¹ Vol. liv, pp. 51–73; vol. lv, pp. 73–85. The main principle has been fairly often indicated; e.g. by Lindsay, W. M., The Captivi of Plautus (1900), pp. 359–60Google Scholar; Winbolt, S. E., The Latin Hexameter (1903), pp. 75, 127 n., 143, 238Google Scholar; Hardie, W. R., Res Metrica (1920), pp. 209–10Google Scholar; Murray, G., The Classical Tradition in Poetry (1927), p. 113 n.Google Scholar; Knight, W. F. J., Accentual Symmetry in Virgil (1939), pp. 8–9Google Scholar. See also Vollmer, F., Römische Metrik, pp. 11–12Google Scholar, in Gercke-Norden, , Einleitung in die Altertumswis-senschaft (1923)Google Scholar. Many German, and almost all French, authorities dissent.

I am indebted for references and suggestions to Messrs. E. Harrison, W. F. J. Knight, A. P. Sinker, and J. B. Poynton.

² His statistics for clash and harmony in each of the first four feet seem unnecessary and encourage the attribution of excessive self-consciousness to the various poets.

³ Cic., Or. xviii. 58Google Scholar; Quint, , i. 5. 30–1Google Scholar. See Lindsay, W. M., The Latin Language (1894), pp. 160 ff.Google Scholar; Westaway, F. W., Quantity and Accent in Latin (1930), pp. 59–81Google Scholar. The best recent conspectus, which shows how clearly scholars are divided on the lines of their native language into a French and an Anglo-German School, is by Leumann, M., in Stolz-Schmalz, , Lateinische Grammatik⁵ (1926), pp. 180–9Google Scholar. The chief rule is that the accent falls on the penultimate syllable if it is long, on the preceding syllable if the penultimate is short. A few details are disputed. In what follows I accept the view that enclitics like -que did not, in the Classical period, draw the accent onto a final short syllable (mégn que, not magn que), though they did onto a long; and that pentasyllables of the type had a secondary stress on the first syllable, not the second (cὄnticuére, not contἴcuére). This last is a case in which English accentuation differs from Latin.

(The sign ″ is used for a secondary accent.)

⁴ The French, having no stress accent of their own, generally refuse to allow that the Romans had one. They do not represent quantity in reading, as we and the Germans do, by putting a stress on the syllable that bears the ictus. Consequently Virgil, as read by a Frenchman, usually sounds like French prose to an Englishman or German.

¹ Leumann, (p. 185)Google Scholar inclines to think that it was pitch combined with a moderate stress, Westaway, (p. 65)Google Scholar the reverse. Of course ‘accent’ is only a convenient term for main accent, emphasis varying considerably over all syllables.

² See Murray, , op. cit., p. 82Google Scholar; Lewis, C. Day, A Hope for Poetry, p. 10.Google Scholar

³ Elementa Doctrinete Metricae (1816), p. 344 (217 in 1817 ed.).Google Scholar

⁴ Opuscula, vol. ii, Introd., p. xii.Google Scholar

¹ De Re Metrica ² (1894), p. 198.Google Scholar

² Only some uncouth inscriptions have spondees here.

³ Birt, T., Ad Historiam Hexametri Latini Symbola (1876), p. 7Google Scholar; Vollmer, loc. cit. E. H. Sturte-vant (Cl. Phil. xiv, 1919, p. 383Google Scholar) shows by statistical comparison with prose writers that Ennius, and his successors still more, must have gone out of their way to use words that ensured this conflict. Crusius, F. (Rōmische Metrik, 1929, pp. 45–6)Google Scholar puts the cart before the horse. Ennius, he says, had a difficulty; he had to get in many iambic, anapaestic, and choriambic words which could not be elided (as sédens, pédibus, praetéritos). The accent on these was bound to conflict with the ictus, so ‘he put them before a caesura, or a place where according to circumstances a caesura was permissible (2nd, 3rd, and 4th arsis)’; in this position the accent would upset the tempoless badly. This presumes that Ennius wanted to avoid conflict. But where eke could he have put such words? The great preponderance of masculine caesuras in Ennius compared with Homer indicates that he did not ‘have to get in’ such words, but purposely introduced them, as Sturtevant's statistics suggest. If not, it is due to the fact that such words are commoner in Latin than in Greek; they are indeed appreciably commoner, and Greek had the help of the Ionic forms, but these facts are not sufficient to account for the difference.

