¹ In his edition of the Pharsalia (1816), pp. 22–4Google Scholar; cf. Norden, E., Aeneis Buch VI³(1934), P. 389.Google Scholar

² e.g. Marouzeau, J., Traité de stylistique latine, pp. 274 ff.Google Scholar

³ Beobachtungen über Vers- und Gedankengang bei Lucrez (1936), pp. 47–103.Google Scholar

⁴ Apart from the fact that these words disturb any true enjambement, it is clear from Quintilian 11. 3. 108, quoting pro Lig. 1, and from Cicero himself, quoting from pro Com. in Orator 225, that such words form an integral part with other words of incisa or phrases.

⁵ T.A.P.A. lxxxii (1951), pp. 241–2.Google Scholar

⁶ Pap. Ox. 30, dated by Mallon, (Emerita xvii (1949), pp. 1–8) to about A.D. 100.Google Scholar

¹ Pap. land, go, published by J. Sprey (1931); also Lowe, , C.L.A. viii. 1201Google Scholar, and Cavenaile, , Corp. Pap. Lot. (1958), pp. 70–1.Google Scholar

² By what might be a remarkable coincidence, a similar example of K is reported as standing at the end of a paragraph at the beginning of Cic. Verr. ii. 4. 32 in the ninthcentury Cod. Reg. Par. 7774^A (so Orelli²-Baiter-Halm, ad loc). This suggests that the Verrines, as a favourite text for rhetorical study, were so marked throughout (or in some selected passages?) by a teacher whose copy has uniquely affected the MSS. tradition. Two other uses of K are reported: inPap.Soc. It. ii. 142 (Lowe, , C.L.A. iii. 289Google Scholar dates it third/fourth century), a Virgilian cento based on Am. 1. 477 ff., where a K with extended tail straddled by a pair of dots is used at the end of every verse in the first of two columns (regardless of sense, if any) and apparently at the end of the last verse in the second column; and, according to Weber, R., in Scriptorium ix (1955), pp. 57–63, in various eighth-century (and perhaps earlier) English MSS. of the Vulgate to indicate mistakes in colometry, and also general errors of division. Whedier K in these contexts signifies kaput (presumably for the beginning of a new sentence rather than the end of the previous one) or kolon/komma (Weber) or even klausula, there is clearly no regularity of practice. For material in this note, and for other palaeographical assistance, I am indebted to Dr. J. D. Thomas and Mr. J. E. Fagg.CrossRefGoogle Scholar

³ This is clear in the clause given as cuius in Sicilia virtutem hostes misericordiam victi / fidem ceteri Sicuti perspexerunt.

⁴ Pap. Herc. 817 (C.L.A. iii. 385)Google Scholar. Zange-meister-Wattenbach, p. 1, pl. 3, has an indecipherable photograph; most transcripts (e.g. H. Degering, Lettering, pl. 7) do not show diacritical marks. Thompson, E.M., Introduction to Greek and Latin Palaeography (1912), p. 276Google Scholar, and Steffens, F., Lateinische Palaographie (1964), pl. 3Google Scholar, give complete transcriptions of part of the poem.

⁵ Pap. Herc. 1067 and 1475 (C.L.A. iii. 386–7).Google Scholar

⁶ Pap. Ox. 2088, dated by Mallon, , Pap.Rom. p. 177–8, as late first century A.D.Google Scholar

⁷ C.G.L. iv. 484Google Scholar. 26–7K, v. 132. 1K, the former discussing the same sort of punctuation that is found in the Medicean of Virgil.

⁸ Hist, de l'éducation dans l'antiquité (1948), P. 553 n. 30.Google Scholar

⁹ pro Mur. 25. Certainly the insertion of puncta at this period appears to have served only to separate words.

¹ vii. 9. 5.

¹ Epp. 40. 11. For a survey of interpretations, see Andrieu, J., Rev. ét. lat. xxiv (1946), pp. 296–8Google Scholar, with his own conclusion that the reference is to the division between words in writing.

³ 1. 5. 27, 11. 3. 52.

