P. Vergilius Maro: Aeneis. Bibliotheca scriptorum Graecorum et Romanorum Teubneriana, 2005

We have a range of antique sources for the Aeneid, and the text has been stable in modern editions. What then can we hope for from a new Teubner? Fresh light on problems new and old, with at times persuasive improvement on the current vulgate; a text that is somehow more clearly or more beautifully presented, e.g. with better punctuation; a fuller, more helpful, more accurate apparatus.

Conte’s elegant and witty new Teubner succeeds in some of these aims, but not in all. The text is no more radical or conservative than e.g. Mynors in accepting readings not in the ancient capital manuscripts, though it varies in the places where it deviates from them. The apparatus is accurate, as far as I can judge, and full and clear in recording variants (cases of imprecision are few, e.g. 4.201 ‘ excubiasque R: excubiasque deum recc.’, which leaves us unsure whether R has the diuum present in the text and MP; 12.709 uiros et ] ‘ et om. cx: uirosque y, Prisc.’), and adds to this well selected material from Servius and Donatus and some incisive new comments in lapidary Latin (missing is a reference to the discussion of 11.290 at Sen. Suas. 2.20). There is a vigorous and carefully considered preface, which economically introduces the history of the transmission and of editing the Aeneid; n. 39 has a useful list of tralatitious errors in the citation of manuscripts, and pp. XIX-XXI have wise words on the difficulties of distinguishing the hands behind alterations: sensible use is made of ‘M ^X ’ and the like where certainty could not be reached. And the volume is for the most part attractively as well as accurately set. One verse where I find the particular excellences of the edition displayed is 4.94 (included in a list of places where the Carolingian manuscripts, given renewed attention here, contribute to the text: p. XIII, n. 23). Conte accepts nomen from three early medieval MSS (pin, of which i is cited here for the first time) rather than the usual numen, and repunctuates so the phrase magnum et memorabile nomen is in apposition to the previous sentence and qualified by the following line una dolo diuum si femina uicta duorum est. The key bit of evidence is Ov. Met. 10.607, where Hippomenes says to Atalanta ‘ habebis/ Hippomene uicto magnum et memorabile nomen ’. The apparatus’s reference to Ovid, and to Silius 4.184, is bolstered by explication of the structure.

Yet there are aspects of the book that have not been thought through to the most satisfying conclusion; and there remain questions about how Vergil should be edited, as well as the problems of individual passages. It is on these that the rest of the review will concentrate.

One persistent irritation is that the indication of which MSS are extant at any one point is given only where the available sources change: this means that there is no entry on many openings, and that what we find on page 101 applies only to the final five lines of text. This may be an efficient method for editor and publisher, but it is not for the reader. Secondly, the apparatus is sometime positive, and sometimes not, which makes it very hard to read. We surely should expect positive statements about the ancient variants in an edition of Vergil, which provides such insight into the kinds of errors made in capitals and in a well-studied tradition. But this apparatus is badly inconsistent in itself: at 7.320 we are given a long list of the MSS that read praegnas (printed) or praegnans, but at 331 we find simply ‘labore M : laborum R ’ and have to check the text to find that Vergil wrote laborem, while being left in the dark about which MSS have this reading (presumably V and γ, the P apograph listed above, but what about the other MSS from 9th and 10th centuries?). The implication can hardly be that neither variant is credible, when the same treatment is applied to e.g. concepit (R) at 7.356, no less plausible than the accepted percepit … flammas (M γ): cf. concepit furias at 4.474, furores concipit at 4.501-2. Particularly strange is 5.764, where the apparatus reads ‘rursus] cursus coniecit Ribbeck cl. Aen. 3, 269 ’: why rursus here when the printed reading is so often omitted? why the addition of coniecit ? Conte wants a lively apparatus (p. XXV), but such apparently insignificant inconsistencies delay and confuse the alert reader, who looks for meaning where there seems to be none.

