1. Let me offer thanks to the editors, the anonymous referee, and especially to Barrie Hall and Anna Ritchie for their provocation, assistance – and good humour when I have taken issue. Citations of MSS readings are for the most part drawn from Professor Hall's collations. I refer to the following editions by the names of their authors: P. Burman (Amsterdam, 1727); S. G. Owen (Oxford, 1889); A. L. Wheeler (Loeb Classical Library; Cambridge, MA, 1924; 2nd ed. rev. G. P. Goold, 1988); J. André (Paris, 1968); G. Luck (2 vols., Heidelberg, 1967–77: in each case I print Luck's text, though not always his punctuation). For an account of the manuscript tradition, and details of the sigla used here, see Tarrant, R. J., in Reynolds, L. D. (ed.), Texts and transmission (Oxford, 1983), 282–4Google Scholar.

2. ‘The poetry of Ovid's exile’, PCPS 11 (1965), 37–49, at 41Google Scholar; he goes on to argue for the irrelevance (but not the inauthenticity) of verses 23–8.

3. Hinds, S. E., ‘Booking the return trip: Ovid and Tristia 1’, PCPS 31 (1985), 13–32, at 28Google Scholar.

4. So Lond. B. L. Add. 18384, and, prior to correction, Copenhagen Gl. Kgl. S.2014.

5. Others are Trist. 2.186, 5.1.45; Ex P. 2.8.35.

6. Plural rather than singular because the corruption (with mali/tui ending the line) looks the more likely, and (with parte following) to vary the number.

7. hoc appears as a variant in Paris B. N. Lat. 8254.

8. Paris B. N. Lat. 7993.

9. Here lies perhaps the origin of the mistake in 6.

10. This conjecture was already in Professor Hall's text when I mentioned it to him: any credit is thus to be shared – but it is hard to believe we have not been anticipated.

11. Already in Bern. 478.

12. “Notes on Ovid's Tristia, Books I–II’, CQ 30 (1980), 401–19, at 413–14Google Scholar.

13. See ‘Notes on the language and text of the Tristia’, HSCPh 65 (1961), 243–61, at 254–5Google Scholar.

14. If it belongs here at all, it must go not with respicis urbem, but with the whole sentence – as in Diggle's (mis)translation: ‘and thus [because your vice-gerent is abroad] you keep watch over the city in person with one half of yourself’.

15. ‘Notes on Ovid's poems from exile’, CQ 32 (1982), 390–8, at 391–2Google Scholar.

16. Cf. Trist. 2.563; and more generally Bailey, D. R. Shackleton, Homoeoteleuton in Latin dactylic verse (Stuttgart, 1994), 56–79CrossRefGoogle Scholar.

17. He has es after an open longum only when the previous word is a spondee, the referee tells me (Met. 5.261 nostro es, Her. 18.169 caelo es); moreover, in his elegiacs he does not have elision after a iambus at 2d, unlike Propertius (cf. Platnauer, M., Latin elegiac verse [Cambridge, 1951], 83–4Google Scholar).

18. ‘Problems in Ovid's Tristia’, PCPS suppl. vol. 15 (1989), 20–38, at 31–2Google Scholar.

19. Thus Hall's abscondere (65, for abducere) is misguided: it is the invaders who destroy ‘what they cannot carry or drive away with them’. Ovid thinks of the pecus et stridentia plaustra of 59.

20. ς, Heinsius: arcem et MAGHPV: arces et F: arces D. Elision before the copula is paralleled in Ovid at his point in the couplet (Fast. 3.585, 6.443; Ex P. 2.6.23), and the apposition, the alternative Luck favours, seems rather ungainly.

21. uotis followed by uota looks suspect, even though one is substantival, the other adjectival. Schrader conjectured laurea uirga in 56. uobis in 55 would be another way to avoid the repetition, or else nostris or arcis … uestrae.

22. Concern with the return of the standards captured by the Parthians is registered also at Hor. Epist. 1.18.56; Ov. Fast. 5.579–94 (n. b. pudor, 587), Trist. 2.228; and in the images on various coins and the breastplate of the Prima Porta Augustus, on which see, e.g., Zanker, P., The power of images in the age of Augustus (trans. A. Shapiro; Ann Arbor, 1990), 186–92Google Scholar.

23. Zanker, , The power of images, 187Google Scholar.

24. As one had been in the clades Lolliana of 16 B.C. (Vell. 2.97.1).

25. They were later recovered, two by Germanicus (Tac. Ann. 1.60.3, 2.25), and one by Gabinius in A.D. 41 (Dio 60.8.7).

26. Cf. Harrison, S. J., ‘Discordia taetra: the history of a hexameter ending’, CQ 41 (1991), 138–49CrossRefGoogle Scholar.

27. For the phrasing we might compare Hor. Serm. 1.10.44–5 molle atque facetum | Vergilio adnuerunt… Camenae; there is a similar generic charge to mollia here.

28. On this, and other passages in which Ovid reworks Callimachus' Heraclitus epigram (2 Pf.), see Williams, G. D., ‘Conversing after sunset: a Callimachean echo in Ovid's exile poetry’, CQ 41 (1991), 169–77CrossRefGoogle Scholar.

