1. Thus Timpanaro, Sebastiano, La genesi del metodo del Lachmann 2nd ed (1981) 13 n. 33Google Scholar. I am very grateful to Dr Timpanaro for an early copy The second edition brings his fascinating story up to date in many respects.

2. Bentley, equalled by few as a conjectural critic, was not only an emender of texts, although reading some recent anti-Bentleian tracts, one might almost think so Cf. my discussion in ‘Studi class. e. crit. test. in Inghilterra’, Annali della scuola norm. sup. di Pisa, Classe di lett. e filos. Ser. III Vol. 8.3 (1978)Google Scholar. section 4, ‘Horatian Poetry Thoughts on the development of textual criticism and interpretation’, Wolfenbütteler Forschungen 12 (1981) 11–14Google Scholar.

3. Nisbet, R. G. M. and Hubbard, M., in their commentary on Odes Book 2 (1978)Google Scholar have usefully taken further my discussion of ‘despised readings’. Agreement on some of these, however, is still to seek.

4. P. Maas's terms, longum and biceps have their uses in metres where one ‘long’ and ‘two shorts’ are interchangeable and are habitually interchanged But this, as Maas himself said, does not apply to all metres (Greek Metre, E.T., §33.2).

5. In two contrary cases such a syllable occurs in ‘fall’ at C. 1.15.36 (Glyc.) ignĭs Iliacos domos and 3.5.17 (Alc.) si non perirět immiserabilis. These two cases cannot convince as they stand The problems are not sufficiently focused in Lenchantin, M.'s discussion of the latter passage Rendic. R. Ist. Lombardo ser. III. 8 (1943–1944) 336–7Google Scholar.

6. All these cases are third-person verbs. The two remaining cases are not, and must therefore be called doubtful S. 2.3.1 would contain second person so used, scribis, if we could be sure of the wording; but this is a well-known textual crux and the variant reading scribes has been considered a strong claimant, especially by Bentley, ad 1. and Housman, , Papers I.153Google Scholar. S. 1.7.7 would contain an adj. so used, tumidus but the variant tumidusque (λ1 φψ) deserves attention.

7. As far as I can make out, the first editors, or commentators, explicitly to follow Bentley in correcting this slip were Cunningham (1721) and Sanadon (1728). Modern editors and commentators have however ceased to remark that the correction of the earlier vulgate, inuenient, is owed to Bentley. After all (they seem to think), the correct reading is here transmitted, and Bentley has become largely a whipping-boy to be punished for what is thought to be - and occasionally is - emendandi cacoethes. Yet there is surely no difference in principle between the critical sense that can correctly choose between variants, and defend, abandon, or emend a transmitted reading. Moreover, if I recall rightly, not a single recent commentator troubles to draw attention here to the Horatian idiom, the striking use of the sing As elsewhere, neglect of textual criticism merges into neglect of style This damaging dissociation is already inherent in Keller, O.'s note on the passage, Epilegomena (1879) 88Google Scholar For there Bentley's results are used, but no longer credited to him, presumably because the correct reading is in the manuscripts: ‘Die alte Vulgata invenient Ist eine unnötige Abweichung von der Lesart aller wichtigen Handschriften inueniet. Inueniet entspricht dem horazischen Gebrauch vgl. carm. I 3, 1–3.34, 9ff.’ Further remarks on the close connexion between textual and stylistic criticism will be found in the preface of Horace on Poetry Vol III (1982)Google Scholar.

8. Ott, E., Über die Congruenz des Prädikats mit mehreren Subjecten im Numerus bei Horaz, Schulprogramm Böhmisch-Leipa I–II (1887–1888)Google Scholar.

