^{page 241 note 1} The opposite is carere numeris suis, as at Tr. I. 8. 48 ut careant numeris tempora prima suis, correctly explained by Owen, , who cites Cic. de fin. 3. 24Google Scholar, quoted above, and Met. 1. 427 animalia … quaedam imperfecta suisque | trunca … numeris: ‘the beginning of our friendship lacks its (remaining) component parts, i.e. does not correspond to the end’. His other parallel, Am. 3. 7. 17–18 quae mihi uentura est, siquidem uentura, senectus, | cum desit numeris ipsa iuuenta suis? must be disallowed: there numeri means ‘function, duty’; cf. Her. 4.88 Veneri mtmeros eripuisse suos, Rem. 372 ad numeros exige quidque suos, Consol. ad Liu. 285 numeros impleuit principis.

^{page 241 note 2} Sigla used in the Ars are as follows: R = Par. lat. 7311 saec. ix (Regius), r = eiusdem man. sec. saec. x an xi incert.; O = Oxon. Bodl. Auct. F iv 32 saec. ix (Book I only); S_aA = Sangallensis 821 saec. xi (I. 1–230 only); A = Lond. mus. Brit. Add. 14086 saec. xi, a = eiusdem man. sec. saec. xi; w = all or most of the recentiores used (some twenty), = some or a few. Some individual recc. are mentioned by name.

^{page 242 note 1} Compare on the similar intrusion of -que Housman at Manil. 4. 776 and on the way of scribes with conjunctions in general Knoche, , Handschriftliche Grundlagen, p. 298Google Scholar, n. 3. To the list of conjectures at Prop. 2. 8. 8. perhaps may be added uincēris, uinces: haec, etc.

^{page 242 note 2} ‘quibus … haec res regatur finibus, nemo tradidit, neque dictu facile est’, Lobeck on Soph. Aj. 7; cf. Hey, , Archiv, xiv. 108Google Scholar, Fraenkel on Aesch. Ag. 504. The harshest enallage with a participle which I have been able to find is in Maiistas 16. 41 (Powell, , Coll. Alex. 69–70Google Scholar). The dissertation of Merz, A.,, Quatenus Ovidius et complures circa Messallam poetae enallages adiectivi quae vacatur figura usi sint (Vienna, 1914)Google Scholar, is known to me only by its title.

^{page 242 note 3} e.g. Cic. de Rep. 2. 26 sine depopulatione atque praeda, Verr. 2. 19. 47 socii praedae acrapinarum, Livy 38. 23. 2 consul captis castris direptione praedaque abstinet militem, etc. I am indebted to Mr. P. G. W. Glare for many of the references on which this discussion is based.

^{page 242 note 4} This passage seems to disable Roby's explanation (Lat. Gr. II. xxxix) of Livy 27. 44. 4 castra … praedae relicta as predicative dative (cf., however, Lucret. 5. 875 praedae lucroque iacebant); the dative seems much more like that at Cic. Fam. 4. 1. 2 urbem … sine fide relictam direptioni et incendiis.

^{page 244 note 1} positus = ‘reclining’ at Her. 4. 98, Met. 13. 638 (reading positique), Hor. Epod. 2. 65; = ‘served’ (of wine) a A.A. 1. 565 (Bacchus), 3. 350 (merum), al. On the repetition positis (229) …positi see 1. 338 n. below

^{page 245 note 1} Possibly as many as three meanings can be discerned in grauts: (i) ‘with wet garments’, cf. Met. i. a66, 4. 729, Virg. A. 5. 178; (2) = uino grauatus, on which sense see Thes. L.L. 6. 2. 2282. 79 ff. (cf. Lee on Met. 1. 224); (3) ‘troublesome’, for which compare nocet in 236 and my remarks.

^{page 245 note 2} Compare the remarks of Butler and Barber on Prop. 2. 12. 13–18 and 1. 9. 23– 24; and on the latter passage see also Davison, , C.R. lxii. 57–58.Google Scholar

^{page 245 note 3} Cf. Dilke at Stat. Ach. 1. 303.

