BMCR 2004.01.13

Ovidius. Carmina Amatoria. Bibliotheca Teubneriana

, , Carmina amatoria : Amores ; Medicamina faciei femineae ; Ars amatoria ; Remedia amori. Bibliotheca scriptorum graecorum et romanorum Teubneriana. München/Leipzig: Saur, 2003. xxxvi, 374 pages ; 21 cm.. ISBN 3598712316. EUR 74.00.

This is the first edition of these poems to appear in Bibliotheca Teubneriana since Ehwald’s (1880, 1910), and the first with a full apparatus criticus. The editor (henceforward R.) lists more than 300 MSS of the amatoria in an appendix which has benefited from the expert assistance of Professor Michael Reeve, with a bibliography of the secondary literature on them. Those selected as the basis of his text and enumerated in the Praefatio he divides into three classes: 1. The antiquiores (10). 2. A selection of recentiores, mostly 12th/13th-century (42). 3. Spanish MSS, mostly 15th-century (8). To these are added florilegia and fragments from the 9th to the 14th century (23); the various excerpts from the notes of scholars reported by Heinsius (5); the Copenhagen scholia; and (? ps.)-Planudes. Nothing is said in the Praefatio about the history of the transmission, which S. regards as satisfactorily covered by others, duly acknowledged together with a list of editions deemed to be of special significance. Heinsius, as is only proper, receives a special tribute, and the Praefatio concludes with a quotation from him to serve as a disarming captatio benevolentiae. There follows a bibliography of items cited in the apparatus by short title. The book is rounded off by serviceable indexes of amatory vocabulary and proper names.

Munari in his edition of the Amores listed 78 MSS,1 of which he cited eight regularly, and nine of the l3th/14th centuries and 25 of the l6/17th sporadically. Even then the results confirmed that, though, as Heinsius had shown, a minority of the later MSS not infrequently offer a reading more likely to be what Ovid wrote than that of the antiquiores, the law of diminishing returns now prevailed, and that the benefit to his text of any attempt at an exhaustive examination of the sources was likely to be out of all proportion to the effort involved. The unique licenda of the Hamiltonensis at Am. 1.10.30 is a warning against dogmatism here; and it is interesting that two of R.’s Spanish MSS, one of the 14th, the other of the 15th century, offer the generally accepted ipse at AA 2.276, hitherto attested only by Heinsius’ lost Argentinensis. Nevertheless it seems clear in the light of R.’s labours that any further improvements to the text of these poems may be expected to be achieved ingenii rather than codicum ope.

It might therefore have been reasonably hoped that an edition on the scale of this one would provide students and scholars with a soundly based, accurately documented, and judiciously edited text which could hold the field as the basis for further work for a few decades, if not for another century. It is then disappointing that I find myself obliged to report that the great and praiseworthy industry which R. has brought to his task, informed with what is clearly a devotion to his author, is not matched by the quality of the result. His critical judgement is erratic and his grasp of Ovidian idiom sometimes uncertain; and he has too often accepted uncritically the judgements of other scholars, particularly Heinsius and Goold,2 without weighing them properly. It is, as I have said, entirely right and proper that the name of Heinsius should figure prominently in an edition of Ovid; but it serves no useful purpose to cite him routinely (or indeed any other editor) as being the first to introduce a reading other than a conjecture of his own into the text, and in many cases the reading so cited turns out to have been the pre-Heinsian vulgate. In particular R. has too often allowed himself to be carried away by the great man’s ‘analogist assumptions’, as Tarrant has termed them.3 I will illustrate these criticisms in my detailed comments. Two general points in addition: R. is excessively fond of the exclamation mark, which as a rule is best used only after an exclamatory particle; and though the doubts expressed by critics of the authenticity of certain passages are duly reported in the apparatus, I find the total absence of the obelus from his text too Panglossian for my taste. Others may not be troubled.