9. Thus the Syntax of Hofmann and Szantyr (1965) 433–5 There is at least a similar implication in the Satzlehre I (1912) 44Google Scholar, of Kühner and Stegmann, and I find it as early as 1828 in the Lat. Gram.⁶ of C. G. Zumpt, §323, perhaps also for the last time, in this context of handbooks, that the importance of Bentley's note on C. 1.24.8 is acknowledged There is hardly any attempt at sifting the evidence in Draeger's, A.Hist. Syntax² (1878) 147–9Google Scholar, but he, like Kühner and Stegmann, at least cites some evidence and does not restrict himself to premature generalizing as do Hofmann, and Szantyr, . Perhaps wisely, the great Madvig, Latinsk Sproglaere (1841)Google Scholar §213, and Riemann-Ernout, , Syntaxe d'accord ch. 4 of Syntaxe Latine (1925)Google Scholar, take up no position as regards usage in poetry and prose, or at different periods.

10. Draeger op. cit.; cf. Weissenborn and Müller on Liv. 1.6.4 Palatium Romulus, Remus Auentinum ad inaugurandum templa capiunt. Contrast 27.36.11 A. Hostilio Sardinia, C. Mamilio Sicilia, L. Porcio Gallia euenit., et al.

11. For the prevalence of pl. pred., see Draeger, , Ueber Syn. und Stil des Tac.³ (1882) §29Google Scholar, Constans, L., Ét. sur la langue de Tac. (1893) §§71–3Google Scholar, Nipperdey and Andresen¹¹ on Ann. 2.42.5 regibus defunctis, A. Gudemann on Dial. 42.3 criminabimur; for the sing Nipperdey and Andresen, , Ann. vol. II ⁶Google Scholar, Appendix to 12.2.2 aduenerat. Sörbom, G.Var. serm. Tac. etc. (1935) 72fGoogle Scholar., mentions some cases of variation. All this cannot replace a full discussion of the whole evidence. Goodyear on Ann. 1.42.1 faciat, 56.2 metuebatur considers some cases where the plural has been conjectured with a good deal of probability.

12. Horace apart, Cicero is the only author whose usage in this field can be said to be known As for Cicero, this knowledge derives from two reasonably full surveys, though now a little antiquated in their approach Anz, H., Ciceros Sprachgebrauch in d. Beziehung des gemeinsamen Prädicats bei mehreren Subjecten, Schulprogramm Quedlinburg (1884) 1–18Google Scholar, and Lebreton's, Jules well-known Études sur la langue et la grammaire de Cic. (1901) 1–31Google Scholar Both scholars divide the material into such groups as position of pred., types of subject (personal, collective, abstract) and of construction (connective, disjunctive, etc) The grouping is useful, but other aspects such as personal preference and levels of style may override it.

13. The examples cited from archaic comedy by Bennett, C. E., Syntax of early Lat. I (1910) 1–3Google Scholar, show at any rate that simple formulae will not explain the evidence.

14. It helps, by contrast, to see ancient usage in proper perspective, if it is remembered that in modern English the pl. predicate is the standard idiom in this collocation. Thus not only with personal subjects (‘Jack and Jill have now left’) but also with inanimate (‘the knife and the fork are on the table’) and with abstract (‘justice and fair dealing do not always prevail’) On the other hand the singular, in such collocations, sounds archaic or solemn or poetic; ‘the tumult and the shouting dies’, says Kipling in his Recessional. One would overstate the case only slightly if one said that in Latin, or at any rate in literary Latin, the contrary expectation prevails The sing. seems to be the common idiom, and the pl. does not, although, especially with personal subjects, it occurs not infrequently. But, here again, probable remains of the ordinary language would repay study.

15. In addition to the passage cited in the text, these are, S. 1.6.55, 2.1.65–7 (this case, as remarked in the text, is not noted by Ott; at 65 read et with cod. K al.; aut, which is also well attested, would unnecessarily conflict with 67 aut, in spite of Lejay's spirited defence; cf. Keller, , Epil. 517Google Scholar), 71–2, 3.227–30, 8.39–40, Ep. 2.1.5–9, 163, A.P. 273–4.

16. Ep. 1.14.22–8, A.P. 376–8; Ep. 2.1.71–2 express personal subjects by epical periphrases: uirtus Scipiadae, sapientia Laeli.

17. S. 1.5.40–1, 6.55, Ep. 2.1.5–9, 163, A.P. 273–4 In these and a few others the verb follows the subjects; it precedes them at S. 2.1.71–2 and 8.39–40 only.