A good use of the apparatus is for pointing out the origins of such errors as Servius’ uoces for questus at 5.780 (in this case the similar line ending effundit pectore uoces at 5.482). Another is the economical presentation of cogent arguments at e.g. 1.426, where Conte explains how the line may function where transmitted (‘spatia definiunt ubi iudicia et munera publica administranda sint’); 3.71, where he mentions Novák’s deletion of et litora complent before arguing for a hysteron-proteron; 10.754, where he explains the superiority of insignis to insidiis; and 12.120, where Servius is cited to explain why limo (‘apron’) is preferred to lino. But in other places the discursive apparatus works less well, e.g. on 10.187-8 Conte comments that Cupavo’s feathers Cycni patris pulchritudinem commemorant, instead of simply saying that insigne formae refers to the father’s transformation into a swan, and the uestrum of crimen, Amor, uestrum is glossed tuum, even though there are no adequate parallels, and ‘a reproach for your family, i.e. you and your mother’ fits Vergil’s style (Harrison (Oxford, 1991), ad loc. cites 1.140, 9.525). At 11.94 praecesserat is glossed with praeterierat. Whether praecesserat is right, or the less well attested but easier processerat (cf. Prop. 4.7.29-30), the sense should surely be ‘had advanced’: the funeral column is put in order from 79 to 88 ( iubet ordine duci, ducitur, ducunt), gradually begins to move ( post … it, 89-90; ferunt, 91), then sets off as a whole ( tum maesta phalanx Teucrique sequuntur, 92). The complete procession accompanies Pallas’ corpse a good way ( omnis longe comitum praecesserat ordo) till Aeneas stops and says his final farewell in 96-8, before returning to the Trojan camp (98-9).

Punctuation has been given serious attention (see p. XXIX); though the results in the text are not always persuasive, the reader is regularly told what alternatives commentators ancient and modern have preferred. For example, at 6.573-4 the apparatus notes the reading mentioned by Servius, which attributes the sentence tum demum … panduntur portae to the narrative voice rather than the Sibyl; brief interruption of the speaker is unusual, but the mix of adverbs and present tense, followed by cernis addressed to Aeneas, as if something had just become visible, makes the break essential. Then Conte mentions Sabbadini’s stronger punctuation at the end of 6.585 uidi et crudelis dantem Salmonea poenas, though he again makes the wrong choice in using a comma: his explanation that dantem poenas means merentem poenas is incompatible with uidi — and with the following dum clause. If 586 dum flammas Iouis et sonitus imitatur Olympi is not an interpolation (so Ladewig), it belongs with what follows. At 7.369-70, omnem equidem sceptris terram quae libera nostris/ dissidet externam reor et sic dicere diuos Conte introduces a single comma before externam, which separates terram from its adjective, and equidem from its verb. At 9.282-3, where he prefers fortuna secunda / aut aduersa cadat to haud aduersa, this involves isolating the preceding tantum as a solitary word on its own, which is scarcely comprehensible even in a pointed text; and in 9.289 a comma follows the parenthesis nox et tua testis / dextera even though that must belong to what follows ( quod nequeam lacrimas perferre parentis) and not the preceding inque salutatam linquo. At 9.462-4 Conte uses a following colon to attach aeratasque acies to the preceding clause. This means he can print in proelia cogit / quisque suos in the next sentence, without emending cogit to cogunt with Cunningham. But this punctuation damages the neat structure of 462-3 Turnus in arma uiros armis circumdatus ipse / suscitat : no further object is looked for here, and in an unpointed text the reader will connect acies with cogit, or rather cogunt, which it seems must have been the original.

