The Dutch Ammianus commentary is a glorious example of collaborative scholarship. Three of the quadriga Batavorum have been working together since the commentary on Book 20 in 1987; the fourth, Drijvers, has been on the team since Book 22 in 1995. With the three original authors in retirement, the frequency of volumes has increased and is now regularly biennial. It is only four years since my review of Books 25 and 26 in JRS 2009, and it is not unlikely that Book 29 will beat this review into press and that the two remaining books will be achieved by 2017.

Before any disagreements uttered in this review, it should be said that the achievement is magnificent, a model of linguistic, literary, and historical learning; this work will be consulted with profit for generations. And before a review focusing mainly on chronology and textual criticism, it should be emphasized that the authors’ coverage is wide-ranging — from the nuances of Latin particles through subtleties of characterization to detailed questions of topography — and the bibliography comprehensive.

Book 26 described the accessions of the brothers Valentinian and Valens in February and March 364 and their subsequent division of the empire, going down to Valens’ suppression of the eastern usurper Procopius in May 366. It also introduced a new narrative principle (26.5.15): that to avoid confusion in readers the organization would be geographical, rather than leaping from place to place to preserve chronological precision. This principle (in which many later historians’ narratives of these reigns have followed Ammianus, including Gibbon, Seeck, and Blockley in CAH XIII) does not greatly affect the reader in Book 26, but Books 27 and 28 see it fully in action. Previously the actions of emperors or campaigns have been described year by year, but Book 27 focuses on events starting roughly between a.d. 365 and 368, including inter alia the German campaigns of Valentinian's generals in a.d. 365–366 (1–2) and Valentinian himself in a.d. 368 (10), Valentinian's promotion of his eight-year-old son Gratian as a third Augustus in a.d. 367, along with some criticisms of Valentinian's cruelty (6–7), Valens’ war on the Goths from a.d. 367 to the treaty in early 370 (5), a sketch of Petronius Probus as praetorian prefect of Illyricum (no chronological indications in the text, but he was in office from a.d. 368 to 375/6) (11), and events in Armenia from a.d. 367 to 370 (12). It is hard to overstate how much this differs from the pattern of previous extant books. In Book 28, narrative blocs cover a still wider temporal expanse. Though the heart of the book treats campaigns of a.d. 369 and 370 (28.2, 3, 5), 28.1 describes the trials of Roman senators for magic and adultery between about a.d. 369 and 374, with a flash forward to the punishment of the prosecutors in a.d. 376, the year after Valentinian's death, which brings a formal end to Ammianus’ coverage of western events; 28.6 describes the travails of the province of Tripolitania from barbarian attacks and the corruption of the military who failed to protect them, a sequence of events beginning as early as a.d. 363 and again with repercussions well after Valentinian's death.

Chronology, then, is the largest single problem in these books, and is given fifteen or so pages in each introduction as well as copious discussion ad loc. On the whole, the commentators show exemplary good sense and clarity, balancing the evidence of Ammianus against that from other authors and from dated constitutions in the Theodosian Code. Good examples are the painstaking examination of the end of the Gothic war in 27.5, Theodosius’ British campaigns in 27.8 and 28.3, and Roman and Persian interactions with Armenia in 27.12; in the latter they engage with the Armenian historiographical tradition and use the new chronology that Noel Lenski set out in the authors’ edited book Ammianus after Julian (2007). In a few places, they can be mildly corrected. In 28.6.30, they place the final fizzling out at Milan of the legal battle between the province of Tripolitania and the comes Romanus at a time after Gratian's court moved there from Trier in a.d. 379. They are surely fundamentally right in arguing for a late date and an extended process — but in fact the court did not move to Milan until a.d. 381 (see Barnes in Ant. Tard. 7 (1999), not cited). The most difficult section in chronological terms is certainly 28.1, the Roman trials, instigated by the odious upstart Maximinus as prefect of the annona and vicarius of Rome and continued under subsequent vicarii when Maximinus had become praetorian prefect of Gaul. Their thorough treatment of the chronology on the whole follows Barnes, who demonstrated that most of the perceived confusions in Ammianus’ account arise from a misdating of Maximinus’ promotion to prefect.

Other questions surround the beginning of the trials, and their end. At 28.1.1, Ammianus dates the trials anno sexto decimo et eo diutius post Nepotiani exitium: the bloodshed associated with the killing in Rome of the usurper Nepotianus in June 350 had been the last major disaster to befall the Roman aristocracy. The sixteenth year would be a.d. 365/366, but all the other indications in Ammianus’ text and outside it point to c. 369/70. It is a pity that they do not give more serious consideration to Barnes’ suggestion of emending sexto decimo to uicesimo, 16 to 20. Their reluctance is perhaps understandable, as Barnes’ solution seems drastic, and Ammianus is certainly capable of errors in chronology (the worst by far, well-illustrated by the commentators, at 27.7.1). However, his text is also capable of serious corruption, as they demonstrate elsewhere, and if numerals were used in the transmission, for which there is evidence, xx and xvi could easily be confused (Barnes also offers xxi as a possibility). The overall sense must be ‘in the nth year after Nepotianus’ death and lasting beyond it’, which works far better if n = 20, since the chapter describes events from a.d. 369 to the mid-370s: et eo diutius is not, as implied on p. xvi and ad 28.1.1, a cover against possible criticism (is the suggestion that Ammianus gave a precise chronological indicator but suspected it was wrong?).