18. The manuscript was not yet known to Bentley; cf. The ‘Ars Poetica pp. 3, 10.

19. Hence Keller, , Epil. 56Google Scholar, vainly cites Cic. Fin. 3.70 nec iustitia nec amicitia esse omnino poterunt, nisi ipsae per se expetuntur. But this is quite a special case for the very reason that moved Madvig ad. 1., and Lat. Spr. §213 b, n. 1, to defend the plural. Nor does Tac. Hist. 3.28 neque Antonius neque Hormus … degenerauere (cit. Nisbet and Hubbard ad. 1.) provide much support, both on inherent grounds, because the case is not at all alike, and seeing that Tacitus’ handling of the plural so differs from Horace's As for Greek, , Diggle, J., Illinois Class St. 6 (1981) 83–4Google Scholar, has shown that most (perhaps all) the alleged instances of pl. verb with disjunctives should be eliminated.

20. The pronoun, as Wakefield and Peerlkamp noted, would be like Prop. 2.1.25, addressing the same Maecenas, bellaque resque tui memorarem Caesaris Stat. Silv. 1.4.55 suo pro Caesare al. For the dat Peerlkamp compared Prop. 4.6.24 signaque iam patriae uincere docta suae; more examples could be cited.

21. Tragoediarum Delectus II (1794) 22Google Scholar.

22. Mueller in particular seems mistaken in joining the whole comparison closely with 18 qui maior absentes habet to the exclusion of 17 comes … metu.

23. Lines 19–20 adsidet … timet corresponds to 17 comes … metu next 21 magis relictis (timet) - 18 qui (metus) maior absentes habet and finally 21–2 non … praesentibus to 15–16 roges … firmus parum. assidere if one can judge by TLL II 877fGoogle Scholar., seems not elsewhere applied to a mother-bird sitting by or upon her unfledged brood; it may possibly be a newly formed metaphor.

24. So it has been explained in Fraenkel, E.'s important paper, Kolon und Satzi (1932) 209Google Scholar, reprinted Kl. Beitr. I 87Google Scholar.

25. Bentley however misreported Cruquius’ ‘Blandinius quartus’ as ‘veternmus Blandinius’, a minor slip There are many examples in Horace of the longer form uti. both in lyrics and hexameter poems. Naturally the unusual form uti is sometimes corrupted in the manuscript tradition as it is likely to be here Thus. S. 1.6.102 uti ne: ut me 2.7.79 uti mos: ut est mos cf. 8.10 ubi: ut.

26. Thus Meineke, Haupt, L. Mueller.

27. Thus A.P. 256–8 (iambus) spondeos … recepit / commodus et patiens non ut de sede secunda cederet etc.; cf. A.P. 11–12 hanc ueniam petimusque damusque uicissim; / sed non ut placidis coeant immitia, non ut serpentes auibus geminentur etc.

28. Markland on Stat. Silv. 1.3.86 citing Virg. A. 7.630 Tiburque superbum Mart. 6.42 superbus Anxur; cf. also Epod 7.5–6 superbas inuidae Karthaginis… arces.

29. It is printed by A. Y. Campbell (edd. 1945, 1953).

30. Adversaria Critica II (1873) 55Google Scholar.

31. But cupressos Klingner's 3rd edition.

32. If we are to believe Keller, , Epil. 369Google Scholar, we need not pay any attention to this ‘medieval conjecture’ But we simply do not know enough about ancient variants to deny the possibility of their survival in dubious cases The assertions of Keller's Textgeschichte are not certainties and its classifications anyhow unacceptable, as I have sought to show, ‘Ars Poetica’ Introd. ch. 2.

33. A. Y. Campbell (1st ed., 1945, ad. 1.) remarks amusingly as well as appositely, ‘mirabor si quis etiam auiam suam docere poterit exsugere aridum iecur’.

34. Proponents of exsucta include C. W. Mitscherlich (1800) in a very competent discussion, Meineke, who printed exsucta without comment, and H. D. Naylor (ed. 1922), noting that ‘the philtre requires dried up organs’. Macleane (4th ed., 1881) hesitated between exsucta and exsuc(c)a, on which see below. Housman 1893 (cit. in the text, p. 38) inclined towards exsuctus.

