^{page 1 note 1} See P. Ouidi Nasonis Amores, Medicamina Faciei Femineae, Ars Amatoria, Remedia Amoris, ed. Kenney, E. J. (Oxford, 1961), pp. v–xii.Google Scholar

^{page 1 note 2} The limitations of the evidence do not permit the construction of a stemma for the whole tradition, a statement which I do not regard as shaken by the stemma for the manuscripts of the Fasti (a somewhat analogous tradition) which will be found at Peeters, F., Les Fastes d'Ovide, facing p. 420.Google Scholar As to unnecessary detail, I have long thought that a case in point is the so-called ‘vetustus Politiani’ (on which see Lenz, , Parerga Ovidiana [Rend. Accad. Line. 1937], pp. 333–56; Munari139, n. 2, 141–3Google Scholar). This was a manuscript of the common class, of no particular importance for establishing what Ovid wrote, and almost certainly anything but vetustus. It is because it is lost that we have heard so much about it: ‘The fish that got away is always the biggest’ (Ullman, , Stud. it. difil. class. xxvii-xxviii [1956], 581).Google Scholar

^{page 1 note 3} The folio numbers are to be taken in all cases as referring to relevant works only, i.e. Amores, A.A., Remedia.

^{page 1 note 4} It was in fact known and used sporadically for the correction of the text before Heinsius exploited it systematically for his edition. Cf., for example, Carrionis, L.Emendationum … Liber Primus (Paris, 1583), p. 12Google Scholar, where he uses it to improve the received text of Am. 2.6. 1; Heinsius, to Scheie, R. H., Syll. ii. 744 (quoted by Blok 226)Google Scholar, Dörrie, 395.Google Scholar For the identification of manuscripts used by Heinsius see Lenz, , Eranos li (1953). 66–88Google Scholar; Munari, , Stud. it. di fil. class, xxiv (1949), 161–5, xxix (1957), 98–114Google Scholar; Dörrie, 399–409Google Scholar. Add incidentally that ‘cod. Bernardi Rottendorphii’ (Munari, , loc. cit. 101) = Wolfenbüttel 4620 (Gud. 313).Google Scholar

^{page 2 note 1} Who appears to have been the first to bring it into notice. He rightly relied on it heavily: (on A.A. 3. 37) ‘sed solus codex Regius hie audiendus: qui nobis inter tantas fluctuationes Cynosura est. Ejus vestigiis insistens, levi mutatione genuinam, nisi fallor, lectionem habebis ….’ The principle that the corruptions of a manuscript such as R are a truer guide than the corrected readings of interpolated manuscripts was instinctively perceived by Heinsius(cf. below, p. 64).

^{page 2 note 2} Not without some errors: the most serious deserve to be mentioned because they have been perpetuated by editors. At 1. 338 O has rabidi, not the inexplicable ualidi recorded by Ellis. At 1. 683 it has mala: Ellis's silence led Marchesi and Bornecque (reasonably enough) to infer that it had sua, the reading of Merkel's text, against which the collation was made. At 1. 513 munditia in Ellis for munditie is a simple misreading.

^{page 2 note 3} Why did Heinsius not perceive, as he was capable of perceiving (cf. Blok 232–3), the obviously close relationship between R and O? The explanation may be that his collation of O, which he made at the age of 21 on his first and only visit to England in 1641 (Blok 21), was not made with the care and minuteness that later distinguished his collations (see on these Lindsay, , Zcntralbl. fur Bibl. xviii [1901], 159–63Google Scholar; Munari, , Stud. it. difil. class, xxiv [1950], 161–5, xxix [1957] 98–114Google Scholar; Blok 226ff.). I notice two references to O in Heinsius' letters: in 1644, writing to Gronovius, he laments that he has mislaid his collation of O, which, he says, ‘reliquorum omnium manuscriptorum, quibus usus sum, instar mihi erat’ (Syll. iii. 125)Google Scholar, and in a later letter (1657) he speaks of it as ‘Uteris Langobardicis scriptus’ (ibid. 370). For this identification cf. Loew, E. A., The Beneventan Script, p. 28, n. 1 (b).Google Scholar

^{page 2 note 4} Ibid., pp. 152, 338.

^{page 2 note 5} The full list of contents is Theodulus, Ecloga; Maximianus; Statius, , Achilleid; Remedia; Her. 1. 1–7. 159Google Scholar; Arator, , De actibus ApostolorumGoogle Scholar. E is interesting as an early representative of the class of manuscripts that includes the Remedia in a cycle of poems thought to be of educational value: see Boas, , Mnem. xlii (1914), 44–46Google Scholar; Rand, , Speculum iv (1929), 260–1Google Scholar; Ullman, , Class. Phil, xxvii (1932), 38–40Google Scholar; Tafel, 45–59.Google Scholar

^{page 3 note 1} On this manuscript, a mainstay of Heinsius' text, but curiously neglected by later editors (Tafel was apparently aware of its existence, but did not use it for his dissertation), see the articles of Lenz (cited above) and myself (Stud. it. difil. class, xxx [1958], 172–4Google Scholar). It is composed of two quite unrelated parts, K itself, and a miscellany manuscript (fourteenth century) of the type described in the preceding note: see Lenz, , loc. cit. 28–30Google Scholar; Boas, , Mnem. xlii (1914), 42.Google Scholar

