This is a bold book, full of wisdom and critical sharpness, and deserves to be read carefully not only by editors and commentators, but by all serious students of Vergil. Kraggerud has collected 109 of his notes on problems of text, interpretation and punctuation, revising some and including 29 new loci. Introductory pages describe his evolution into a textual critic as he saw how literary and philological arguments could lead to texts that seemed superior to those in the published editions. He also became aware of how corrupt the revered capital manuscripts are, a point nicely illustrated on pp. 6-7 with a list of the 23 substantive errors to be found in M, P and R, individually and together, within one passage of 101 lines (Aeneid 5.500-600): ‘the ancient paradosis is too … arbitrary to serve as the sole basis for the text’ (xv). He is therefore inclined to think that there may be ‘80, perhaps even 100’ places in need of conjectural emendation (7).
It has been said that no one agrees with a radical textual critic, the conservatives because they know that any changes to the familiar text are wrong, other radicals because they regard these changes as wrong. Having indicated my approval for Kraggerud’s approach, and commended the methodological underpinning with which he opens the book, let me turn to the details — and the disagreements. In fact there are an impressive number of places where, on first reading at least, I find the notes persuasive (a) on matters of interpretation or punctuation: e.g. Ecl. 5.65-6 (interpretation of aras … altaria); Aen. 2.433 (uices Danaum), 6.153 (pecudes, prima), 6.822 (full stop after infelix, comma after minores), 9.141 (no comma after satis), 9.391-2 (question mark after siluae not sequar), 12.161 (ingenti mole, ‘with a huge multitude’, belongs before the comma); and (b) on textual decisions: Ecl. 2.32 (primus), 4.28-9 (paradosis), 5.3 (considimus), 5.8 (certet), 5.38 (purpurea); Geo. 1.35 (relinquit, as at 3.519, Aen. 12.470, and perhaps 6.716), 1.36 (sperent), 1.500 (nunc*)1, 3.303-4 (dum; but there is no vagueness in extremo … anno — ‘year’s end’ means the second half of February); Aen. 1.380 (retaining et genus ab Ioue summo, deleted by Conte),2 1.458 (Atriden), 1.604 (iustitiae), 2.121 (paret), 2.598 (omnes), 4.176 (initu), 4.469-73 (Euiadum, and scaenis, ‘stage-buildings’, taken as an allusion to Euripides, Orestes), 5.300 (Panopeusque, reading of the 8th-century fragment p, which Kraggerud rightly values as a source of neglected readings), 6.293 (cauae), 6.561 (auris; but the claim that ‘ad auras is found only twice in Vergil’ is a bizarre error, for ‘16 times’), 6.615 (-que*), 6.659 (siluas*), 6.761 (luci), 6.846 (restitues), 7.129 (exiliis), 7.741 (paradosis), 9.85 (retained), 9.130 (exspectans), 9.363 (deleted by Wagner), 9.471 (uidebant, taking simul as simul ac), 9.539 (Schrader’s recedunt), 9.709, 733 (clipeus twice*), 9.789 (pugna), 11.256 (deletion of ea*), 12.648 (nescia), 12.790 (certamine). Also valuable are notes that present a serious case for an alternative to the vulgate text, such as Ecl. 6.74 (ut), Aen. 6.852 (mores), 9.243 (fallet), 9.599 (marti).
However, there are many notes that are not convincing for various reasons, though many even of these will be useful in advancing critical debate. I begin with a quick list of unpersuasive notes on the Aeneid where I have little to say: the repunctuations of 1.1-7 and 6.791, uidet at 2.485 (those inside the palace, evoked by 483-4, see Pyrrhus and his soldiers in limine primo, as at 469), avoidance of the archaic orthography ni (for ne*) at 3.686, Tyrias … res spectat* at 4.224-5 (Housman’s Hesperiam gives better sense and is hardly more distant palaeographically), amissae classis* at 4.375 (the paradosis means ‘I restored [to you] the squadron of ships you had lost, and brought back your comrades from [imminent] death’: a syllepsis rather than a zeugma), 5.851 (et ‘even though’ is easier than the awkward enjambment of fallacibus auris | et caelo together with the separation of caelo and sereni = sereni caeli), 6.466 (‘this is the reason why …’ does not do justice to the stress on extremum, or work in context), 7.377 (without anything to limit the reference to Latium immensum … per orbem would mean ‘over the whole immense globe’), 7.598 (rapta is ingenious, but robs nam of its function), 9.402-3 (altam et lunam supposes an easier corruption than torquet, and aptly links suspiciens with precatur), 9.464 (the argument for quisque suas ignores the principle utrum in alterum abiturum erat? and the placement of the words, which encourages an appositional reading, as at 12.525, Accius, Carm. 3.5), 10.366 (quando is confusing and wasted: I suspect it has replaced an adjective, e.g. aquosi, or Courtney’s uanos [BICS 46 (2002-3), 193]), 10.705 (Ellis’s creat: Paris instead of Bentley’s Parin: Paris leaves quem, i.e. Mimas, as the object of creat: the parallelism of the clauses requires an object, and not a verb), 12.218 (se uiribus aequos*; the sense is not incomplete if we delete non uiribus aequis, and punctuate with a comma at the end of 217, as Conte and Tarrant do). At Georgics 3.159 Kraggerud fails to consider the run of thought or to explain how he thinks ecquos* is an improvement on et quis: he has fallen for once into the palaeographical trap.
To end, detailed comment on a few passages.