^{page 247 note 1} At [Tib.] 4. 5. 16, whether quam or hane be read, queat cannot be apodotic; at Prop. 1. 19. 20 tum makes all the difference; at Lucan 10. 192 we cannot know that relinquam is subj. and not fut. indie.; Ovid, Her. 21. 183–4, 213–14 are simply not relevant.

^{page 247 note 2} I have never seen Pichon's work, which seems to be scarce in this country. This passage was kindly transcribed for me in Paris by Mr. I. D. Hood.

^{page 248 note 1} A puzzling case is that of Am. 1. 5. 1–2, which I have postponed discussing until now: aestus erat, mediamque dies exegerat horam; | adposui medio membra leuanda toro. Such repetitions as median … medio in adjacent verses can be paralleled in other poets: Virg. A. 1. 504–5,12. 883–4, Lucan 3. 621–2, 4. 426–7; but I know of no sufficient parallel in Ovid. The last line of the poem, proueniant medii sic mihi saepe dies, seems to echo the first (a fact which I had overlooked until a member of my lecture class acutely pointed it out to me), which makes it more likely that medio in v. 2 is corrupt than mediam. It is not difficult to think of corrections: either Burman's uacuo or uiduo would do very nicely; but medius is vox propria in this context, indeed it is almost true to say that in the erotic vocabulary medio toro iacere is a t.t. for ‘sleep alone’. Have we then a play on words? ‘It was the middle of the day, and I was lying in the middle of my bed.’ The figure tradnctio, the use of the same word in a somewhat differing sense, is dear to Ovid (see, for example, Owen on Tr. 2. 376); but is the difference between the spatial and temporal senses of medius striking enough to lend itself to rhetorical exploitation in this way ? Compare perhaps Juv. 11. m–12 uox | nocte fere media mediamque audita per urbem. The conceit seems a feeble one, but it is the best explanation I am able to suggest. On the whole problem of unconscious repetitions see Vahlen, , Opusc. Acad. i. p ff.Google Scholar, Cook, A. B., C.R. xvi. 146–58, 256–67Google Scholar, Jones, W. H. S.,, Proc. Camb. Phil. Soc. 181 (N.S. I). 19–22Google Scholar, Housman, , Lucan, p. xxxiiiGoogle Scholar, Poutsma, , Mnem. xli. 397–425Google Scholar, Laughton, , C.P. xlv. 73–83Google Scholar, Gries, , C.P. xlvi. 36–37Google Scholar, Jackson, J.,, Marginalia Scaenica, p. 17.Google Scholar

^{page 249 note 1} Vollmer, , Herm. lii. 467Google Scholar, defended quam at Rem. 325, but not even Bornecque has followed him.

^{page 250 note 1} Cf. Brandt, ad loc. The Romans insisted strongly on fides as an essential element in friendship: Cic. de am. 65 firmamentum autem stabilitatis constantiaeque est eius, quam in amicitia quaerimus, fides: nihil est enim stabile, quod infidum est. Sex. Rose, III in priuatis rebus si qui rem mandatam rum modo malitiosius gessisset… uerum etiam neglegentius, eum maiores summum admisisse dedecus existitmabant, 112 ergo idcirco turpis haec culpa est, quod duos res sanctissimas uiolat, amicitiam et fidem. nam neque mandat quisquam fere nisi amico neque credit nisi ei quern fidelem putat, de not. deor. 3. 74 inde tot iudicia defide mala, tutelae mandati pro socio fiduciae, Plaut. Most. 25–28 haecine mandauit tibi, quom peregre hinc it, senex? | hocine modo hie rem curatam qffendel suam? | hoccine boni esse qfficium serui existumas | ut eri sui corrumpat et rem et fdium ? For mandatum see Thes. L.L. 8. 267. 46 ff. ‘in sermone forensi iuris privati mandatum est contractus, quo amicus (procurator) iussu amici (mandatoris) huius negotia gratis gerenda suscipit’, and the further literature cited there. Ovid's penchant for legal phrases is notorious: Iddekinge's useful work (De insigni in poeta Ovidio Romani iuris peritia, Amsterdam, 1811Google Scholar) did not say the last word on this subject.