My quick count of the number of readings adopted here that lack ancient authority reached 41, of which 2 are transpositions and 13 might be described as orthographical (I omitted deletions). This is no departure from recent editorial practice, and is a reasonable way to edit a text that is better preserved than other classical Latin works; but it should be clear that if our editions are erring, it is likely to be towards conservatism. There are many oddities in our texts of the Aeneid, and it would be good to see too editions that treat Vergil as any other author, and respect the several centuries that separate publication from evidence as a long, and not a short, time. And if editors are not willing to countenance a radical approach in their own texts, they might in the apparatus provide readers with the conjectures that are plausible or diagnostic. Conte uses his space to give rather more than Mynors, but there are still many excellent suggestions not to be found here, e.g.: 1.405 at ante ille Trappes-Lomax ( PCPhS 50 (2004), 152: a fundamental article on hiatus; p. 153 suggests atque post tui in 9.291); 2.121 parent ] paret Madvig; 2.154 uestrum ] Vestae Markland; 2.255 lunae ] noctis Giardina ( QUCC 83 (2006) 93-5); 4.469 Eumenidum ] Euiadum Allen (see E. Courtney, BICS 28 (1981) 22, E. Kraggerud, SO 83 (2008) 60-3); 4.471 scaenis ] Poenis Markland; 7.664- 9 post 749 Warde-Fowler ( Virgil’s ‘Gathering of the Clans’ (2nd ed. Oxford, 1918), 46-50); 8.216 clamore relinqui ] clamare propinqui Peerlkamp; 8.219 exarserat ] exarsit et Dorville; 8.358 del. Pichon; 8.455 tecto ] lecto Wakefield (cf. 461; also Suet. Aug. 73.1 ne toro quidem cubuisse aiunt nisi humili et modice instrato); 9.47 ut ] at Schenkl; 9.214: id del. Hoffmann (so also Mackay, CJ 34 (1938), 171-3; Goold combines this with Powell’s suggestion aut solitas); 9.579 et ] ut Wakefield; 9.671 caelo ] telo Wakefield; 10.366 del. Trappes-Lomax ( CQ 55 (2005), 315- 16); 10.857 quamquam uis ] quamuis dolor Peerlkamp; 12.25 simul hoc animo hauri del. Peerlkamp (see Courtney, BICS 28 (1981) 19); 12.842 excedit ] cedit Feeney ( CQ 32 (1982), 184, n. 33). The reader is implicitly referred to Geymonat’s edition for rejected conjectures (p. XVII), but inevitably not all of this selection are to be found there.

More commonly affecting the text than decisions on accepting conjectures is the choice between possible variants. There is much that is well done, e.g. when at 12.641 Conte follows Traina in accepting P’s nostrum ne against the straightforward order ne nostrum found in MR ωγ, Seru.; and he makes out a good case for rupit in 2.129 rather than rumpit despite the present tense of destinat following (comparing 6.193). But the cult of the lectio difficilior often leads Conte astray. At 9.481-2, e.g., he prints tune ille senectae / sera meae requies (with MP; illa is in R, and some later mss). If this were right, we would surely need a comma after ille to show that ille means ‘the man’ and does not look ahead to a later noun. But we have already had the sense of disbelief this would convey in the mother’s opening words hunc ego te, Euyale, aspicio? Scholars should realise that the only validity of the lectio difficilior principle is to encourage a critic to print the reading that is more likely to have given rise to the other, better expressed through the question utrum in alterum abiturum erat? So at 11.657 it is unlikely that the reading dia has emerged from corruption of the other ancient readings diua and dura, though they can easily have derived from it; moreover, the form is both Homeric and Ennian (see Skutsch on Ann. fr. 60, e.g.). In 9.481, however, it is true that requies might encourage change of ille to illa, but so might the masculine tu have encouraged change in the opposite direction. The proper question then is which is Vergil more likely to have written, or to put it another way, ‘which is the lectio facilior ?’ One is straightforward Latin (Stat. Silu. 3.1.8, Theb. 7.500, Sil. 5.649; and cf. Aen. 12.57-8 spes tu nunc una, senectae / tu requies miserae), the other so odd that P. Hardie in his commentary ad loc. (Cambridge, 1994) cites as a parallel Aen. 2.274-5 quantum mutatus ab illo / Hectore qui …, but omits the last two words, as if they were not basic to the linguistic structure; where no noun or relative is added, there is at least an adjective in other instances (Ov. Met. 1.757-8 ille ego liber,/ ille ferox; Stat. Silu. 2.1.168 tune ille hilaris comisque uideri?).