At Ecl. 1.69 the asyndeton patrios longo post tempore finis | … et … culmen | post aliquot mea regna uidens mirabor aristas is awkward; together with the ambiguity of post before an accusative this renders the supposed resumption of longo post tempore hard to follow. Kraggerud therefore prefers to take post as a preposition meaning ‘behind’, comparing 3.20 tu post carecta latebas. However, latebas there clarifies the usage, whereas neither uidens nor mirabor encourages such a reading. It is surprising that commentators do not regularly adduce verses 29-30, where longo post tempore is followed by postquam: this may provide some support to the resumptive reading. (There is an odd error in this note, when it is said on p. 11 that Courtney ‘rejects post … and reads instead post ah!’; presumably ‘reads ah! instead of post’ is meant. In general the proof reading has been carelessly done, e.g. subsistit for substitit at 2.739, Cereis and curia in the quotation of Aen. 2.741-51, vertitur for volvitur twice on 234, Latin semi-colons in the citations of Greek text, but I did not spot anything else likely to cause problems.)
The note on Ecl. 3.60-3 prints Ab Ioue principium musae and claims ‘musae = carminis’ (81); the analysis thus falls far short of the author’s desire (4) for a thorough analysis of each context. For if we take Musae as a vocative plural, we have a passage that reworks the structure of Theocritus 5.80-2. As Kraggerud says, ‘The Muses and Apollo constitute a group of inspiring deities belonging closely together.’ Damoetas assumes the support of the Muses in invoking them to begin from Jove; and Menalcas adds Apollo (Et me Phoebus amat), just as Lacon does in the Idyll: there is no competition of Jupiter against Apollo, and not the slightest reason to read At* for Et. At Ecl. 4.8 quo is undeniably difficult, but Kraggerud’s cum/quom* is no solution: he provides no translation and fails to show how the temporal clause works: the rest of paragraph stresses that the new age is being born now (iam … nascitur … iam … iam … modo … iam), so it is barely comprehensible for the birth of the child to be put into the future. Perhaps editors might consider cui/quoi (so Burman at 1.20); for the dative after desino cf. 8.11 tibi desinam.
Discussion of Ecl. 6.1-12 begins with the arbitrary addition of ‘Roman’ (35) to enable Prima to mean ‘the first to’, rather than the unusual but paralleled ‘at first’, which is shown to be correct not only by the sequence in this passage (which reflects the development of the book to this point, and which Kraggerud tries to break by introducing a new paragraph after verse 5), but also by the matching Extremum at the start of poem 10. Vergil wanted primus as the first word of the second half of his book, but as he preferred to avoid a spondaic word at the start of a line, and never begins a poem or a book with one, naturally avoided primum. Some of the material later in the note is useful, however, in explicating tamen in 9, e.g., and exploring evocations of Theocritus 16.
At Ecl. 6.33-4 Kraggerud sees that if you accept his ex omnia primis, rather than the better transmitted his exordia primis, the subsequent epanalepsis of omnia is pointless. He therefore follows Kirsch in extracting omnis from P’s omnisa in 34 omnis et ipse tener mundi concreuerit orbis; but this creates another problem, the incoherence between the repeated forms of omnis, one of which is the subject (‘everything’) and the other a consequently ineffectual predicative ‘in its entirety’. It is time to return to the inoffensive exordia. Though I am persuaded that Vergil wrote te, not me, at Ecl. 10.44, I cannot follow Kraggerud to his further conjecture inermem*, which seems an ineffectual epithet to apply to an elegiac puella. It is erotic love, not desire for war that has taken her away: one should take duri Martis not with insanus amor but with in armis (so Hollis, FRP 141, in supporting Heumann’s te).
The discussion of fors et at Aeneid 2.139 makes no mention of several other instances of the combination (e.g. 11.50, which gets its own note, Horace, Carm. 1.28.31, Propertius 2.9.1, Statius, Siluae 5.3.62). The et could be pushed (with difficulty in the Horatian and Propertian cases) to have independent value, but such a collection of examples suggests that OLD s.v. fors2 is right to imply that the et is part of the usage. We might conceive that each of these was corrupted from an original forsit*, as Kraggerud suggests; but it seems more likely that the different manuscript traditions are right.
The note on the Helen interpolation (2.567-88) argues that there is no need to mark a lacuna. Given that we do not mark a lacuna wherever there is a half-line, this may be logical, but he does not produce the evidence to show that Vergil intended no change here: in the four passages cited from Cicero and Statius where inverted cum follows preterite verbs, the cum is always followed immediately by repente or statim (so cum subito at Aen. 1.535); there is no equivalent in 2.589. The case for expunction of 2.749 is thin, essentially based on the hysteron proteron urbem repeto et cingor armis, but without mention of the parallel at 11.535-6 graditur bellum ad crudele Camilla, | o uirgo, et nostris nequiquam cingitur armis.
The discussion of Aen. 6.460 inuitus, regina, tuo de litore cessi is out of the normal run, concerned not with philological issues, but with what the author regards as misguided over-interpretation of the imitation of Catullus 66.39: ‘The original context if remembered should exert no power to change the meaning of Vergil’s own context’ (215). As it happens, I largely agree with him in his reading of the passage and the relationship with the Catullian verse (Vergil has subverted a comic line and made it tragic), but the principle is fundamentally misguided: if intertexts cannot affect meaning, the foundations of meaning are severely limited. How are dictionaries constructed?
1. I mark Kraggerud’s own conjectures with an asterisk.
2. I came to the same view, with an alternative interpretation, in my review of Conte, BMCR 2010.10.03, where I also argued for retention of Aen. 4.126, excluded by Kraggerud, and, like him, for a strong pause after 6.585, not 586, and illa at 9.481. I missed Schaper’s caelo, printed by Conte at 7.543, and now supported by Kraggerud (wrongly, I fancy: no example is offered of the dative after conuersa, and caeli … per auras, ‘through the air of heaven’ marks the ascent, as it marks the descent from heaven at 11.595).