However, the very full apparatus criticus with which the edition is equipped at least provides users with a D.I.Y. kit to construct their own. R. could have made that task a good deal easier. In the first place, besides the superfluous citations of earlier editors mentioned above, the notes are bulked out unnecessarily by the use of the cumbrous ‘recc.’ and ‘recc. aliquot’ in place of the now generally accepted symbols ω (omega) and ς (final sigma), five and ten keystrokes respectively instead of one, taking up space which could have been more usefully used to give page references to the secondary literature, a particularly irritating omission entailing much waste of time by the conscientious enquirer. The clutter is further compounded by the illogical practice of citing the Spanish MSS separately instead of including them with the other recentiores. There are a number of places where text and note do not agree.4 A particularly unhelpful proceeding is that of dividing conjectures which embrace more than one word under separate lemmata, making it troublesome or sometimes impossible to reconstruct them. Again and again the passages purportedly supporting a reading merely go to show that it is good or possible Latin, not that it is preferable to the alternative(s). There are also a good many inaccuracies. In what follows I may appear to lay too much emphasis on what I see as the failings of this edition to the neglect of its positive qualities. Well, as Housman said, ‘finding faults, if they are real and not imaginary, is the most useful sort of criticism’, but that might strike some as an evasion. It is true that the number of places where R. prints a reading that is clearly correct or, though not what has been generally preferred or what I should print myself, is at least defensible, outnumber those where I consider his choice impossible to defend; but in texts that have for the most part not been gravely corrupted in transmission and have been worked over as intensively as these by scholars, of an author who writes perspicuously and whose usage has been well studied and documented, that is faint praise. We have a right to expect something better than what R. has given us. I hope also that the passages which I have chosen to discuss may be of some exemplary and admonitory value. They are the survivors of an originally much longer list; and I have noted only a small sample of the places where the apparatus criticus leaves something to be desired.

Amores I.

Epigr. 3 ‘iam] non Naugerius‘. As it stands this makes no sense; what Naugerius proposed was ut non ulla tibi iam sit legisse uoluptas.

1.10 Virginis (Goold). This raises a point of some general interest. In this instance capitalization is appropriate when the word is used as a recognised proper name, as of the Aqua Virgo ( AA 3.385) or the sign ( AA 3.388). Here, where the epithet pharetrata identifies the goddess, it is fussy and superfluous. R. I think overdoes capitalization with sometimes odd results, as with Puer, Pueri for Cupid ( Rem. 435, 701). There are borderline cases, as at Rem. 395-6 Elegi … Epos (Heinsius, Muretus); whereas at Rem. 487 Artes (Henderson) is defensible, at 385-6 Burman’s Arte seems to me to blur a crucial ambiguity. The problem, born of the imposition of alien conventions on texts innocent of such graphic distinctions, is in principle insoluble.

4.16 It is not helpful to report Roverius’ ut accumbas without also reporting the repunctuation entailed. Similarly at 4.50 ‘detecto Riese‘ is guaranteed to puzzle; what he printed was detecto … toro.

7.1 ‘ non dist. Hall 1999‘. Hall’s point was that editorial brackets are an expedient to save what otherwise ‘would be awkwardly ambiguous’; his solution was to print meas … manus. Similarly at 7.48 ‘ post opem interrogationis signum posuit Kenney‘ is literally true but hopelessly puzzling as it stands.

8.36 R. prints prosit ut adueniens, en aspice, diues amator etc., and notes ‘ sic dist. Naugerius, Madvig 1873‘, which is false. Naugerius placed a colon after aspice, Heinsius a full stop, Madvig an exclamation mark, noting ‘Effectus hic est Veneris advenientis’. R.’s punctuation obscures the sense.

8.65 ‘ueteris y, recc., Marius, prob. Müller‘. What Marius read was ueteris quinquatria cerae in his text, quamque atria in his notes.5 What Müller read according to the accurate Munari was ueteris plena atria cerae.

10.13 animam Luck; not one of his happier inspirations. ‘Disposition’, not ‘soul’, as the variation mentis … figura in the next verse shows, is what the sense requires; the antithesis animus )( corpus is illustrated in McKeown’s note.