^{page 3 note 2} Cf. Lenz, , Eranos li (1953), 75. As regards the Remedia the identification is certain: 565 dormit, 645 querendo, 726 ilia] mille; and I regret to have to record that a hand of the seventeenth century which looks very like that of Heinsius has supplied some missing verses and made some corrections (these annotations are not noticed by James).Google Scholar

^{page 3 note 3} D is an immense Corpus Poetarum Latin rum: besides Ovid it contains Statius, Virgil, Horace, Lucan, Persius, Juvenal, Avianu Cato, Homerus Latinus, Sedulius, Prospi Aquitanus, Theodulus, and Maximianu See Boas, , Mnem. xlii (1914), 28–29.Google Scholar

^{page 3 note 4} The identification was made for tl Ibis by La Penna, (Publi Ovidi Nasonis lbis [1957], p. xcviii)Google Scholar, and for Her. by Dorr, (403)Google Scholar; but see also Lenz, , ‘Ovids Remedi und der Codex Iureti’, Stud. it. difil. class xxxi (1959), 169–74Google Scholar, contending that it doi not hold true for Rem. But cf. Am. 1. 8. t repressa, 1. 11. 23 lassere retento, A. A. 1. 34 suos, 360 luxuriatur, 389 temptabis, etc., Ret, 213 tandem firmis quamuis, 307 non cessent, 407 ueneri, etc. Lenz's suggestion that by ‘code Iureti’ Heinsius meant ‘a MS belonging to J.’ may be the right explanation; or H.'s notes may have been in a muddle (cf. below, n. 2).

^{page 4 note 1} The leaves from 33 to 37 are out of order, as follows: 32b ends with A.A. 3. 528, then follow 3. 593–656 (fol. 33). 529–92 (34), 720–83 (35), 657–719 (36), 784–fin. (37).

^{page 4 note 2} Merkel, , Fasti (1841), p. cclxxxGoogle Scholar, and Lenz, , Phil. Wochenschr. li (1931), 441–2, 677Google Scholar, identify it as the ‘alter Regius’; but some readings agree with those cited by Heinsius from his ‘tertius Regius’: A.A. 3. 377 talis (u.l.), 383 teretesque (u.l.), 476 e duro … ore. On the other hand, Heinsius gives A.A. 2. 244 apposita, 3. 694 tremunt as from the ‘alter Regius’ when they are in fact in Par. Lat. 7998 (a. 1305); and he reports A.A. 2. 355–6 as missing from the ‘prior Mentelii’ ( = P_b), which they are not. It is difficult to resist the conclusion that his notes were in some disorder or that he frequently misread them.

^{page 4 note 3} This identification was made by the late Mr. Wardrop, James (cf. C.R. N.s. viii [1958], 134)Google Scholar. In dating the manuscript to about 1470–80 he compared its writing with a Horace in die Library of King's College, Cambridge (James 34). Cf. the Martial in his hand in the Bodleian (Auct. F. 4. 33), reproduced as no. 9 in Humanistic Script (Bodleian Picture Book No. 12, Oxford, 1960).Google Scholar

^{page 5 note 1} See Lenz, , Eranos liii (1955), 63–64.Google Scholar

^{page 5 note 2} In e and p¹ but not in p³, the speech of Ulysses from Met, 13 intervenes between A.A. and Rem.

^{page 6 note 1} Some words, and probably also the complete verse 1. 662, have been lost by the cutting of the margins. As Tafel (17) does not give the order of the verses it may be worth while for me to record it here: 1. 3–4, 19, 21, 35–38, 41–42. 44–45. 49–50. 52, 57, 59, 61–62, 64/65–66, 84, 99–100, 132, 136–7, 139–40. ‘143–4. 146, 149–51. 159–60, 162, 221–2, 238–40/243–4, 246, 250, 269–70, 274–5. 343–5. 356, 394. 433. 442–3. 485. 503–4. 574. 578, 580, 595–6, 606, 615–16, 643, 645, 655–6, 661, <662>/663–4, 741–2, 752, 753–4, 2–197, 199–201, 279–80, Priapeia 5. 3–4, 2. 341–2, 464, 463, 363–4, 390, 351, 409–10, 437/438, 459–62, 501, 505–6, 519, 535, 548, 669–70, 702, 719–20, 3. 31, 41–42.

^{page 6 note 2} The leaves are bound in the wrong order: they should run 2, 1, 4, 3. The fragment contains 1. 64–89, 97–121, 129–52, 160–84 (184 covered by repair), 192, 320–41 (lacking 323), 349–71, 380–404, 412–37.

^{page 7 note 1} Fol. 103 is torn roughly across, about a quarter of the way from the bottom, in a manner that makes Tafel's theory (30) that the division was, though maladroit, intentional seem somewhat unlikely.

^{page 7 note 2} This assumes that in R′ the Heroides followed the Amores and that the order of the two works was at some date reversed when P was rebound (Tafel 31).

