidence that Ammianus engaged with the writings of Christian contemporaries.

87 Sabbah, op. cit. (n. 7), 348 is, broadly speaking, correct to point out the dangers of assuming a direct dialogue between the Res Gestae and the works of Themistius, inasmuch as the latter may well be no more than the sole extant witnesses to views that were widespread in right-thinking court circles at the time. Readers will need to judge for themselves whether or not the examples presented here rise above that caveat.

88 The date cannot be determined precisely, but must be before the start of the year's campaigning season; Dagron, op. cit. (n. 64), 23.

89 The termini are provided by the presence of Athanaric at dinner with the emperor: we know that the Gothic king arrived in the capital on 11 January and that he died there on 25 January a.d. 381. Given those dates, Theodosius' dies imperii on the 19th seems the most appropriate date for the speech. See Scholze, H., De Temporibus librorum Themistii (1911), 51Google Scholar; Dagron, op. cit. (n. 64), 23; Vanderspoel, op. cit. (n. 67), 199–200.

90 Them., Or. 15.190c–d; 191d; 192d; 194b–d. Heather, P. and Moncur, D., Politics, Philosophy, and Empire in the Fourth Century. Select Orations of Themistius (2001), 231–5CrossRefGoogle Scholar, is a very apt summary of Themistius' technique and rationale in this speech.

91 Heather, P., Goths and Romans, 332–489 (1991), 165–8Google Scholar; and Heather and Moncur, op. cit. (n. 90), 216–17, are more plausible on Themistius' relationship to Theodosian propaganda — indeed imperial propaganda generally — than are Dagron, op. cit. (n. 64), 103–12 and Vanderspoel, op. cit. (n. 67), 201–5. If Penella, R. J., The Private Orations of Themistius (2000), 34Google Scholar is correct (as seems plausible), then even a seemingly anodyne progymnasma like Themistius' Or. 30, in praise of agriculture, might be put at the service of Theodosian propaganda.

92 The date is uncontroversial: Dagron, op. cit. (n. 64), 23. See Them., Or. 15.187a–b; 189a–c; Or. 16.201b.

93 The terms of this treaty are widely discussed: Heather and Moncur, op. cit. (n. 90), 259–64, following Heather, op. cit. (n. 91), 157–81, systematically overstate the independence allowed the Goths by this treaty; cf. Kulikowski, M., Rome's Gothic Wars from the Third Century to Alaric (2007), 150–3Google Scholar.

94 Or. 34.xxi–xxvi.

95 Or. 16.206d: ‘the unspeakable Iliad of disasters on the Danube, when no king yet ruled over the affairs of state, with Thrace laid waste, with Illyria laid waste, when whole armies had disappeared completely, like a shadow’. The careful chronology of Lenski, op. cit. (n. 62), brings this point out for the first time.

96 Amm. 31.5.11: ‘negant antiquitatum ignari tantis malorum tenebris offusam aliquando fuisse rem publicam, sed falluntur malorum recentium stupore confixi.’ This concern with the actual consequences that flow from an ignorance of the past is a consistent aspect of Ammianus' historical thought: Brodka, op. cit. (n. 24), 26–31.

97 Note the parallel to Herodian 2.15.6–8 here, missed at Baaz, E., De Herodiani fontibus et auctoritate (1909), 71Google Scholar.

98 Amm. Marc. 31.5.10: ‘Et quoniam ad has partes post multiplices ventum est actus, id lecturos, si qui erunt umquam, obtestamur, ne quis a nobis scrupulose gesta vel numerum exigat peremptorum, qui comprehendi nullo genere potuit. Sufficiet enim veritate nullo uelata mendacio ipsas rerum digerere summitates, cum explicandae rerum memoriae ubique debeatur integritas fida.’