29. Already in the codices HP.

30. At e.g. 1.11.20, 3.7.27, 5.2.70; perhaps also 4.3.76, 5.2.21.

31. See in general Galinsky, G. K., ‘The triumph theme in the Augustan elegy’, WSt 3 (1969), 75–107 (91ff. on Ovid)Google Scholar.

32. Amongst the longer such passages are Am. 1.2.23ff.: Amor over the poet (imaginary); Ars 1.175–228: Gaius over the Parthians (never achieved); Ex P. 2.1: Tiberius over the Dalmatians, and a forecast of a full triumph for the addressee, Germanicus; Ex P. 3.4: uncertain (perhaps Tiberius over the Dalmatians again), with a forecast of a further triumph over the Germans.

33. E. J. Kenney, in his notes to the translation by A. D. Melville (Oxford, 1992), p. 150; similarly Galinsky, , WSt 3 (1969), 102Google Scholar.

34. Cf. the note above on Trist. 3.12.45–6 (pp. 143–4).

35. spectare (19) recalls Prop. 3.4.15 and Ovid, Ars 1.217.

36. Verse 20 echoes Prop. 3.4.16 titulis oppida capta legam; cf. also Cons. Liv. 462. The Rhine (41–2; with Germania in 43–6; cf. the pairing of Egypt and Nile at Prop. 2.1.31) replaces the Tigris and Euphrates of Propertius 3.4.4 (cf. also 3.3.45 Suebo perfusus sanguine Rhenus, and Verg. Aen. 8.727 – following the Euphrates – Rhenus bicornis: Ovid aptly breaks the horns) and Ars 1.223–4.

37. Add Prop. 3.1.10 to the passages listed by Luck.

38. ac (et Paley) … plausu ς.

39. Millar, F. G. B., ‘Ovid and the domus Augusta: Rome seen from Tomoi’, JHS 83 (1993), 1–17, at 10–12Google Scholar. He finds in the exile poetry, which he sees as an important source for late-Augustan history, ‘not the voice of the subversive dissident, but that of the outraged loyalist’ (p. 1). The paper admirably demonstrates (what is hardly to be disputed) that Ovid is ‘Augustan’; it pays lip service to ‘ambivalences of attitude’ (‘critical ingenuity can in any case discover these in any text’, p. 7), but nowhere looks for signs of Ovid's outrage in the exile poetry. And here lies a problem: given what has happened to Ovid, we need to suppose him an outraged individual, but to those who choose to read superficially so that they can discover loyalty (less than 2 pages reveal the Metamorphoses as shaped by Augustan loyalism, without mention of any episode between 1.205 and Aeneas in book 13!), the text remains so polite that the anger against Augustus never appears. Search for Ovid's anger, and it begins to subvert the ‘overt loyalism’ (p. 9).

40. On which see Wiedemann, T., ‘The political background to Ovid's Tristia 2’, CQ 25 (1975), 264–71CrossRefGoogle Scholar; Nugent, S. G., ‘Tristia 2: Ovid and Augustus’, in Raaflaub, K. A. & Toher, M. (edd.), Between Republic and Empire (Berkeley, 1990), 239–57Google Scholar; Barchiesi, A., Il poeta e il principe: Ovidio e il discorso augusteo (Bari, 1994), 15–25Google Scholar; Williams, G. D., Banished voices (Cambridge, 1994), 154–209Google Scholar.

41. WSt 3 (1969), 102Google Scholar.

42. Cf. Trist. 4.2.37.

43. Galinsky, (WSt 3 (1969), 102)Google Scholar claims that the later passage differs in that ‘no imaginary answers are given’. This seems to ignore the concessive clause in Trist. 4.2.26.

44. Today it is easier to defend a novel about paedophilia or suicide than a handbook.

45. A reading that was happy to make light of the specific allusion might see this as inevitable human behaviour. In that case, the message is that banishing the author is in vain: naturam expelles furca …

46. The lovers from the Ars are not the only Ovidian figures I catch sight of in the crowd. Amongst Caesar's creatures, applauding as he passes, is the sinister figure cursed in that other poem of Ovid's exile: hos super in curru, Caesar, uictore ueheris | purpureus populi rite per ora tui, | quaque Ibis, manibus circumplaudere tuorum (Trist. 4.2.47–9). See Heyworth, S. J., PLLS 7 (1993), 85–6Google Scholar, for discussion of another place where this ambiguity seems to operate (Hor. Epod. 1.1); and for playful ambiguity between noun and verb in Ovid, cf. Met. 1.294 (the Flood) ducit remos illic ubi nuper Ararat, noted by Ryder, T. J. B., ‘Ovid, the flood, and Ararat’, G&R 14 (1967), 126–9Google Scholar.

47. Calling to mind the similar prediction of the same uates at the end of the Metamorphoses.

48. Nor indeed Germanicus' real father, Drusus: there is a sinister edge to verse 59 (maturosque pater nati spectabit honores).