^{page 7 note 3} These titles must be older than a, since some were already missing from it: (R)PS give no title at 1. 2, 1. 6, 2. 18, 2. 19, 3. 7. No doubt a was copied from a manuscript in which the titles were written in the margins (with no space left between elegies), where they might more easily become illegible or escape the copyist's notice. They can hardly be antique: some are inept (1. 5 compositus est ad Corinam, 2. 2 suasorium ad se [unless this is misplaced, as suggested by Merkel, or a truncation of ad se<ruumy>]), and all are banal. Some of the recc, notably DFHZ, exhibit certain agreements widi PS in the form of the titles, but the resemblance is not close enough to guarantee their existence in the archetype (supposing that there was such a thing: cf. below, p. 26). Certainly in the earliest state of the tradition of which we have evidence the introductory Epigram was not separated from 1. 1; and if, as I believe, the suggestion of Lucian Muller was correct (see below, p. 13), 2. gb and 3. lib were already then merged with the poems that preceded them.

^{page 7 note 4} The intermediary p is inserted in anticipation of the identification made below, p. 24.

^{page 8 note 1} P is our sole authority for the use of this word by a classical author.

^{page 8 note 2} A_c has guaeras, not -is, as reported by Munari. I perhaps ought to say expressly that Munari's edition is extraordinarily accurate.

^{page 8 note 3} Unless one counts 2. 8. 7 num (prius) S: nam Pf: nunc P: non ω. Cf. 3. 4. 8 occlusis SA_bV_b: exdusis Pωφ: inclusis ς.

^{page 8 note 4} Possibly its exemplar was difficult to read: at 2. 16. 5 and 2. 17. 9 words are omitted and a space left. I have given a sufficiently ample selection of S's errors in the Appendix to my edition; a scrutiny of it should make an editor reflect seriously before laying any weight on its unsupported testimony at, for example, 1. 3. 3 petii, 2. 12. 20 mouit, 3. 7. 20 amata (this last accepted by Munari on the recommendation of Knoche).

^{page 8 note 5} Also exemplified in my Appendix. They differ from those of S in being for the most part obvious blunders. The list could be lengthened enormously if all mistakes of the type of ad for at, iubet for iuuet, iacerem for iacerent, and the confusion of terminations, were included. Thus the editor should beware of taking seriously, for example, I. 6. 65 pruinosus (cf. 1. 9. 17 infestus), 2. 14. 15 negasset (cf. 2. 4. 48 ingestum), 3. 3. 37 metuere (cf. Munari ad loc), 3. 11. 7 perferre (cf. 1. 6. 54 adesse). It is perhaps unlikely, though it is by no means impossible, that a future editor will emulate Bornecque and print stellantia from P at 1. 8. 11. quaerenti at a. 11. 22 is perhaps in a different class, as a probable interpolation rather than a blunder: it is certainly not what Ovid wrote (cf. Lee, , C.R. N.S. ii [1952], 175).Google Scholar

^{page 9 note 1} Cf. Knoche, , Gnomon viii (1932), 522.Google Scholar

^{page 10 note 1} Apropos of 3. 12. 15, where it gives, correctly in my opinion, thebae. But cf. Lindsay, W. M., An Introduction to Latin Textual Emendation, pp. 68–69Google Scholar; also Housman, , J.P. xxxiii (1913), 59: ‘When it is a question between e and ae, not even the best and oldest of Latin MSS are competent witnesses.’Google Scholar

^{page 10 note 2} This seems certain: cupido repeated so soon after 47 is pointless and inelegant; it probably stems from an untimely reminiscence of 1. 6. 11 risit, ut audirem, tenera cum matte Cupido. rogantem may be due to conjecture, for an object is wanted after exaudis, and precantem or rogantem are obvious possibilities; the hyperbaton is characteristic of Ovid.

^{page 11 note 1} I once proposed, but have never ventured to print, a conjecture which if accepted would eliminate this example. Is it possible that equo … tauro is a corruption which conceals the truth and that Ovid wrote non equa munus equo, non a boue uacca poposcit? For the antithesis bos-uacca cf. Fast. 4. 826 alba iugum niueo cum boue uacca tulit, Met. 9. 740–1 imagine uaccae / passa bouem est; for posco + a, cf. Am. 1. 10. 53, A.A. 3. 25, Rem. 289; and for theposition of the preposition cf. Am. 2. 19. 31–32, A.A. 1. 333, 723–4, 763, Her. 6. 107–8 (reading Tanai) and cf. Notes I, p. 55; II, p. 254. But to postulate the ousting of a boue by a gloss tauro perhaps strains credulity, and I do not press the suggestion.

^{page 11 note 2} Munari sets forth the position admirably, but declines to pronounce the Somnium indubitably spurious. Once it is recognized that it was not written by Ovid—and I do not see how the attribution can be maintained—Munari's anxious questionings (149–50) are seen to be unnecessary: the Somnium was never ’separated’ from the Amores because it never belonged to them in the first place (Munari notices this solution as a possibility, 150, n. 1).

^{page 12 note 1} For manuscripts not on my list see Munari, 147–9.Google Scholar

^{page 12 note 2} Pace Ellis, R., The Amores of Ovid, p. 7. I cannot guess the source of his statement that the elegy ‘was thought spurious by Heinsius’. Cf. Munari 144: ‘Nessuna traccia di dubbi in N. Heinsius.’Google Scholar

^{page 12 note 3} In Va the Somnium stands after 3. 15; it was not in Va's exemplar, but the copyist evidently took it from a manuscript in which it stood after 2. 5, as in D (Munari 148, n. 3).