35. Other examples for exsucus lie between the evidence accepted for this word at TLL V.2, 1942 4ff., and TLL ibid. 1945 25ff. exsuctus. Two centuries before these examples became conveniently available, Bentley cited most of them In his note he pointed out that some of them, though thought to contain exsuctus, were more likely to contain exsucus Unfortunately, however, being convinced of the overriding virtue of N. Heinsius’ unlikely emendation exes(t)a, ‘eaten up, consumed’, he did not himself vindicate exsucus for Horace A case for this word is especially convineing when not only aridus but exsanguis is in the context, as Quint. I.O. 12.10.14 (Atticorum imitatores) nunc quoque andi et exsuci (-ucti -uti var ) et exsangues; and this case will draw Sen. Contr. 7.5.15 after it: declamator aridus et exsucus (-uctus var.). The reasons why in the Horatian passage exsuca and not exsucta should be read are set out in the text above.

36. Valart said, ‘Exsucca: ita regius codex 7977^us Habent Exsucta mss. tres, Exsecta duodecim, Execta quindecim’.

37. The only recent mention of the codex I have seen occurs in Lenchantin-Bo's list of the manuscripts containing the shorter Vita Horatiana (Vol. I, p. LXII)Google Scholar.

38. Not only Venice 1490, which Bentley mentioned, but also Florence 1482 and others, as was pointed out by Fea This may have been an editorial correction or was taken from some of the deteriores (not now noticed) on which the early prints were based.

39. Cf. responsum date.

40. Thus A. Y. Campbell.

41. So already noted by Fraenkel, 33 n. 1.

42. Fraenkel's structural notions are taken further by Schetter, W., Philol. 115 (1971) 249–55CrossRefGoogle Scholar Schetter notes that, to a certain extent, Fraenkel's analysis was anticipated by Plüss, T., Das Jambenbuch des Horaz (1904) 62Google Scholar, Olivier, F., Les Epodes d'Horace (1917) 33fGoogle Scholar., and also by a work I have not seen, Kirn, B., Zur lit. Stellung von Horazens Jambenbuch (Thesis Tübingen 1935), 13fGoogle Scholar. For a different approach see R. W. Carrubba's paper cited below.

43. Carrubba, R. W., Eranos 63 (1965) 161Google Scholar. My reason differs because I am not so sure that praeda and caper are ‘humorously one and the same thing’ There is after all an agna as well as a caper to be sacrificed to the winds.

44. This conclusion, up to a point, supports Carrubba's contention that the break after line 14 is stronger than is usually accepted. But it does not necessarily imply a bipartition of the whole poem.

45. An example of varying addresses, but irrelevant to the Horatian poem, is the central portion of Prop. 3.21 There socii are addressed at 11 nunc agite, o socii, propellite in aequora nauem, followed by uos … amici and tu … puella; 15–16 Romanae turres et uos ualeatis amici, / qualiscumque mihi tuque puella, uale.

46. C. 1.27.10–11 dicat Opuntiae/frater Megyllae is quite another matter. Nor will amici, nominative, do at all.

47. Housman cited CIL x 1403Google Scholar d 3.22 L.Amicius Fortunatus XIII 6385Google ScholarL. Amicius Donatus So does TLL I 1939, 34–7Google Scholar, adding Amicia from a lost inscription, vi 28147. I have not looked any further.

48. Jacoby, , F.Gr.H. 275 (Juba) 43–4Google Scholar, citing Plin. N.H. 6.202–5 (Sebosus and Juba), Ptol. Geog. 4.6.33 al.

49. For later Roman references, see TLL vi 1, 1197 25ffGoogle Scholar., for the notion Fischer, C. T., R-E vii 42Google Scholar.

50. Plut. Sert. 8.2 sailors, , report to Sertorius, 5 they make reference to Homer (cit. above); note the romantic touch, reminiscent, in expression, perhaps of Alexander's 9.1 ( v.1.) Cf. Sall., Hist. 1Google Scholar, frr. 100–3 (M).