^{page 251 note 1} The elaborate series of transpositions proposed by Tolkiehn, J. (Festschrift für L. Friedländer, pp. 433 ff.Google Scholar, Neue Jahrbücher xi. 336 ff.Google Scholar) in Book I represent mere wasted ingenuity. Müller, , Rh. M. xvii. 532Google Scholar f. rightly observed that 2. 669–74 are out of place, but I cannot agree that the best place for them is after 2. 702. Damsté's transposition (Mnem. xxxix. 444) of 3. 487–8 to follow 490 is more attractive than most of his suggestions and is adopted by Bornecque, but it is hardly necessary.

^{page 252 note 1} Ael. uar. hist. 10. 18, Parthen. erot. 29, Diod. 4. 84, Seru. ad E. 5. 20, 8. 68 (Daphnis); Apollod. bibl. 1. 4. 3, [Erat.] cat. 32 (Orion).

^{page 252 note 2} Ael., Parthen., Diod., locc. citt. (Daphnis); Parthen. erot. 20, Σ Nic. ther. 15, [Erat.] loc. cit. (Orion).

^{page 254 note 1} I am bound to admit that in all the examples of os used as a collective singular that I have found in the Augustan poets it means ‘moudi’, not ‘face’. But cf. perhaps Ter. Heaut. 572 concsdas aliquo ab ore eorum, Cic. de rep. 3. 15 ut esset posteris ante os documentum, Phil. 8.20 ante osoculosque legatorum, etc.

^{page 254 note 2} I learn from Mr. H. C. Gotoff that testificere is found in five manuscripts, all of the fifteenth century, in Roman libraries: independent conjecture, or survival of tradition ? There is usually no means of knowing.

^{page 257 note 1} It seems to have been a feature of α, the common ancestor of ROS^a: I. 207 consistens S^a. 277 convenient RO, 423 inspiciens O, 729 hinc R, 738 amans O, 2. 466 habent R, 3. 280 ferens R, 771 sint R (see below), etc.

^{page 257 note 2} See Bonner, S. F., Roman Declamation, pp. 154 fGoogle Scholar. There is a good example at v. 397 of this book: quod latet, ignotum est: ignoti nulla cupido.

^{page 257 note 3} 1. 53 andromedon O, 327 obstenuisse 0,455 peroretur R, perhometur O, 514 labet ogo R, 723 undo O, 2. 79 noxosque R, 179 orbore R, 262 collidus R, etc.

^{page 258 note 1} Additional sigla in the Remedia: E = Etonensis Bl. 6. 5 saec. xi, Put. = Par. lat. 8460 saec. xii (Puteaneus).

^{page 258 note 2} Naples MS. II C 32 (c. 1400) contains extensive excerpts from a Greek prose ver sion of Ovid's amatory poems (Amores, Ars, Remedia). The excerpts from the Remedia were published, with some inaccuracies, by Schenkl, (Graz 1909), pp. 105 ff. I hope to discuss the value of this translation for the editor of the text in a later paper.

^{page 259 note 1} The idea was no doubt a rhetorical commonplace; it is allied to the proverbs about concordia (Otto, Sprichwörter, s.v.): cf. Aesop 53 Hausrath, Babrius 47, Plutarch, , de garrul. 17Google Scholar (511c), Publ. Syr. 289, Sail. lug. 10. 6.

^{page 260 note 1} I cannot forbear adding a postscript to the instructive history there retailed. In the summary of my note printed in C.R., N.s. viii. 195 for ‘first Aldine’ read ‘second Al– dine’.

^{page 260 note 2} It reads ‘735 coni. operae. multa libri.’ Besides envying Merkel his printer one should note the confident assertion (founded on complete ignorance) that all the manu scripts have multa. Just so did Müller, , Rh.M. xvii. 528 fGoogle Scholar. reprove Heinsius: ‘die abweichende Angabe des Heinsius ist irrig’, preferring to believe Merkel, who had never seen the reading of R at 1. 293. the passage in question, rather than Heinsius, who had handled the manuscript himself. But Heinsius lived in the seventeenth century and it stood to reason that he could not collate.

^{page 260 note 3} I cannot pretend that I find mis entirely compelling. Perhaps it is a mistake to trouble about the ductus litterarum; no correction based on quid of those kindly suggested by various friends has seemed to me convincing. Should one be bold and write (exempli gratia) tu caueas: actor eqs.?