^{page 12 note 4} Cf. Munari 145–6. It might be argued that 3. 13 is just as unlike any other poem in the Amores, indeed much more unlike; and it is true that this attractive piece (on which cf. Lenz, , Stud. it. di fit. class, x [1933], 312–13Google Scholar; Fabula i [1958], 255–62Google Scholar) reads more like a foretaste of the Fasti than anything else, but it could have been written by nobody but Ovid. What is really odd about it is the mention of his wife, as remarked by Fränkel, H., Ovid, p. 235, n. 25.Google Scholar

^{page 12 note 5} Lenz, , Bursian 226 (1930), 112–13, produces parallels from the Fasti for the beginning in medias res. But the narrative technique of a long poem (continuous, though not an ) is different from that of a short self-contained elegy. Munari (145) shows himself too tolerant of Lenz's contention.Google Scholar

^{page 12 note 6} See, for example, Housman, , J.P. xviii (1890), 7Google Scholar; C.R. xi (1897), 428Google Scholar; Postgate, , C.R. xxx (1916), 142 ff.Google Scholar; Platnauer, , Latin Elegiac Verse, pp. 105–8.Google Scholar

^{page 12 note 7} The closest are perhaps A.A. 1. 399–400tempora qui solis operosa colentibus arua, / fallitur, et nautis adspicienda putat, Her. 3. 19 si progressa forem, caperer ne, nocte, timebam, 10. 110 illic, qui silices, Thesea, uincat, habes.Google Scholar

^{page 13 note 1} Compare 6 lene sonantis aquae, a tag borrowed from the Fasti (where it occurs twice) and quite unsuitable in its new context.

^{page 13 note 2} Cf. Lee, A. G., Ovidiana, pp. 468–9.Google Scholar

^{page 13 note 3} For a separate appearance of the couplet in miscellany manuscripts (including p6) see Baehrens, , P.L.M. iv. 17.Google Scholar

^{page 13 note 4} Some miscellanies combined genuine extracts from Ovid together with pseudo-Ovidiana: e.g. Bb has Am. 2. 15 (Anulus) along with A.A., Pulex, De Medicamine Aurium, Cuculus. The Somnium is first attributed to Ovid by Servius auctus ad Buc. 6. 54, but this does not necessarily imply, as Ellis thought (The Amores of Ovid, p. 7)Google Scholar, that the poem was cited from a copy of the Amores. The sequence was probably (1) juxtaposition with Ovidiana and pseudo-Ovidiana, (2) attribution to Ovid, (3) incorporation into the canon. Our Somnium is not to be confused with the medieval poem of the same name discussed by Lehmann, P., Pseudoantike Literatur des Mittelalters, pp. 63–65.Google Scholar

^{page 13 note 5} Bornecque did not use Sa for his edition of the Ars. Vollmer's ‘Kritischer Apparat zu Ovids Remedia’ (Hermes lii [1917], 453–69) was based on the work of Tafel, who was killed in the First World War.Google Scholar

^{page 13 note 6} Absurdly accepted by Ehwald into his text. Romulus was a and primus is the mot juste for him: cf. Am. i. 7. 31–32, 2. 3. 3, 2. 11. 1–2, 2. 14. 5–6, 3. 10. 11–14; for the idea in general see Leo, , Plaut. Forsch.², pp. 151 ff.Google Scholar, and for this passage in particular Bürger, R., De Ovidi carminum amatoriorum inventione et arte (diss. Wolfenbiittel, 1901), pp. 48–50.Google Scholar

^{page 14 note 1} A reminiscence of a common Ovidian phrase (fiducia formae at A.A. 1. 707 and five times in Met.). Housman's ferme was uncharacteristic (from an avowed enemy of palaeographical conjecture) and unhappy: the word does not occur in the poets between Lucretius and Juvenal (Axelson, , Unpoetische Wörter, pp. 136–7).Google Scholar

^{page 15 note 1} There is a small clue, partially obscured by Ellis in his collation of O and overlooked by Tafel, to the provenance of η At 1. 267 ubiq for ubique, 655 uterq … neq, for uterque … neque point to a confusion of the Continental symbol for quod (q) with the Insular symbol for quod (q). Cf. Lindsay, W. M., Notae Latinae, pp. 228 (‘This … ancient Nota [q] … is characteristic of the earlier minuscule of all parts of the Continent, except Spain’), 254.Google Scholar

^{page 15 note 2} Examples in the Appendix to my edition. At 1. 448 it is interesting to note that Tanaquil Faber might have claimed manuscript authority for his emendation of praeteritum to nam pretium (see Tanaquilli Fabri Epistolae [Saumur, 1659], Ep. liiiGoogle Scholar [not lii, as reported by Burman, ], p. 179Google Scholar; the correction was seriously debated, more seriously than Faber's merits dictated—cf. Heinsius' famous note on A.A. 2. 660; Bourchenin, D., De Tanaquilli Fabri Vita et Scriptis [diss. Paris, 1885], p. no). O has pretium for praeteritum, obviously through simple error. Tafel (16) refers O's deo at 1. 76 (+Pa) and 416 to ’Interpolation eines chrisdichen Abschreibers’.Google Scholar