99 cf. Amm. Marc. 22.6.2 where the plurimis refers to Aristophanes of Corinth, the subject of Lib., Or. 14. Themistius also lurks behind a plural in one other place, where Ammianus contradicts him about the severity of Valens' proscription of the partisans of Procopius (26.10.4: ‘proscriptiones et exilia et quae leuiora quibusdam uidentur, quamquam sint aspera’), where Sabbah, op. cit. (n. 7), 361, correctly detects Themistius behind the quibusdam. There are probably also points in the portrayal of Jovian, e.g. at Amm. Marc. 25.9.7 and 25.10.11, where Ammianus is specifically correcting Themistius (see Sabbah, op. cit. (n. 7), 358, and Boeft, J. denet al. (eds), Philological and Historical Commentary on Ammianus Marcellinus XXV (2005), 331–2Google Scholar, much more plausibly than P. Heather, ‘Ammianus on Jovian: history and literature’, in Drijers and Hunt, op. cit. (n. 22), 105–16, at 108). However, Sabbah, ibid., 351–2 is also correct to reject two other potential references to Themistius, at Amm. Marc. 29.2.18 and 30.8.14, respectively.

100 Them., Or. 15.190a. The equation of the adulatores and Themistius is drawn, inter alia, by Heather and Moncur, op. cit. (n. 90), 201. And note that the same Themistius had himself stated that Valens hated flatterers, even as he loved philosophers equally with generals: Or. 10.129 (elsewhere, at Or. 22.276, he engages in the classroom exercise of distinguishing friends from flatterers). That Themistius could be disparaged as flatterer even by those who stood to gain from his flattery is shown by Julian, Ep. ad Them. 254B–C.

101 Sabbah, op. cit. (n. 7), 349–50, for more signs of Ammianus' dislike of the Constantinopolitan orator.

102 Zosimus, HN 4.25–6. For Julius, PLRE 1: 481 (Iulius 2).

103 Zuckermann, C., ‘Cappadocian Fathers and the Goths’, Travaux et Memoires 11 (1991), 473–86Google Scholar; previous scholarship is exhaustively summarized in Elbern, S., ‘Das Gotenmassaker in Kleinasien (378 n. Chr.)’, Hermes 115 (1987), 99–106Google Scholar.

104 PG 46: 736–48, at 737A (17 February a.d. 380); PG 46: 416–32, at 424C (undated). See Kulikowski, op. cit. (n. 93), 145–7, for the full argument: put briefly, Julius decided that both the Goths in eastern army units and the young Gothic hostages of a.d. 376, now nearing military age, were a menace, whose destruction was necessary to prevent a repetition in Oriens of what had happened in Thrace. Beginning with the frontier forts, Ammianus' castra, his actions were imitated elsewhere, and provoked riots in Asia Minor (attested in the Cappadocian Gregory) that were put down with the indiscriminate brutality attested in Zosimus.

105 The corrected clausulation is that of Sabbah, op. cit. (n. 11), 98 n. 38, while the error goes back to the chapter titles in the edition of Valesius (see Kelly, G., ‘Adrien de Valois and the chapter headings in Ammianus Marcellinus’, CP 104 (2009), 233–42, at n. 41Google Scholar).

106 Office: ILS 773.

107 Lib., Or. 2.9.

108 Lib., Or. 1.169–70, with PLRE 1: 752 (‘Protasius’ 2); Martin, J. and Petit, P., Libanios: Discours 1 (1978), 256–7Google Scholar; Petit, P., Les fonctionnaires dans l'oeuvre de Libanius: analyse prosopographique (1994), 217Google Scholar.

109 Theoderet, HE 5.2–3.

110 It does not, however, demand Ammianus' close personal connection to Libanius and his friends tout court: as we saw above, and as Sabbah, op. cit. (n. 7), 250–3 shows, the points of overlap between major Antiochene figures in Ammianus and the correspondants of Libanius include as many striking absences as connections, while the men frequently provide very different judgements on the qualities of various individuals.