⁴ See Müller, , op. cit., p. 234.Google Scholar

⁵ Humphreys, M. P. (Trans. Am. Phil. Ass., 1878, p. 41)Google Scholar tried to show that the percentage of conflicts in the fifth foot in Ennius is about the same as in Homer and Hesiod read with a Latin accent, if allowance is made for the fact that words in and are more numerous in Greek than in Latin; and that Ennius therefore simply copied Homer. But Sturtevant's statistics (loc. cit., p. 379) suggest that Ennius definitely preferred to end with words of two or three syllables.

¹ As R. S. Radford maintained: A.J.P. xxv (1904), pp. 424–6.Google Scholar

² See especially Murray, Gilbert, op. cit., p. 114Google Scholar: ‘The last verse of an Alcaic is extraordinarily delightful in rhythm; but it would be nothing in particular if it were not reached by a struggle—and just the right kind of struggle’; cf. p. 112. Mr. P. F. Radcliffe tells me that conflict ending in harmony is a feature of Elizabethan madrigal.

³ Humphreys, , loc. cit., p. 44.Google Scholar

⁴ Virgil admitted more exceptions than Cicero and Catullus, in order to produce various effects. Winbolt, , op. cit., p. 127Google Scholar, gives the percentages of abnormal endings as: Ennius, , 14Google Scholar; Lucretius, i, 8½Google Scholar; Cicero, (Aratea), 2½Google Scholar; Catullus, , 2 (excluding Hymenaeus)Google Scholar; Virgil, , 3.Google Scholar

⁵ In his first two Satires Horace has 40 conflicts in the last two feet. ‘The result thereby produced is certainly striking, and, as he meant it to be, unpoetical.’ Munro, H. A. J., Trans. Camb. Philos. Soc., 1864, p. 394.Google Scholar

⁶ This was first observed by G. Cortius, who edited Lucan in 1726. It is familiar especially from Munro's Introduction to his Lucretius (1864), vol. ii, p. 105Google Scholar. It was ‘discovered’ in Germany in 1922 by Marx, F. (Abh. d. Sachs. Ges. pp. 197–232)Google Scholar, and is now known there as Lex Marxii. Crusius, (op. cit., pp. 50–1)Google Scholar thinks the object was to prevent a diaeresis in the middle of the line, though he recognizes conflict of ictus and accent as a secondary motive.

⁷ Figures are given by Knight, W. F. J., op. cit., pp. 38–9Google Scholar: Catullus, lxiv, 61 per cent.Google Scholar; Lucretius, 51 per cent.; Virgil, , Ecl., 37 per cent.Google Scholar, Geor., 33 per cent., Aen., 36 per cent. Cf. Hardie, W. R., J.Ph., 1907, pp. 266–79Google Scholar. In Horace's ‘neoteric’ Epode xviGoogle Scholar, of the 14 hexameters in the description of the Isles of the Blest (37–66) 12 have coincidence in the fourth foot. Mr. Knight, in his chapters V and VI, makes an interesting study of Virgil's use of such lines for special effect and his combination of them with ‘heterodyne’ types. His conclusions seem to me legitimate, allowing for differences of sensibility in given cases, and provided that too much conscious theory is not attributed to Virgil and that the method is not pressed too far.

¹ Crusius, , op. cit., p. 48Google Scholar, arguing against the ictus-accent theory, quotes the first 31 lines of Ovid's Metamorphoses and shows that here coincidence predominates almost as often as conflict. But (1) he admits that, if you read on, the case is different; (2) he allows no accent on any pyrrhic, whether or not it is immediately preceded by an accented syllable; (3) Ovid liked his Unes to be fluent, not complex, so he avoided coincidence less. Knight (p. 39) gives the percentage of coincidences in the fourth foot in Metamorphoses i as 49. This represents a notable reversal of Virgil's tendency.