11.28 ‘Veneri … acer titulo inclusit Némethy : at … acer titulo non inclusit Merkel‘. In fact it seems to have been Jahn who first limited the extent of the inscription unambiguously to Veneris … dedicat.

13.11-14 The apparatus is in a thorough muddle here. (i) Riese’s comment that Aurora is addressed in the second person throughout the poem referred to lines 33-4 not this passage. (ii) The explanation of the omission of these verses by PYS as due to homoeteleuton, taken from McKeown and due originally to Ehwald, is pertinent only if they fell out after line 18, which is not made clear. (iii) The report of their whereabouts in the margin of P is duplicated and compounded by a misprint. (iv) The vulgate reading of line 14 miles et armiferas is ascribed to ‘PY, recc.‘. (v) R.’s comment on the reading et miles saeuas as ‘ ex u. 18‘ makes no sense. Incidentally I cannot imagine what Bentley was thinking of when he proposed et miles seras; serus is hardly the right word for stand-to at first light, and Ex P. 3.4.62 is quite irrelevant.

14.21 The credit for Thracia is not mine; it was the vulgate until Heinsius ‘corrected’ it. Similarly at line 24, where mala was installed in the text by Marius and was the vulgate until male was reinstalled by Müller or Riese.

Amores II.

2.61 I suppose it is R.’s predilection for the exclamation mark that has led him to prefer quam from the excerpts of Juretus and Politian to the transmitted quid, which is faultless.

4.46 corporis … sapit is supported against moribus … placet by wholly irrelevant parallels. Ovid is not interested in girls with their heads screwed on; he likes them complaisant.

5.29 The passages cited in support of differs do indeed exemplify the iunctura; they have no bearing on the sense.

6.21 Pliny presumably had ancient authority for the belief that emeralds are brittle, and it is in character for Ovid to prefer the mildly recherché fragiles to the banal virides. That the belief was erroneous is neither here nor there: Lucretius, echoed by Ovid, on the behaviour of sling-stones should remind us that poets do not always trouble themselves about technical accuracy. Has it not been suggested that Tennyson thought that railway trains ran in grooves?

9.36 hac means, as Booth and McKeown point out, ‘on my side’, precisely what Ovid is complaining Cupid is not.

11.25 By following Heinsius in printing qua R. sacrifices syntax and sense to transcriptional probability ( qua RY). After ubi temporal in line 23 the asyndeton with what follows dictates that we must have another temporal conjunction, viz. ubi, here.

17.19 Reading Vulcano Venerem introduces enjambment between this and the preceding couplet. This runs counter to Ovid’s fixed practice of presenting such lists of exempla in self-contained distichs.6

18.1 It is peculiarly perverse to print Achillen here with Heinsius, while at the same time citing Housman’s conclusive demonstration that it is impossible Latin, and then go on to print Achillem at Am. 3.9.1, AA 1.11, in both cases without comment.

Amores III.

1.51, 7.81 recincta is preferred to the better-attested soluta; the fact that Ovid uses solutus in this sense only here in these early poems is not a reason for believing that interpolation has been at work. There is nothing wrong with soluta ( OLD s.v. 6a) — and to quote a maxim which has repeatedly occurred to me while weighing R.’s critical choices, utrum in alterum abiturum erat?

1.53 R.’s parallels for incisa, which all refer to carved inscriptions, destroy the case for printing it here.

3.14 There is no case for preferring en (y ς) to et (PYS ς). The parallels are irrelevant: et introducing an indignant question or comment is well attested and characteristic of Ovid ( OLD s.v. 15a, Bömer on Met. 9.203-4). Cf. line 33 of this very elegy, and see below on AA 3.241.