Again, at 11.229-30 Conte prints alia arma Latinis / quaerenda aut pacem … petendum : there is ancient evidence for this, but MPR all have the straightforward petendam, surely to be printed in view of the parallelism. The argument for petendum is lectio difficilior potior (explicitly so in Gransden (Cambridge, 1991) and Horsfall (Leiden, 2003) ad loc.); but when most of the ancient evidence for the gerund is in commentators, who love oddities, one can see that the anomalous form only has to have been written once to command a position in subsequent texts. Sometimes the arguments utrum in alterum and lectio difficilior may actually point in different directions. At 11.340-1 genus huic materna superbum / nobilitas dabat, incertum de patre ferebat Conte follows other editors in preferring ferebat (MP ^X ωγ, Seru., Tib. [i.e. Donatus]) to ferebant (PRack γ ¹). nobilitas cannot be the subject of ferebat, which must rather be hic understood from huic. The sense ‘he offered an uncertain origin from his father’, with the awkward change of subject, seems rather ‘more difficult’ than ‘men reported it was uncertain about his father’; but I doubt it is therefore to be preferred when ferebat repeats the ending of dabat earlier in the line and is thus more likely to have arisen from ferebant than the other way round. Or what of 11.480? Which of the ancient readings causa mali tanti (so Conte) and causa malis tantis is more likely to have been corrupted into the other? Conte presumably regards the dative as an interpolation designed to remove the hiatus that follows tanti. That is possible, but seems less likely than that Vergil did not compose the line with the hiatus in the first place, and that the less familiar, but idiomatic dative (3.305, e.g.) has been corrupted into the genitive, through reminscence of causa mali tanti at 6.93.

Finally interpolations. A number of times I am not persuaded by Conte’s decisions in this area. For example, at 1.380 Italiam quaero patriam et genus ab Ioue summo, Conte himself brackets the last five words as a reader’s addition drawing on 6.123 et mi genus ab Ioue summo, where the expression is certainly far more pointed. If we punctuate as Mynors does, with a comma before et, ‘this is presumably a new statement … in which Aeneas claims descent from Iuppiter’: so Austin (Oxford, 1971), ad loc., who goes on to argue that the phrase implies the kinsmen in Italy, also descended from Dardanus, but I wonder if it might be better to take it of the future race Aeneas seeks to establish: compare the request to the oracle of Apollo at 3.85-6 da moenia fessis / et genus et mansuram urbem (also 1.6, 4.230, 5.737; and the way that genus is put parallel to territory also at 1.339). He is descended from Jupiter, and his race will be; but he needs to find a home, a wife and a people to match that descent. Also deleted is verse 4.126 conubio iungam stabili propriamque dicabo. It is true that the line repeats 1.73, where propriam is more pointed, but Dido can be understood as the object here; the sense of her being ascribed to Venus’s son is pointed as an expression of divine power and indifference; and the repetition reprises Juno’s etymologically determined role as the goddess of marriage, which makes the offer hard to decline and gives the wedding divine authority. In any case, such a line may be Vergil’s stop gap, not a later reader’s embellishment. We may contrast how Conte handles 5.777-8, which repeat 3.130 ( prosequitur surgens a puppi uentus euntis) and 3.290 ( certatim socii feriunt mare et aequora uerrunt): given that the mss are split on the order, the chances are very high that one is a marginal addition, and even if Vergil had intended to imitate Od. 12.144-52, where oars are used to leave harbour until a fair wind is sent by Circe, it is unlikely that he would in the end have used these unintegrated lines to do it. Moreover, this hardly fits with the following paragraphs, which have Venus appealing to Neptune for a safe passage and his granting her request, at which any signs of storm disappear (779-826); only then does Aeneas order masts to be raised, sails to be set (827-32; n.b. sua flamina at 832). Announcement of a following wind at 777 is thus doubly awkward: Günther’s deletion should be adopted. Not even mentioned are a couple of obvious interpolations. At 11.309 the words spes sibi quisque; sed haec quam angusta uidetis (deleted by Burgess and Porson) intrude awkwardly between Latinus’ dismissal of any hopes in help from Diomedes, and the cetera which introduces other aspects of the situation. As a clinching argument there is a metrical anomaly: Vergil offers no parallel for the final e of ponite treated as short before the double consonant at the start of spes. The halfline Euryali et Nisi (9.467) was deleted by Heyne: the transmission puts the phrase in the wrong clause and there is obvious pressure for a gloss on capita in the preceding verse praefigunt capita et multo clamore sequuntur.