51. Epod. 16.17–22 relate to the ‘Phocaean execration’ known from Herod. 1.165, Call. fr. 388.9 Pf., Suda Φ 635 et. al.; cf. Orelli ad loc. Hor., Pfeiffer ad 1. Call.; for Teucer's emigration in Horace, see C. 1.7.21–9.

52. ‘Meliùs antiquus interpres, circumuagus arua beata, vt de certa Oceani parte, qua insulas fortunatas ambit, loquatur Pöeta’.

53. TLL III 1176.67ffGoogle Scholar. only cites Laus Pis. 19 aethereae moles circumuaga flammae Stat. Theb. 10.112 adsunt… circumuaga somnia, usually printed circum uaga coll. Ov. Met. 11.613–14 hunc circa passim … / somnia uana iacent At Prud., Psych. 812Google Scholar the adjective is absolute, but at Hil. Trin. 8.48 circumuagus is qualified by prepositions, ex locis in loca.

54. circumfluus, in absolute usage known from Ovid and Silver Latin, is construed with the dative by Stat. Theb. 3.5 Styx inde nouem circumflua campis, but not known with the accusative before the late Epistula Alexandri ad Aristot… Jul. Val. p. 204.17 (Kübler) uel orbem terrarum circumfluum nauigare oceanum; cf. TLL III 1145 56ffGoogle Scholar.

55. The conventional punctuations seem to be both halting and awkward in different ways, whether circumuagus arua; beata/petamus arua diuites et insulas, or circumuagus: arua beata/petamus, arua diuites et insulas.

56. The parallels for aere … aere adduced by Keller, , Epil. 406Google Scholar, all differ. In particular anaphora amore … amore, Epod. 11.2–3, is not ‘wholly alike’; cf. above p. 44 top.

57. Vnxere a R u (post ras.) V δ(in ras.) π ψψ Luxere ABC λ¹l: Vinxere u (ante ras.).

58. In his editions, and as late as the third of 1959, Klingner prints what must be a mere slip – additum for addictum. He was obviously not trying to support the nonsensical reading, additum of ψ pr. (itself a mere slip afterwards corrected to addictum by the scribe), since he has no critical note.

59. I have noted T. Obbarius (ed. Jena 1848). Neither Meineke nor L. Mueller follow suit. Peerlkamp does, but seeks to improve on luxere by printing planxere. The only recent editor to impugn unxere is, as far as I know, A. Y. Campbell (1st ed. 1945), who is, not rarely, worth listening to when he criticizes the vulgate, but very rarely indeed when he emends the text. In his second edition (1953) this emendandi cacoethes had become grotesque.

60. Hom. Il. 24.582, 587.

61. Il. 24.720–3 For the part played by the women in the rites, see Alexiou, Margaret, The ritual lament in Greek tradition (1974) 10ff, 102fGoogle Scholar.

62. Thus, Il. 16.669, 679, 18.345, 350, 24.582, 587; for Roman evidence, see Enn. Ann. 155 (V²), Virg. A. 6.218–20, 9.485–7, al.

63. Il. 23.9 .

64. Il. 22.386, of Patroclus, Lambinus might have added the corollary feared by all that the corpse might fall a victim to birds of prey and scavenging dogs, as Il. 1.3–5 / ( var.), al.

65. Lambinus’ first example has, since Valckenaer, often been impugned as an interpolation from Soph. Ant. 29–30; the passage is Eur. Phoen. 1634 He compared Eur. Hec. 28–30, but also the consequential request for lament and burial, Androm. 1158–60, Med. 1377 He was learned enough to cite in addition Solon's supposed reproof (ap. Plut. Popl. 24.5) to Mimnermus on and thus to put in perspective Ennius’ remarkable nemo me lacrimis decoret nec funera fletu/faxit etc. He compared also the Hellenistic epigram Alc. Anth. Pal. 7.247 and such Latin counterparts as Virg. A. 11.372 inhumata infletaque turba.