The date of the last trials, those of Aginatius and Anepsia under the vicarius Doryphorianus, is debated. Ammianus’ narrative clearly implies that Doryphorianus entered office and that the executions took place before the death of Valentinian on 17 November 375: since his predecessor Simplicius is attested in office on 23 March 374, the date must lie between those termini. The commentators point to a letter of the emperor Gratian from a.d. 379 (Collectio Avellana 13.3) which refers to an earlier letter he had written to Simplicius as vicarius, who they argue must have remained in office after Valentinian's death. However, since Gratian had been Augustus since a.d. 367, it could have been written under his father's authority but included his name in the heading. The commentators claim ad 1.53 that ‘when citing constitutions issued when he was a minor member of the imperial college, [Gratian] attributed these explicitly to his father’ (they cite CTh 1.6.8, 16.6.2, and 16.7.3) and conclude that the final trials belong after Valentinian's death. However, all of these citations come in lists of earlier legislation, and it is not hard to find counter examples: CTh 10.19.8 (1 March 376) and 16.5.4 (probably 18 April 376) are constitutions from very soon after Valentinian's death in which Gratian refers back to previous legislation using the first person plural, and though that legislation is lost, chronology means that it should belong to his father's reign. So there is no reason to doubt Ammianus’ implications that the trials belonged exclusively in Valentinian's reign — and indeed Ammianus would be guilty either of serious error or an extraordinary and wilful deceit if the authors’ chronology were correct on this point.

The most unequivocally successful aspects of the commentaries are philological: in explaining usage, in detailing intertextuality, in exploring the nuances of pronouns they cannot be bettered. There are many fresh observations, including at 28.4.21 the fact that editors have printed a sentence with no main verb, simply two present participles: perhaps an authorial error? I turn to their textual choices. As in the previous volumes, Den Boeft et al. diverge frequently from the standard Teubner edition of Seyfarth from which they take their lemmata. I counted over sixty divergences, excluding patently corrupt and lacunose passages where they reject overly optimistic attempts at rescue (there is a marked increase in such passages in Book 28). At only three points, by my count, do they vindicate the manuscript reading of the Vaticanus against other readings printed by Seyfarth (27.1.2, 28.2.4, 28.4.28); at another dozen they argue for readings of Gelenius’ edition of 1533, which may represent either the readings of the lost Hersfeldensis or simply his conjectural acumen. In just over forty they argue for the conjectures of others (ten by Petschenig, six by Henri de Valois), and they make about ten conjectures of their own (personally I would alter his text still further). In half a dozen or so cases where they disagree with Seyfarth, Ammianus’ prose rhythm, which is remarkably regular, is mentioned as favouring their change, but in another half dozen cases, they do not mention the fact that their solutions repair the rhythm. At 27.7.7 their solution breaks the cursus, but justifiably, given Ammianus’ practice in pithy excerpts of direct speech. There are also places where cursus should have been taken into account and was not: at 27.4.10 in favour of Clark's <de>fluentem; at 27.7.9 perhaps tipping the balance in favour of Adrien de Valois’ efficere rather than Madvig's effici; at 28.1.37 as an obstacle to their proposed punctuation. Whereas some of their disagreements attest Seyfarth's perverse conservatism more than their good judgement, there are countless astute choices and some outstanding conjectures: at 28.1.22 tutus for V's tectus, while rescuing the ms reading tectius a line before; at 28.1.47 coartato for V's contracto makes lurid sense of a Roman matron's suicide by self-suffocation. Of course, my focus on emendation does not mean that they do not just as often explain the unexplained: for example by identifying eiusdem in 28.1.27 as Lollianus mentioned in 28.1.26 (the two sentences therefore should form a single paragraph). I read through the commentaries while writing a translation of the two books, and can rarely remember learning as much about Latin in as short a time.

A few minor corrigenda. 27.3.9: Gelenius’ reading is not fremitu but fremituque; 27.3.15: lemma and commentary have been accidentally duplicated from 27.4.14; 27.5.9: Augustus’ grandson Gaius Caesar is confused with his namesake and nephew the emperor Caligula; 27.6.2: the emperor Gratian is better described as ‘assassinated’ than ‘executed’; 27.12.2: the praetorian prefect ‘Sallustius’ (or to be precise, Saloustios) described in John Lydus, Mag. 3.51.6–52.4 should have been identified with Saturninius Secundus Salutius; 28.2.10: the villa Murocincta, normally identified as Parndorf near Vienna, is certainly nowhere near Sirmium. The authors probably assume that readers will have a critical text, but if they do not, they will not realize that at 27.2.6 insueta is the reading of Accursius and Gelenius, not C. F. W. Müller's conjecture; at 28.2.4 His is not added in Gelenius’ edition but is a conjecture by Müller; and at 28.1.38 Valesius’ conjecture implacabilitate is anticipated by the scribe of manuscript E.