⁴ Gram. 24. 3.

⁵ See Bonner, S.F., Hermes lxxxviii (1960), pp. 354–60.Google Scholar

⁶ This conclusion tallies with the normal assumption of the handbooks: e.g. Thompson, , op. cit., p. 61Google Scholar; Kenyon, F.G., Books and Readers in Greece and Rome (1932), p. 67Google Scholar. It is unfortunate that Norden, in his important note on the nature of ancient punctuation (op. cit., p. 386)Google Scholar, does not really consider the problem of its continuity in the tradition of MSS.; and Pearce, T.E.V., in C.Q., N.s. xvi (1966), p. 145Google Scholar, quoting M as evidence for ancient punctuation, does not ask whether this represents Virgil's own practice. For a typical editorial view see Dilke, O.A.W.'s Horace, Epistles I³ (1966), p. 35 n. 2Google Scholar: ‘Editors can use their own judgment in punctuation.’

¹ I shall refer mainly to Mediceus (M), reproduced by Rostagni (1931); Palatinus (P), by Sabbadini (1924); and the Schedae Vaticanae (F), by Ehrle (1899, 1930). For other authors see Lowe, E.A.'s list in C.Q. xix (1925)Google Scholar, showing very few indications of punctuation, or even word-division.

¹ So Rostagni, , op. cit., pp. 8–10.Google Scholar

³ Yet he never contemplates the improbable break in 2. 433, vitavisse vices, Danaum et si fata fuissent, recommended by Peerl-kamp, Nettleship, and others; nor does he consider any variants at 7. 298, fessa iacent, odiis aut exsaturata quievi, as MP and modern editors have it.

¹ Cunningham, M.P., in Proc. Am. Philos. Soc. ci (1957), p. 504Google Scholar, enunciates a general rule for determining phrasing in Latin prose: ‘Distinctio of some sort occurs at points where what follows does not go closely in grammar with what precedes’; and conversely: ‘Conti-nuafiois normal between words when thesecond goes closely in grammar with what follows’. It will be found that our major dilemmas concern adverbs, whose relation to what precedes is precisely the question at issue.

¹ I am thinking of such follies as the punctuation attributed by Servius to the Vergiliomastix (in jest?) at Ecl. 2. 22, lac mihi non atstate novum, nonfrigore. defit, or his own perverse preference at Ecl. 3. 108, non. nostrum inter vos tantas distinguere lites. Little more legitimate appear such readings as the commonly accepted sera, tamen pietas in Prop. 3. 15. 35, found in most texts since Müller (1885).

¹ This would also have the advantage of making the inanimate (ova…terram) subject of the infinitive in either half of the sentence.

¹ Elsewhere turn resuming after a cum clause stands first in the apodosis, except thrice (2. 44, 6. 281, 6. 1151), where it is preceded by an emphatic word, always in the same line.

¹ Cf. 5. 1010, and Bailey's Prolegomena V.B. § 8, p.96.

² For the form of the phrase thus given cf. 2. 886, animum quodpercutit ipsum; though no enjambement is there involved.

³ The T.L.L. (C. 1114) records the phrase as unique. Lachmann was so suspicious of it as to read victum for circum, with quod causal. Heinze explains it as a conflation of the familiar ideas seen in 5. 931, solis volventia lustra, and in 1. 311, solis redeuntibus annis; but it is not clear that Lucretius mixes ideas in this way. A closer parallel, given by Heinze, is Manilius 3. 524, sideris… re-deuntis in orbem. Heinze's verdict, ‘bold, but not liable to misunderstanding’, takes no account of the position of circum in the verse. In Lucr. 2. 1107–8, addita circum semina, the adverb belongs with the preceding participle; but the whole phrase in enjambement is bound closely together.

¹ Virgil appears to have no examples of a break before a phrase of this sort.

² As in 3. 79–80, mortis formidine vitae / percipit humanos odium lucisque videndae. Bailey unnecessarily separates off mortis for midine with commas.