4.8 Here R. has McKeown on his side in reading exclusis, but after omnia in line 7 omnibus can refer only to means of access, not to who or what is excluded. (Nisbet’s ostia gets round that difficulty, but at the expense of the rhetoric.) Pace Timpanaro, whose discussion R. does not mention,7 the fact that occludo does not otherwise occur in Augustan literature does not mean that Ovid would have avoided using it if it was the mot juste, as it is here.8 Indeed, Timpanaro in the end drew back from a definite verdict.

6.46 pomifera, besides being poorly supported, would be the only instance in Ovid of the elision of -a at this place in the verse (M. Platnauer, Latin elegiac verse, Cambridge 1951, 89). It is unsound method to introduce a metrical anomaly not imperatively demanded by the sense.

7.55 Not, I think, one of the places where reluctance to obelize can be tolerated, at least when the reading printed is patently indefensible: blanda est, besides being in itself too feeble to be credible, is disqualified by the laws of anaphora, which rule out a verb in the present tense.

7.67 rigent, approved by Heinsius, strikes a coarse note characteristic of satire and lampoon (as the parallels show) but alien to elegy; it also renders the following ualent anticlimactic and removes an effective alliteration.

8.5 More irrelevant parallels; of course pulchra is an appropriate epithet for the elegiac mistress, but it totally lacks point here. pulchre identifies the girl as a connoisseur of literary style ( OLD s.v. 1b); at line 7 bene laudauit carries the suggestion ‘has exclaimed “bene”‘. Martinon saw the point: ‘Femmes curieuses de poésie’.

8.28 R.’s text hoc tibi, si uelles, posset, Homere, dari! would give perfect sense if it were indeed the command of a cohort rather than access to a mistress that is in question.

9.23 Printing ‘ aelinon! … aelinon!‘ robs Ovid of a little stroke of allusive wit: et Linon … ‘aelinon!’ explains in three words the aition of a ritual. That the cry is repeated at Aesch. Ag. 121 is irrelevant.

9.61 uenias gives inferior sense and is refuted by erit in the preceding verse and est at line 65.

11.46 dent, approved by Heinsius, makes Ovid hope that the gods will continue to countenance her infidelities.

12.3 votis is preferred to fatis for no sufficient reason; similarly at line 21 the banal canos is preferred to caros, which reminds the reader not only that Nisus’ magic lock was cherished by him and the city it protected but also that there was a high price on it.

13.29 ore is correctly preferred to ora, but it would have been better to pass over in silence the support of Madvig and Goold for the latter, offered in unawareness of the regular idiomatic use of populi pl. = ‘the people’ (Bömer on Met. 6.179).

15.6 A good example of the pitfalls of arguing from analogist premisses. At Tr. 4.10.8 fortunae munere is apt, here less apt than militiae turbine, which picks up the theme of 3.8.9-10 and suits an appeal to Amor. At AA 1.33 R. cites the analogous variation at Tr. 2.249 of nil nisi legitimum for nos Venerem tutam but has evidently not grasped its implication for this passage.

15.18 The colourless nostris preferred to the suitably assertive magnis, making the point that Ovid will meet the challenge of the grander genre with a matching style.


27-8 pro se quaeque parant nec, quos uenerentur amores, | refert: munditia crimina nulla merent. R.’s version of the pentameter offers a reasonable solution, the hexameter is a mess. (i) R. takes quaeque as nom. sing. fem., but Goold’s parant makes no sense without a direct object (his se sibi, neat in itself, is ruled out by est etiam placuisse sibi cuicumque uoluptas at line 31). (ii) uenerentur in this context is absurd. The note offers an example of the recurrent fault mentioned above: it is unhelpful to report Postgate’s cui se in isolation and leave it to the reader to work out how he read the verse.

35 The ‘parallels’ for uos are beside the point; with urget it simply makes no sense.

60 contere in haec: solidi etc. This punctuation is attributed both by R. and myself to Kunz, erroneously; what he actually says is that if it is adopted, ‘et sensus et constructionis exsistunt difficultates’, as indeed they do. We have a dilemma: punctuating after haec and reading solidi with Heinsius for MS solida we have acceptable sense but barely acceptable Latin; punctuating after contere, as the syntax dictates, leaves the substance which is to be added unspecified. Kunz was right to diagnose hopeless corruption.