^{page 15 note 3} C. Marchesi in his edition of die Ars (Corp. Parav. 1918) plumed himself (p. ix) on restoring feras on the authority of R at 1. 199. Similarly editors have at various times preferred 2. 87 dispexit, 3. 150 hyblae (cf. Clausen ad Pers. 1. 131), 261 mendo (cf. Notes II, p. 257Google Scholar); and the supposed authority of R has maintained in the text a manifestly false reading at 1. 147, which on inspection turns out to be the work of r: see C.R. N.S. iii (1953), 7–10Google Scholar. A selection of R's errors in the Appendix to my edition; it would be possible to add innumerable instances of omission of letters and syllables, and confusions of d and t, u and b, and the like. In the face of these propensities one appeals to the authority of R with diffidence: e.g. at 3. 170 rubes (also in AN) may very well be due to colores in the following verse, as Mr. A. G. Lee points out to me; the apostrophe seems singularly pointless and is not necessitated, as often, by the metre. A more important case at 3. 343 deue tribus Aς: deque tribus ς: deue cerem R: deie cerem r. How is it possible to assert confidently, as was done by, for example, Schanz-Hosius, , Gesch. d. rom. Lit. ii 4. 211Google Scholar, that the reading of R conceals the truth and that de(q)ue tribus must be interpolated? The vulgate reading may be inconvenient to literary historians striving to establish the chronology of Ovid's early work, but an editor cannot help that. I do not positively assert that the scribe of R wrote cerem for tribus, but it cannot be proved that he did not and in view of his other errors it cannot be called unlikely. ‘Die Wahrheit hat Regel und Einheit, Irrthum und Zufall dagegen sind regellos und können daher nicht stets bis auf ihren Ursprung verfolgt werden’ (Boeckh, A., Encyklopädie u. Methodologie d. philol. Wissenschaften [1886], p. 194).Google Scholar

^{page 16 note 1} His article may be consulted with advantage for a description of die manuscript, but his observations on the text, in view of the inadequacies of the editions on which he had to depend, are worthless.

^{page 16 note 2} As seems to be demonstrated by the fact that the rubrication includes the scholia, which are also by a. It is difficult to be sure whether A and a are identical; they are certainly very similar, a, like A, several times agrees with R in error: 2. 690 atque (+Og), 3. 288 cum risu usa est (+Pa), 398 ueste (+BfoOg). It looks as if A and a stem from the same exemplar and the choice between text and variant was governed by the whim of the copyist: cf. on Pb* below, p. 18.

^{page 17 note 1} For no apparent reason: cf. the omission by R of Rem. 9–10 and by P of Her. 9. 63–64, 147–52. When such things can happen, it is difficult to see why such a coil should be made about the omission in a of Am. 1. 13. 11–14 and 2. 2. 18–27 (see above, p. 9), textbook cases of homeoteleuton and homeoarchon respectively.

^{page 17 note 2} Cf. the even more remarkable case of Wolfenbüttel Gudianus 297 in the Heroides (Dorrie179–84).Google Scholar

^{page 17 note 3} Cf. Ker, A., Ovidiana, p. 224Google Scholar. For the expression compare 197 genitor patriaeque tuusque; for the flattery Tac. Ann. 13. 6 imperatori quantum ad robur deesse, cum octauo decimo aetatis anno Cn. Pompeius, nono decimo Caesar Octauianus ciuilia bella sustinuerint?; and for the syllepsis in auspiciis annisque—almost Ovid's favourite figure— Fränkel, , Ovid, p. 197, n. 10Google Scholar; Bell, A. J., The Latin Dual and Poetic Diction, pp. 304 ff.Google Scholar; Korn-Ehwald, ad Met. 9. 135. The confusion annus/animus is common: e.g. A.A. 2. 669, Rem. 392 (Tafel 40–41). At Am. 2. 6. 8 I believe that Ovid wrote expleta est annis ista querela suis = ‘mat complaint of yours is satisfied by its years’, i.e. by the long passage of time, a sentiment that he repeats and varies at v. 10 magna sed antiqua est causa doloris Itys. At Am. 1. 9. 5 Rautenberg's animos (transmitted by BVb,) is not to be dismissed out of hand, though it is safer to keep annos.Google Scholar

^{page 18 note 1} I. 592 bella Οαωο: uerba RAς: tela H: Theodulf, , Carm. 28. 640ad fern nefaciles sint tibi uerba mantis (cf. Tafel 69). (The error uerba for tela at Am. 2. 1. 19 tela Itali, T: uerba PSς: bella ς.) Cf. also the echoes of the Somnium in Theodulf and Paulus Diaconus noted by Munari 146.Google Scholar

^{page 19 note 1} I cannot be absolutely certain of the reading of Oa¹ here, r's reading may be no more than a lucky guess: cf. 3. 151 positus ω: positos RAς: posito rL1: cultus ς, where the attempt was less successful.

^{page 19 note 2} Cf. C.R. N.S. iii (1953), 7–10.Google Scholar

^{page 19 note 3} See above, p. 17, n. 3.