111 See, e.g., H. Sivan, ‘Ammianus’ terminus and the accession of Theodosius I', in Vogel-Weidemann and Scholtemijer, op. cit. (n. 13), 113–20, though note that her account of Theodosius' accession is now superseded by McLynn, N., ‘Genere Hispanus: Theodosius, Spain, and Nicene orthodoxy’, in Bowes, K. and Kulikowski, M. (eds), Hispania in Late Antiquity: Current Approaches (2005), 77–120Google Scholar. Kelly, op. cit. (n. 2), 24–9, is the best treatment of the way Ammianus' language throughout Book 31 implies criticism of the Theodosian settlement.

112 It was also a further implicit attack on Libanius, who had made his peace with Julius' Christian successor almost immediately: Lib., Or. 2.9. Sabbah, op. cit. (n. 9), xxxvi suggests that the praise of the elder Theodosius in Book 29 might be read as dispraise for his son.

113 Straub, op. cit. (n. 62), 199–200.

114 Exhaustively treated in Brandt, op. cit. (n. 47), 108–19, building on Seager, R., Ammianus Marcellinus: Seven Studies in his Language and Thought (1986), 76–80Google Scholar, though Ammianus' usage is drawn from Cicero: Michael, H., De Ammiani Marcellini studiis Ciceronianis (1874)Google Scholar.

115 In the same way, Julian had inflicted discrimina multa on the fleeing Alamanni at Strasbourg (Amm. Marc. 16.12.57).

116 Them., Or. 16.211a. These improving sentiments continue down through 212a.

117 Lib., Or. 24.5. Libanius is critical of the soldiery elsewhere in his oeuvre, especially Or. 2.37–40, but the context is altogether different.

118 cf. Amm. Marc. 17.12.17 and Pan. Lat. 2.27.3.

119 Amm. Marc. 16.10, with the demonstration of Sabbah, op. cit. (n. 7), 325–30.

120 Blockley, R. C., The Fragmentary Classicizing Historians of the Later Roman Empire: Eunapius, Olympiodorus, Priscus and Malchus (1981), 25Google Scholar suggests en passant that the whole of the first draft of the Res Gestae may have been written in Greek in Antioch, a suggestion which has never, to my knowledge, been seriously pursued.

121 On Sallust, Hertz, M., De Ammiani Marcellini studiis Sallustianis (1874)Google Scholar. On Tacitus, Riedl, P., Faktoren des historischen Prozesses. Eine vergleichende Untersuchung zu Tacitus und Ammianus Marcellinus (2002)Google Scholar. On the general difficulty of finding direct evidence of Tacitus in Ammianus, Kelly, op. cit. (n. 38), 348–61.

122 Blockley, op. cit. (n. 11), 60 n. 27, argued for Book 31 as a monographic treatment in which ‘Sallust's monograph on the Jugurthan war immediately comes to mind’. Sabbah, op. cit. (n. 9), xlii notes that there is much more Sallustian content in Book 29, but the focus on Africa would have made that virtually inevitable.

123 Note that it is genre, not audience that matters here — many of Libanius' orations, as Liebeschuetz, J. H. W. G., Antioch. City and Imperial Administration in the Later Roman Empire (1972), 24–6Google Scholar, reminds us, were meant for small, selected, and at times influential audiences, despite the larger public implied by their generic form.

124 Millar, F., ‘P. Herennius Dexippus: the Greek world and the third-century invasions’, JRS 59 (1969), 12–29Google Scholar, remains the best account; the fragments are edited in FGrH IIA: 452–80.

125 Eun., fr. 1 (Blockley) = 1 (Müller), from the Excerpta de Sententiis 1.

126 Indications of Eunapius' book numbers are similarly problematic: Paschoud, F., ‘Eunapiana’, Bonner Historia-Augusta-Colloquium 1982/1983 (1985), 149–62Google Scholar, reprinted in Paschoud, F., Eunape, Olympiodore, Zosime (2006), 153–94Google Scholar.