86 utrumque, though accepted by most editors, does not seem to me to be possible Latin: can an ingredient ‘be’ a weight? Heinsius’ utrimque, one hasta different, sets the matter right literally at a stroke.

98 multus was espoused by Goold, explaining it as describing ‘a full/lasting complexion’. I should be surprised to learn that he had read Kunz’s note, which I think clinches the case for nullus.

Ars Amatoria I

40 The arguments adduced by Luck and Goold for terenda do not hold water. Their parallels show only that it is good Latin, and Goold’s contention that premo only = ‘press down’ was mistaken; it is the idiomatic word for hugging a coastline vel sim. ( OLD s.v. 10c), as Hollis points out.9 The case for terenda is further weakened by its recurrence at line 52 in a different sense and with particular emphasis.

62 What the parallels for bella demonstrate is its banality not its superiority to uera, which gives pointed sense, ‘a girl who really is a girl’, the genuine article ( OLD s.v. 4).

255 Editors who print uelis here no doubt appear to be playing safe, but compared with Bais — ‘Baiae and its (notorious) shores’ — it reads as a banalization. As to praetexo + dat. being ‘Silver’ Latin, see Kenney, CR 48 (1998) 212, and ‘ Vt erat novator : anomaly, innovation and genre in Ovid, Heroides 16-21′, in (edd.) J.N. Adams and K.G. Mayor, Aspects of the language of Latin poetry (Oxford 1999) 400-3, 406 n. 34.

328 uni posse placere uiro of course makes perfect sense per se, but not here: it was not Thyestes who fancied Aerope but vice versa.

Ars Amatoria II

91-2 decidit, utque cadens ‘pater, o pater, auferor’ inquit | clauserunt etc. This calls for a grammatical excursus: utque was first suggested by me, not by Luck — suggested and rejected for what still seem to me good grammatical reasons. Ovid uses ut temporal + present interchangeably with perfect only with uideo : cf. Met. 1.163 quae pater ut … uidit, 324-6 Iuppiter ut … uidet, 11.471 ut nec uela uidet, 12.426 ut uidet extinctum and Bömer ad loc. That is to say, to pass muster syntactically inquit must be perfect. Neve-Wagnener III 635-6 however cite only Catull. 10.27 inquii, Cic. De or. 2.259 inquisti, where there can be no doubt of the tense. I do not think it acceptable to father on Ovid a syncopated form inquît not secured unequivocally by metre, as are forms such as petît in its two occurrences in Ovid ( F. 1.109, Tr. 1.10.25)

94 In reading quoue Heinsius was merely reproducing the vulgate. It was either Müller or Riese who first saw that ‘Where are you or where are you?’ is hardly a sensible question. For -que‘proceeding from the general to the specific’ see OLD s.v. 9b: ‘Where are you? Where in the world are you?’

308 R.’s restoration of this much-vexed verse is predicated on the supposition that bowdlerization has been at work. In view of what the copyists have acquiesced in elsewhere one would have expected a more inflammatory Ur-version than R.’s et quaedam gaudia uoce nota, ‘call attention by your voice to certain joys’. This R. takes as referring to the words uttered during lovemaking, but, apart from the unOvidian oddity of expression, that adds nothing to uenerentur, which can hardly connote silent adoration. And what in the world are these ‘certain’, special joys? One does not get the impression that Ovid either went in for or approved the, shall we say, less orthodox forms of sexual behaviour.

345 R. prints nil without a note. The old editors printed nihil assuetudine; Heinsius’ substitution of consuetudine of course necessitated nil, but, now that assuetudine has by general consent been reinstated, nihil should come back too; cf. Housman, CP 1002.

690 Platnauer did not read or approve atque; his comment ( op. cit. 80) is weighted in favour of utque.

708 I should like to know what R. thinks figit means in this context; hardly that Amor shoots his arrows at the pudenda ? tingit, as Janka shows, is both obscene and apt.