66. As in another celebrated dispute amongst leading scholars of the time (cf. JRS 41 (1951) 36, 51Google Scholar for the case in which he seems to have been the injured party), Muretus was involved. This comes out very clearly in the rebuttal that closes the discussion in Lambinus’ first edition (1561): ‘Cum huius loci explicationem propemodum eo, quo nunc est, ordine dispositam Muretus apud me legisset, probauit ille quidem coràm: sed mutata posteá sententia, consilio suo dissimulato, neque mihi praesignificato, in libris Variarum lectionum conatus est ostendere utranque scripturam aequè probabilem esse’ This discussion is likely to have taken place during Lambinus’ second stay in Italy, 1555–60 (Sandys, Hist. of Class. Schol. II 189), when he had established himself at Venice (Sandys 149). The reference is to the first series of Muretus' Variae Lectiones (1559) 3.19 But the point argued there is irrelevant to the issue and certainly would not have been denied by Lambinus: ‘cum constet vngi solita apud antiquos cadauera’ - from which Muretus concludes, illogically, that either reading is possible. It does not then appear that the standard argument against luxere, which is noted above in the text, was urged by Muretus.

67. Special rituals were employed when cenotaphs were used (Hug, in R-E XI 171 fGoogle Scholar.). But they cannot be relevant to this issue; there is no hint of a cenotaph here.

68. The pun at Pl. Cap. 526 could denote ‘on behalf of’ rather than ‘in place of’.

69. Cic. Att. 10.8.7 Sardanapalli uicem … mori ‘like S.‘. Cf. Sall., Hist. 2Google Scholar (=4.67 Maurenbrecher) ceteri uicem pecorum obtruncabantur ap. Non. 9.497 (M.), in a section headed ‘accusatiuus uel nominatiuus ablatiuo’ There are indeed examples of uice, abl., denoting ‘like’, and if we are to believe Nonius, such was the normal usage. The two examples for uicem, ‘like’, appear to be completely isolated. Nevertheless they shield each other, in spite of the doubts that have been expressed, e.g. by Shackleton Bailey on Att. 10.8.7. The step from ‘in the place of’ to ‘after the fashion of’, ‘like’, is not larger in the case of uicem than in that of uice.

70. Thus Liv. 1.25.6 (on the text see below), 2.31.11, 8.35.1, 23.9.10, 25.38.3, 26.21.2, 28.19.17, 43.9, 34.32.6, 39.14.4, 40.23.1, 44.3.5.

71. Liv. 1.9.15, 3.36.3.

72. Cat. 64.68–9 sed neque tum mitrae neque tum fluitantis amictus / illa uicem curans. Doering (ed. 1788) implies the substantival notion, ‘condition’, with transitive curans which is clearly expressed by Ellis (ed. 1889) ‘uicem, here strictly substantival after curans, “what happened to”, Suet. Aug. 66 Vicem suam conquestus est’. Baehrens had been more judicious a few years earlier (ed. 1885), “uicem non curans” potes intellegere “sortem uilipendens”, ut Cic. ad. Att. 8.2 cuius ego uicem doleo Liuius 44.3 sollicitus eorum uicem habent, quamquam in his “uicem” rectius sumunt aduerbialiter, quod nostro loco malim, ut aequet siue “pro curare” (m. eta. curam habere) siue “de … curare” (=cogitare)’ Kroll, Fordyce, and other recent commentators are astray in taking the adverbial usage (‘on account of’) for granted without even mentioning the alternative.

73. Liv. 1.25.6 exammes uice unius quem tres Curiatii circumsteterunt Traces of the true text, uicem, are seen in uicet D, uices L, duly noted by Conway and Walters, who remark, ‘fortasse recte, sed nihil muto’ fortasse indeed.

74. Keller, , Epil. 411Google Scholar, ‘uicem … um die Uebereinstimmung mit der wirklichen lateinischen Sprache zu retten’.

75. Villeneuve (Budé) dubiously so: ‘uicem … nescio an recte; me tamen mouit codd. optimorum auctoritas’!

76. The comparative frequency of end-stopped couplets is a feature to which Fraenkel has drawn attention in the paper cited above, n. 24.