^{page 19 note 4} On the full reading of Sa see above, P. 15.

^{page 19 note 5} Also [Plan.] ): see below, p. 21, n. 3.

^{page 20 note 1} abscedere, though dismissed by Heinsius as ‘contrario plane sensu’, and translated with marvellous perversity by Brandt and Bornecque, is defensible: cf. 2. 349 ff. cum tibi maior eritfidttcia, posse requiri, / cum promt absenti curafuturus eris, / da requiem eqs.; Prop. 2.14. 19–20 hoc sensi prodesse magis: contemnite, amantes; I sic hodie ueniet, si qua negauit heri; Plaut. Mil. 1034–5, and cf. Prinz, , Wien. Stud, xxxvi (1914), 53, n. 1.Google Scholar For abscedere used of feelings and emotions see Thes. L. L. i. 145. 79Google Scholar (cf. abeo, ibid. 70. 27 ff.). But at this stage of the affair the advice implied by accepting abscedere would be premature (cf. also below, p. 25, n. 1).

^{page 20 note 2} Cf. 763 R. The reference must be to the diversity of the quarry, not the locale.

^{page 20 note 3} He seems to have known of its existence, though he did not use it for his dissertation: Lenz, , Stud. it. difil. class, xxix (1957), 2.Google Scholar

^{page 21 note 1} K² has in fact written ars fere over the line without altering temporibus.

^{page 21 note 2} The true explanation of uino, that planetrees were thought to benefit from the infusion of wine on their roots, was given by Gronovius, , Observ. I. v; Lenz's note is beside the mark.Google Scholar

^{page 21 note 3} This is the Greek prose version mentioned in Notes II, p. 258, n. 2.Google Scholar

^{page 21 note 4} citharae lotosque was in fact first read by Salmasius, ex codd., and was reported by Heinsius from exc. Seal.: it may well be that R was the source of both reports (cf. below, p. 30). Lenz's critical note is inaccurate: below, p. 22.

^{page 21 note 5} At A.A. 3. 158 R preserves the correct spelling cnosi. Editors still do their best to render Ovid uncouth: cf. the index to Breitenbach's, H. edition of Met. (Zurich, 1958): ‘Cnidos s. Gnidos! Cnossos s. Gnosos!’Google Scholar

^{page 21 note 6} Lenz, , Stud. it. difil. class, xxix (1957), 10, well compares Am. 1. 2. 33 tendens R: tendent Sω.Google Scholar

^{page 21 note 7} Examples in Appendix to edition.

^{page 21 note 8} Examples in Appendix to edition. Several of E's peculiar readings recall those of S in Am. (see above, p. 8, n. 4): 71 uoluistis, 85 potantibus, 102 lentae, 647 queraris (loquaris for queraris at Am. 1. 4. 23 queraris ς, exc. Put. et Seal.: loquaris ΡSω. I have not included this in the list at p. 11, above, but queraris must surely be right: Munari's suggestion that loquaris = male loquaris is ingenious, but reference to the passages on which it is based will show that in all of them the context makes the pejorative sense unmistakable, which cannot be said of the Amores passage).

^{page 21 note 9} Examples in Appendix to edition. Peculiar readings which recall S and E are 93 profer, 186 torta, 522 prompiius.

^{page 22 note 1} Stud. it. difil. class, xxix (1957), 13.Google Scholar

^{page 22 note 2} Tafel, 45–59Google Scholar; Boas, , Mnem. xlii (1914), 17Google Scholar; Ullman, , C.P. xxvii (1932), 38.Google Scholar

^{page 23 note 1} Cf. 111 partem REK¹ς: parte K²ς.

^{page 24 note 1} For the presence of Her. in o see Tafel 22–23, 30–31. Here a complication arises, however: Dörrie has shown, to my mind convincingly, that the tradition of Her. cannot be traced back to two independent copies of the archetype (Dörrie, 124 ff.Google Scholar; cf. my review, Gnomon xxxiii [1961], 479, 484Google Scholar), i.e. there is no equivalent in the tradition of Her. of the β-branch. It seems to follow that the archetype of Her. and the archetype of Am., A.A., Rem. (using the term ‘archetype’ for convenience, as explained) were distinct, and that the two traditions were first amalgamated in a (cf. Dörrie, 127Google Scholar). This hypothesis helps to explain why the text of Her. is so much less well preserved than that of the other amatory works.

^{page 24 note 2} De Re Metrica2, pp. 24 ff.Google Scholar

^{page 24 note 3} Am. 2. 1. 24 funds P, A.A. 1. 9 oui ROS_a, Rem. 433 istamquere R, Her. 9. 160 eat P (Tafel 34). At A.A. 1. 9 oui must have been the reading of a, therefore it was a's exemplar that was in majuscules: cf. Timpanaro, , Stud. it. difil. class, xxxii (1960), 57 ff.Google Scholar

^{page 24 note 4} Possibly Spanish symptoms are faulty aspiration (cf. Loew, E. A., The Beneventan Script, p. 102Google Scholar), the confusion of e/ae (common in R), it/a (e.g. A.A. 2. 660 sit] sa R), i/y (e.g. 1. 201 parihy R), ci/a (2. 358 uanescitque] uanesatque, 444 eliciendus] eliaendus R). On the other hand, the confusion a/ti (1. 612 arte] tirte, cf. 525 uatem] uirtim R) recalls the Corbie ab script; in, e.g., the example at Lowe, , Codd. Latt. Antiqq. v. 650Google Scholar the confusions e/i, o/u, b/p, c/g, d/t, all characteristic of a, are found. Cf. also Knoche, U., Handschriftliche Grundlagen des Juvenaltextes, p. 246, n. 2Google Scholar, on the quod/qui, quod/quo confusion in early Continental minuscule.