127 Stein, F. J., Dexippus et Herodianus rerum scriptores quatenus Thucydidem secuti sunt (1957), 4–65Google Scholar for the Thucydidean linguistic framework, with Blockley, R. C., ‘Dexippus and Priscus and the Thucydidean account of the siege of Plataea’, Phoenix 26 (1972), 18–27CrossRefGoogle Scholar, for detailed analysis of one significant episode.

128 See briefly but comprehensively Barnes, T. D., The Sources of the Historia Augusta (1978), 114–23Google Scholar, building on the important insights of Chalmers, W. R., ‘The Nea Ekdosis of Eunapius’ histories', CQ n.s.3 (1953), 165–70CrossRefGoogle Scholar. Paschoud, F., ‘Quand paru la première édition de l’Histoire d'Eunape?', Bonner Historia-Augusta-Colloquium 1977/1978 (1980), 146–62Google Scholar, reprinted in Paschoud, op. cit. (n. 126), 93–106; Goulet, R., ‘Sur la chronologie de la vie et des oeuvres d'Eunape de Sardes’, JHS 100 (1980), 60–72, at 66CrossRefGoogle Scholar; Blockley, op. cit. (n. 120), 24–5; and Sacks, K. S., ‘The meaning of Eunapius’ history', History and Theory 25 (1985), 52–67, at 64CrossRefGoogle Scholar, do not undermine Barnes' early date, nor is Blockley's tripartite composition for Eunapius' work necessary. Liebeschuetz, J. H. W. G., ‘Pagan historiography and the decline of the Empire’, in Marasco, G. (ed.), Greek and Roman Historiography in Late Antiquity (2003), 177–218, at 181–2Google Scholar correctly states that Eunapius would not have written a history of Theodosius' reign while the emperor was alive.

129 The evidence is usefully compiled in Goffart, W., ‘Zosimus, the first historian of Rome's fall’, American Historical Review (1971), 412–41CrossRefGoogle Scholar, although he attributes to Zosimus what are in fact Eunapian sentiments and probably Eunapian language. Straub, J., Heidnische Geschichtsapologetik in der christlichen Spätantike (1963), 183–93Google Scholar, draws the apt comparison between Orosius and the Historia Augusta and one that holds equally good of Eunapius. Brodka, op. cit. (n. 24), 37 notes how different Ammianus' explanatory framework is from that of Eunapius, Zosimus or Orosius.

130 Though the use of Eunapius in the Historia Augusta, laid out by Barnes, op. cit (n. 128), 120–4, is denied in Paschoud, F., ‘À propos du nouveau livre de T. D. Barnes sur Ammien Marcellin’, Antiquité Tardive 7 (1999), 353–63CrossRefGoogle Scholar, and elsewhere, his analysis of the sources of the Historia, and indeed of the later fourth century in general, is vitiated by his appeal to the lost (and thus reconstructable ad libidinem) Annales of Nicomachus Flavianus, ‘patère à laquelle on peut d'autant mieux accrocher les défroques les plus diverses que cet personage est un fantôme mal localisé d'auteur d'une oeuvre des plus évanescentes’, words which Paschoud, F., ‘Nicomaque Flavien et la connexion byzantine (Pierre le Patrice et Zonaras): à propos du livre récent de Bruno Bleckmann’, Antiquité Tardive 2 (1994), 71–82, at 73CrossRefGoogle Scholar, reprinted in Paschoud, op. cit. (n. 135), 293–316, reserves for Eusebius of Nantes but which better apply to his own historiographical phantom. Since Paschoud is in the habit of alleging linguistic monoculture in barbarous spécialistes anglo-saxons who fail to appreciate the centrality of Flavianus to the history of the fourth century, let it be stated that my own careful reading of Bleckmann, op. cit. (n. 34) — a reading that has included physically placing allegedly comparable passages of different late Greek authors side by side, something with which neither Bleckmann nor Paschoud has deigned to grace their readers — and the varied Paschoudiana published on the topic since, convinces me that Bleckmann has indeed identified a lost fourth-century history, one that certainly covered the third century after the text of Dio ceased and which probably extended to the death of Jovian. This text is not Eunapius and it is not Dexippus, but it is most certainly Greek. Every alleged Latin feature is petitio principii to Flavianus. The whole matter of Flavianus is now treated at devastating length in Cameron, op. cit. (n. 1), 627–90; one can only hope it will finally lay this phantom source to rest.