Ars Amatoria III10

61 The combined efforts of Heinsius and Housman that produced uernos … educitis, apart from spoiling an amusing point, pay insufficient regard to the context: Ovid is not addressing himself exclusively to those still in the springtime of life.

95 ‘det Heinsius leniter [? leuiter ] distinguens post mulier‘ is guaranteed to mystify; he also read inquis.

146 I take the opportunity in passing to query the propriety of Gibson’s substitution of est for sit. There are certainly examples enough and to spare of the pleonastic use of jussive subjunctive + gerundive in prose (W.A. Baehrens, Beiträge zur lat. Syntax (1912) 501-8) to counsel caution.

241 As noted on AA 2.91, ut (Heinsius) here must be accounted a solecism. Goold was, moreover, mistaken in his contention that ‘As a conjunction, et merely joins’; see above on Am. 3.3.14.

274 I am not at all sure that inflatum (Par. Lat. 7993) can properly = ‘large’, and, though it receives support (!) from Rem. 337, it is reasonable to believe that both exaggerative and reductive bras were in use.

335 The comma after ullis, however it got there, makes hay of sense and grammar.

440 Priamei … tuis is printed on the authority of Madvig and, to my surprise, not ruled out of court by Gibson. This was a parlous lapse on the part of the great critic and grammarian. His parallels, faithfully reported by Cristante, are totally irrelevant: the synizesis of the vocative of such names is unthinkable in Ovid, who is punctilious in his respect for orthodox Greek prosody. Cf. AA 2.713 Briseï, Met. 1.504 Peneï, 13.858 Nereï, F. 6.553 Cadmeï.

764 R. prints ne … uide, citing parallels which merely show that it is acceptable grammar; he does indeed record Watt’s uides, but neglects to make it clear that it makes sense only if nec is also read.

803 R. resurrects the pre-Heinsian vulgate quod iuuat, which makes as little sense here as it does at Rem. 756, q.v. What does he think it means?

Remedia Amoris

18 Another case of misapplied analogy. R. follows Heinsius in reading e trabe against the a t. of the MSS on the strength of e trabe at Met. 5.127. Given that Latin uses a, e and de indifferently with pendeo, how can we be sure that Ovid stuck to one usage? For all we know he might have recollected Virg. Aen. 12.603 trabe … ab alta (cit. Pinotti) or it may be simply that his ear told him that a sounded right.

34 capto is printed on the strength of Burman’s parallels, which have no bearing on the situation here, where the uir is the girl’s husband or keeper not one of Cupid’s captives. R. does not seem to have grasped the purport of furtim in line 33.

467-70 To read ut for id in 467 and punctuate what follows as R. does destroys the sense of the passage.

481 R. reads sim with Heinsius, commenting ‘ ut postulat coniunctiuus qui sequitur‘. It is true that the syntax is informal, but Agamemnon cannot be made to say ‘If I were to be king’: he is king.

598, 606 More gratuitous tinkering after Heinsius, querentis for the faultless loquentis ( querenda at 644 provides no warrant for this), fleres + silua vocative — is Ovid likely to have apostrophized both Phyllis and the woods in consecutive verses?

704 tuque, fauens Par. lat. 7994, Goold, a choice predicated on unawareness of Ovidian idiom. For utque facis = ‘as indeed you do’ cf. Met. 7.815, Ex P. 3.3.47-8, OLD s.v. facio 26a ( TLL does not appear to recognise this usage). Here it expresses the poet’s grateful acknowledgement of past favours and confidence in more to come. R. in fact cuts the ground from under his own feet by adding ‘ lege Lucke‘, whose note should have disposed of tuque, fauens once for all.

756 R.’s version of this tormented verse is cui faueas actor quod iuuat arte docet, which unless I miss some subtlety should mean ‘the actor by his art teaches you whom to favour, which [sc. as a matter of fact; cf. on AA 3.803] is pleasing’, hardly the expected warning against theatre-going.