^{page 24 note 5} Not 26, as suggested by Müller, , loc. cit., p. 27Google Scholar; cf. Tafel, 33–34.Google Scholar

^{page 25 note 1} Cf. Tafel, 74.Google Scholar Corruption is sufficiently attested by the common errors noted above, pp. 7, 13–14. Forgetfulnessofthe fact can lead to unwise editorial choice: e.g. at Am. 2. 11. 21 at ω. ad PS, edd. nonnulli, distinctione sic posita—haec alii referant ad uos: quod quisque eqs. Cf. 3. 1. 57 at ilia] adilla P: ad illam S. The borderline between accidental and wilful depravation (‘corruption’ and ‘interpolation’) is difficult to draw, particularly in a tradition such as this, but readings of a which seem to me to belong in the class of interpolations are the following: Am. 1. 10. 49 pepigisse ς: tetigisse PSς; 1. 14. 24 mala ς: male PSς (cf. Notes I, p. 58Google Scholar); 2. 4. 5 non esse ςφ: non nosse PSς: odisse ς; 2. 4. 46 moribus VbW: corporis PSω (cf. Notes I, p. 60Google Scholar); 2. 6. 33 ducensque DF: ducitque PSω; 2. 7. 23 omandis ω: ornatis PSς … operosa ω. operata PSN; 2. 11. 40 aestus ω: eurus PSς (cf. Notes I, p. 62Google Scholar); 2. 18. 26 dicat pς: dictat PSς (but dictat has, incredibly, been defended); 3. 3. 11 aeterni S²ω. aeterno PS¹ς: aeternum ς (cf. Notes I, p. 64Google Scholar); 3. 7. 9 cupida … lingua ω: cupidis … linguis PbPc: cupide … linguis PSς; A.A. 1. 64 cogeris Itali, aW² (ut uid.): cogens et ROS_aAbω (cf. Notes II, pp. 241–2Google Scholar); 269 cunctos Aω: formae ROb (cf. above, p. 14; the error may, however, have been wholly unconscious); 684 duos O (u.l.) Pf: uenus ROAω (cf. above, p. 15); 715 accedere ω: abscedere ROAς (cf. above, p. 20, n. 1. It seems probable that abscedere arose from a misunderstanding of the preceding a); Rem. 19 fodit Itali, Og: fodiat REKω (doubtful); 446 diducto Kςφ: deducto Eς: seducto Rς: subducto ω (cf. Notes II, p. 259Google Scholar); 599 subnubilus rK² (u.l.) ω: sub nubibus REK¹ς; 798 an ueniat K¹ς: adueniat REK²ς. (It will be appreciated that A.A. 2 and 3, where R is the sole surviving representative of a, do not come into consideration.)

^{page 25 note 2} In my notation these are E_aHNO_bV_aV_b.

^{page 26 note 1} Observe the caution of Tafel (32): ‘Wir kennen weder die Mittelglieder noch den Gang der Zersplitterung des Korpus.’

^{page 26 note 2} Cf. Dörrie, 419 on the Heroides.Google Scholar

^{page 26 note 3} Cf. Pasquali, 15–21Google Scholar; and on Am. Munari, 150Google Scholar; on Juvenal see Housman's, edition, p. xl, n. 1.Google Scholar

^{page 26 note 4} For the term see Pasquali 140ff.; and for the analogous situation in the Metamorphoses see Pasquali 388–90, Slater, D. A., Towards a Text of the Metamorphosis of Ovid, P. 19.Google Scholar

^{page 26 note 5} On these variants see Rosa, L., ‘Sulle varianti della tradizione manoscritta degli Amores di Ovidio’, Annali della fac. di lett. e fil., Naples, iv (1954), 41–60. Her selection of examples is uncritical, but the discussion demonstrates well the spirit in which medieval copyists approached their task, as reflected in the variants which they introduced into the text. Many readings transmitted solely by s clearly come into the class of ‘metrical variations’: e.g. Am. 1. 13. 48 adsueto] est solito; 2. 16. 40 saxa cruore] sanguine saxa; 2. 16. 44 sidera nostra] qui rapuere (ex 3. 11. 48); 3. 9. 52 dilaniata]dilacerata; 3. 15. 6 militiae turbine] fortunae munere (ex Tr. 4. 10. 8); A.A. 1. 591 stimulata]stimulante;610 cupias] incipias; 2. 282 turba sed] sed tamen; 370 cubare] iacere; 380 aonii cornibus icta dei ] aonio concita baccha deo (ex 1. 312); 603 exigua] eximia; 3. 297 quoniam prosunt impendite] magnam prodest impendere; 421 speciosd] formosa; 722 micante] tremente; Rem. 75 incipiens] 0 uates; 233 tristissima] strictissima; 640 suppliciter] simpliciter, etc. Such 'variations‘ (for the term cf. Dörrie 125) need not usually be signalled in the apparatus criticus; but qualitatively most of them are not very different from, for example, Am. 1. 13. 7 aer ] humor Sf; 1. 13. 15 colentes] colonos Sς; A.A. 1. 63 cupis] petis Aςs; 627 ostendii] ostentat Aς; 761 modo se] sese Aς; Rem. 45 salutares] salutiferas K2r; 398 fortius] fortiter EK'r; 653 euanidus] radicitus Kf, etc. These are not real ‘variants’, but an editor is (rightly) expected to record them; an apparatus in which all obviously false or unauthoritative readings were suppressed might in a sense be truly critical, but it would not fulfil one of its functions, that of giving a conspectus of the tradition.Google Scholar