131 Rosen, op. cit. (n. 80), 86 acknowledges the overlaps and the contrasts between the two authors' arguments, but does not believe that this proves their knowledge of each other's work.

132 His excursus on the Huns and Alans in Amm. Marc. 32.2–3 may be in direct response to Eunapius' botched attempt at describing them: Matthews, op. cit. (n. 11), 337 on Eun., fr. 41.1 (Blockley) = fr. 41 (Müller).

133 Chalmers, W. R., ‘Eunapius, Ammianus Marcellinus, and Zosimus on Julian's Persian expedition’, CQ n.s. 10 (1960), 152–60CrossRefGoogle Scholar, established beyond reasonable doubt the direct connection between the texts of Eunapius and Ammianus in their accounts of Julian's Persian campaign, but he is too quick to dismiss the possibility of Ammianus' direct knowledge of Oribasius' lost Ὑπόμηνα, the solution favoured by Sudhaus, H., De ratione quae intercedat inter Zosimi et Ammiani de bello a Iuliano imperatore cum Persis gesto relationes (1870), 89–102Google Scholar. The Ὑπόμηνα was clearly a personal diary, but such things tended to trickle continuously through the channels of fourth-century amicitia and there is no way to rule out Ammianus' access to a copy that had found its way to Antioch. That said, however, the fact that Ammianus had read Eunapius by the time he wrote the books on the Persian campaign does not mean that he had read Eunapius when composing the Greek original of Book 31.

134 Sacks, op. cit. (n. 128), 56 cites good evidence for a less than wholly favourable approach to Julian in Eunapius, especially with respect to the Persian campaign, and it is not at all implausible to believe that, in this, Ammianus found another point of dispute with the Lydian.

135 Note that Eunapius may in fact have drawn little but his starting date from Dexippus, instead relying on Herodian for style and approach: Giangrande, G., ‘Herodianismen bei Eunapios. Ein Beitrag zur Beleuchtung der imitatio in der späteren Gräzitat,’ Hermes 41 (1956), 328–44Google Scholar; Buck, D. F. ‘Dexippus, Eunapius, Olympiodorus’, Ancient History Bulletin 1(1987), 48–50Google Scholar. Pace Paschoud, op. cit. (n. 126), 190, 199, the argument of Baldini, A., Ricerche sulla storia di Eunapio di Sardi. Problemi di storiografia tardapagana (1984)Google Scholar, does not demonstrate that the first edition of Eunapius began with Augustus, rather than in a.d. 270.

136 Stein, op. cit. (n. 127), 65–71, shows convincingly, at least so far as the limited number of fragments permits, that Dexippus used his Thucydidean model precisely in order to place historical causation on the human, rather than the divine or supernatural, plane. By contrast, Baldwin, B., ‘The language and style of Eunapius’, Byzantinoslavica 51(1990), 1–19Google Scholar on the limits of Thucydidean imitation in Eunapius.

137 See Seyfarth, W., ‘Vom Geschichtsschreiber und seinem Publikum im spätantiken Rom’, Wissenschaftliche Zeitschrift der Universität Rostock, Gesellschafts- und Sprachwissenschaftliche Reihe 18 (1969), 449–55Google Scholar, on the literary attainments Ammianus expected of his audience.

138 I imagine an ageing and bitter Ammianus, his literary career in Rome as much a dead-end as his earlier career as a protector, suddenly confronted with the likelihood that he would die with his great work unfinished. Rather than let that happen, he resurrected an old monograph that could, with tinkering, stand at the end of the final hexad of a history he had started and could not now complete — but that is, of course, to give fancy free rein with the evidence.