778 toro is preferred to uiro on the strength of Her. 19.100. Aside from the fact that the cases are hardly parallel, with toro, as Henderson pointed out ( CQ 30 (1980) 1b7), the reference is to Plisthenes, not to Agamemnon. And, after toro at 770, utrum in alterum?

The book on the whole is accurately produced. In the text the last word of the line has dropped out at Am. 1.3.21, AA 2.243, 381, 445, 3.239, 445, Rem. 359; at Am. 1.8.50 read ut for et, AA 2.138 closing punctuation omitted, Rem. 285 Ulixes for Vlixes; p. 73 marginal figures haywire. In the apparatus, at AA 1.97 the ed. Bonon. 1471 reads in not et; at Rem. 214 delete the (incomplete) duplicate report of Damsté’s emendation of 221. One inelegancy which a printer’s reader would have corrected when they still existed is the use of capitals for the inscriptions at Am. 1.11.27-8, 2.13.25, AA 3.812; small caps as at Am. 2.6.61-2 are much more suitable. A small but telling symptom of modern methods of book production is the number of turned-over colons in the apparatus. And while I am on that sore subject, let me conclude by noting a handful of errors in my own edition which still survive from the process by which that was produced: at AA 1.414 read fuit for fluit (reproduced unfortunately in good faith by R.), 2.217 delete comma at end of verse. Less important but still irritating are Med. 88, AA 2.78, where read murris, harundine respectively.

[[For a response to this review by Richard Burgess, please see BMCR 2004.01.14.]]


1. See his 5th edn (1970), adding 19a, 26a, 37a, 56a to the original 74.

2. Amatoria Critica was the beginning of a lifelong friendship, and I took Goold’s criticisms in good part. Forty years on I am more often inclined to take issue with them than I was then.

3. R.J. Tarrant, ‘Nicolaas Heinsius and the rhetoric of textual criticism’, in (edd.) P. Hardie, A. Barchiesi, S. Hinds, Ovidian transformations. Essays on the Metamorphoses and its reception (CPS Suppl. vol. 23), Cambridge 1999, 298.

4. Am. 2.9.17, 2.11.17, 2.15.21; AA 2.300, 3.359; Rem. 501.

5. R. reports Marius’ notes from Burman’s transcription. This does not matter very much, though at Am. 1.13.39 the words he quotes are Burman’s paraphrase, not what Marius wrote. Burman did not bother to identify his source, but it is not hard to find. Dominicus Marius Niger’s edition of Amores, Medicamina and Nux was published at Venice in 1518 by Tacuinus de Tridino. It does not seem to be all that rare, and it is duly recorded in the Bipontine Notitia Literaria, reprinted in Valpy’s Delphin.

6. Am. 1.1.7-12, 1.7.7-10, 13-18, 1.9.33-40, 1.10.1-6, 1.15.9-30; 2.4.11-30, 33-34, 2.5.35-40, 2.12.18-26, 2.14.13-18, 2.18.21-6, 29-34; 3.4.19-24, 3.6.25-44, 3.12.21-40; AA 1.326-40, 3.109-20, 137-48, 173-84, 321-6, 329-38; Rem. 55-68, 373-86, 453-60, 735-40, 743-8, 759-66 (only sequences of six lines or more cited). In many cases the sense is carried on from couplet to couplet, but there is no instance of true enjambment, which indeed is very rare in Ovid’s elegiacs.

7. Nuovi contributi di filologia e storia della lingua latina (Bologna 1994) 291-7.

8. See Kenney, in (ed.) Barbara W. Boyd, Brill’s Companion to Ovid (Leiden-Boston-Cologne 2002) 36-8 and n. 57.

9. My apologies in passing for failing to record that Vsus at 1.29 was first mooted by Mr Hollis.

10. R.K. Gibson’s edition of this book (Cambridge Classical Texts and Commentaries 2003) appeared too late to be used by R.