^{page 27 note 1} artis rB2F: irae Aω: ite R.

^{page 27 note 2} capto Itali: coepto RPC: cepto OAω. Cf. Notes II, pp. 244–5.Google Scholar

^{page 28 note 1} It is perhaps worth pausing to establish that Ovid certainly wrote omisi. The verses (15–20) run as follows: in manibus nimbos et cum loue fulmen habebam / quod bene pro caelo mitteret ille suo. / clausit arnica fores: ego cum loue fulmen omisi,‘ / excidit ingenio Iuppiter ipse meo. I Iuppiter, ignoscas: nil me tua tela iuuabant; / clausa tuo maius ianua fulmen habet. The only variants of importance are those at v. 17: fulmen omisi ς: fulmen amisi PDN²: fulmina misi pSω—the steps by which the vulgate reading arose are clear. At 15 cum loue is arro Ovid was just embarking on the composition of a Gigantomachia and, poet-like, identified himself with his subject: ‘With Jove I held in my hand the clouds and the thunderbolt, a bolt worthy to have been thrown by him in defence of his heaven [the conventional rendering of 16 is feeble and pointless: Ovid surely means that he was doing his subject proud—cf. 12 et satis oris erat]. My mistress shut her door: I dropped Jove and thunderbolt both—Jove himself fell from my mind [17 is an example of Ovid's favourite figure, syllepsis; cum loue now has to be taken in a somewhat different sense, but excidit in 20 insures against misunderstanding]. Forgive me, Jove; your weapons were of no use to me; her closed door has a more powerful bolt than yours.‘ The correctness of omisi is guaranteed (a) because the sense demands a word meaning ‘dropped’, and misi means ‘threw’ (as Burman pointed out in 1727); (b) because the singular fulmen is vital to the word-play (for fulmen in this sense cf. 1.6. 16; the joke was pointed out by Lee, , C.R. N.S. ii [1958], 176). It may be remarked that this passage may not be used as evidence that Ovid really wrote or began a Gigantomachia.Google Scholar

^{page 28 note 2} See, for example, Housman, , Iuuenalis, pp. xxvi–xxviiGoogle Scholar; Knoche, U., Handschriftliche Grundlagen des Juvenaltextes, p. 74Google Scholar; Pasquali, 43–108Google Scholar; Clausen, W. V., Per si Saturarum Liber, pp. xviii–xxii.Google Scholar

^{page 28 note 3} A good example in Clausen, , op. cit., p. xxii, n. 1.Google Scholar

^{page 29 note 1} It is instructive to compare S. G. Owen‘s ed. maior of the Tristia with his O.C.T. and to ask oneself which type of apparatus really tells the reader more.

^{page 29 note 2} The indirect tradition for the carmina amatoria is sparse and of little assistance to the editor: see Tafel 62 ff.; Lissberger, E., Das Fortleben der Rom. Elegiker in den Carm. Epig., diss. Tubingen, 1934. When Seneca (Contr. 2. 2. 8) puts rursus into the mouth of Porcius Latro at Am. 1. 2. 12, he is almost certainly misquoting; similarly the author of C.L.E. 936, who gives A.A. 1. 475 as quid pote tan durum saxso aut quid eqs.Google Scholar

^{page 29 note 3} In particular I think of Axelson‘s wholly convincing subeslur at A.A. 1. 620 (Hermes lxxxvi [1958], 127–8).Google Scholar

^{page 29 note 4} On these and other florilegia see Tafel, 46–47Google Scholar; Ullman, , C.P. xxvii (1932), 11–19Google Scholar; Boas, , Mnem. xlii (1914), 22 (p₂)Google Scholar; Lenz, F. W., Parerga Ovidiana, pp. 404–10.Google Scholar

^{page 30 note 1} See Palmer's edition² of the Heroides, , p. 422Google Scholar; Ullman, , C.P. xxvii (1932), 13.Google Scholar

^{page 30 note 2} It can hardly have been earlier in date than the twelfth century (so Ullman, , C.P. xxiii [1928], 131, n. 1)Google Scholar; certainly not the ninth, as suggested by Purser, (loc. cit.).Google Scholar

^{page 30 note 3} The two sets can hardly have been identical, or Heinsius would not have bothered to cite both. Either one set was copied, with some omissions and a few additions, from the other (i.e. Put. from Seal.), or both derived from a common source.