³ Nevertheless, it should be observed how often a single passage in Lucretius does in fringe the poet's normal metrical or prosodi-cal practice: e.g. 5. 949'50, omnia deber' / ut (his one hypermetric line); 6. 755, natura loci opus efficit (his one hiatus); 3. 198, coniectum spicarumque (plausibly accepted by Bailey, who observes that everywhere else a spondaic fifth foot is preceded by a dactylic fourth); 3. 493, anima spumas (Tohte's emendation, followed by Bailey, giving Lucretius' sole example of a final short vowel making a long syllable before initial sp, etc.). A unique transgression of his observed practice can never automatically be dismissed as impossible.

¹ In his note on Lucan 1, 231 (see p. 330 n. 1 above).

² Always including a conjunction or relative (as qui se, at tu, si quern), so that the run of the sense is unmistakable. The only exceptions are quoted below.

³ For a similar example of see Norden's note on Aen. 6. 358, paulatim ad-nabam terrae iam tuta tenebam, where Servius knew of punctuations before and after terrae. Norden rightly sees that terrae (as dative) cannot be separated from adnabam; but it is also essential (as genitive) with tuta, as the alliteration emphasizes.

¹ R. G. Austin (whom I wish to thank for many hints and corrections in this paper) points out that aret introduces the third member of a tricolon, the preceding members of which each begin likewise with emphatic words constituting the predicate.

² Op. cit., pp. 400 ff.Google Scholar

³ Elsewhere in Virgil -ne is attached over 90 times to the first word of a sentence or clause (cf. especially Aen. 1. 132, 10. 845, both beginning with tantane); only once to the second word, in 2. 597, superet coniunxne Creusa / Ascaniusque puer? (on which see Austin's note). There is no question there of enjambement, and coniunx carries considerable weight.

⁴ Contrast 2.348, fortissimo frustra / pectora, where the adverb is sandwiched in the manner described in the next paragraph.

⁵ So Austin on Aen. 2. 18: ‘Such “indifferent’ words belong usually to the following line and do not find their full force till then.’

¹ Perhaps only in 2. 1107–8 addita corporasunt extrinsecus, addita circum / semina, where the relation of extrinsecus to the first addita determines that of addita circum; and in I. 66–7, mortalis tollere contra / est oculos ausus primusque obsistere contra, where the sandwiching is not close, and there is certainly no ambiguity.

² About half the texts I have inspected print a comma after attollere; most translators follow suit, as Stanyhurst, ‘Feeble agayne weixing she droups’; Mackaii, ‘Swoons back’; Rhoades, ‘Sinks back swooning’; and the Budé, ‘Et de nouveau s'éva-nouit’. No text appears to print a comma at the end of the line; but Austin prefers to break the sense there, translating ‘rursus’ as ‘in turn’ or ‘to meet her’, while Paratore likewise has ‘tenta risollevarsi’.

³ But the strict sense of rursus cannot be pressed in Aen. 3. 31, 4. 557, 11. 621, 12. 571.

¹ The suggestion that interlinear hiatus (as here ultro / hortantem) might be an argument against linking the two words too closely proves to have nothing in it: cf. Aen. 1. 13–14, Tiberinaque longe / ostia.

² For a similar self-contained line sandwiched between present participles, cf. Aen. 2. 380–1, (anguem) repente refugit / attollen-tem iras et caerula colla tumentem, based on Geor. 3. 421, quoted here. In neither, however, is the first participle entirely on its own, as hortantem will be here if ultro is taken exclusively with inferre.

¹ Gear. 3. 501 remains the one place where punctuation seems essential, if only to explain the complete lack of misunderstanding in antiquity. Despite Oliver's theory, our evidence is totally inadequate to indicate what sign Virgil would have used at this point if he wanted to mark a break.

² I am unaware of any work on the effect of alliteration, in Virgil and elsewhere, on rhythm and phrasing in verse. It struck me some years ago, when considering a problem of the rhythm of the fourth foot of the hexameter (A.J.P. lxxi (1950), p. 374), that an isolated spondaic word in this foot was often related alliteratively to the word occupying the fifth foot.Google Scholar