² -que occurs in this position a few times; but there is some reason for supposing that, when it was elided, it did not alter the accentuation of the word to which it was attached; see Lindsay, , A.J.P. xiv (1893), p. 313Google Scholar. But it is rarely added to dactylic word-ends even in Cicero (26 cases in the speeches): Shipley, F. W., Cl. Phil. viii (1913), pp. 33–5.Google Scholar

³ 114 instances in Virgil. Full figures given by Eskuche, G., Rhein. Mus., N.F. liv (1890), pp. 385 ff.Google Scholar

¹ Edmiston, H. D., C.R. vii (1903), pp. 458–60.Google Scholar

² Harkness, A. G., Cl. Phil. iii (1908), pp. 41–58Google Scholar; Radford, R. S., A.J.P. xxv (1904), pp. 157, 160, 420Google Scholar; Broadhead, H. D., Latin Prose Rhythm (1922), p. 55 fGoogle Scholar. Harkness sums up: ‘Pyrrhic and iambic and monosyllabic words are not used in such a way that they could be accented as part of a word-group in cases where the word-group accent would be contrary to the verse rhythm. On the other hand we find these words employed in such a way that they naturally form word-groups where the word-group accent is in harmony with the verse-rhythm.’

³ Winbolt gives the facts (op. cit., pp. 137–43) and notes (with regard to Pyrrhics) that ‘in effect these conditions secure coincidence of verse-and word-stress’ (p. 137 n.).

⁴ Of the 155 cases in Ovid almost all are enclitics: Harkness, , loc. cit., p. 54Google Scholar. See also Norden, E., Aeneis VI ³ (1934), p. 448Google Scholar. Norden's Appendix IX, on irregularly formed verse-endings, is a most valuable collection of material and statistics.

⁵ Sandbach, F. H., C.R. xlvii (1933), p. 217.Google Scholar

⁶ Norden, , op. cit., pp. 447–8.Google Scholar

¹ Norden, , op. cit., pp. 438–46Google Scholar. Virgil has 85 abnormal endings due to the presence of Greek words. Sometimes, as in Atlantides abscondantur, the Greek word is not itself included in the last two feet, but has suggested a Greek rhythm. See also in general Maxa, R., ‘Lautmalerei und Rhythmus in Vergils Aeneis’, Wiener Studien, xix, 1897, p. 78 f.Google Scholar

² Lindsay, , The Latin Language, pp. 159, 161Google Scholar; Westaway, , op. cit., p. 72.Google Scholar

³ ‘Accent greatly weakened a following short syllable or two following short syllables’; Rad-ford, , loc. cit., p. 420.Google Scholar

⁴ Winbolt, , op. cit., p. 128Google Scholar. Cf. Norden, , op. cit., p. 441Google Scholar, on the effect of the ending lacrimisque (Aen. x. 505Google Scholar). It is by no means true that Latin has so few eligible words of four or five syllables as to account for their rarity at the end of hexameters.

⁵ Harkness, , loc. cit., p. 43Google Scholar; Vollmer, , Sitz, bayer. Akad., 1924, p. 17.Google Scholar

¹ Abh. bayer. Akad. (1886), p. 9Google Scholar. Cf. Leo, F., De Stati SiluisGoogle Scholar (Proem. Gott., 1892–1893), pp. 7–8Google Scholar. The Asiatic Dichoreus, Choriambus, and Ionicus a minore do not seem to have come under the ban, being conceived of as single feet.

² Cf. Diomedes, , p. 469 (quoted by Leo)Google Scholar: ‘Quamuis enim idem pedes eademque sint tempora, tarnen, ubi duae sunt paites orationis, nescio quo modo in utriusque confinio retentus Spiritus et restrictus adfert quandam compositioni firmitatem; at in una parte orationis properare nerba et continua Spiritus celeritate labi uidentur.’ Müller, , op. cit., pp. 242–3Google Scholar, suggests further aesthetic objections to such words as endings.

³ The main principle is indicated by Lindsay, , The Captivi, p. 359Google Scholar; Butler, and Barber, , Propertius, p. xviGoogle Scholar. Hermann was again the forerunner (op. cit., p. 227).

⁴ The Greeks too avoided it. In the rare exceptions in Greek (e.g. Theognis 456 and 520Google Scholar), and abo in Latin (e.g. Catullus, , lxxvi. 8Google Scholar; Martial, xii. 68. 6Google Scholar; Ovid, , Pont. i. 6. 26Google Scholar), the monosyllable goes closely with the preceding word. Even in prose a non-enclitic monosyllable was a uitiosa clausula; Bornecque, H., La prose métrique dans la correspondance de Cicéron (1898), p. 20Google Scholar. It was probably felt to cause too much of a bump; for Virgil certainly used it when he wanted to produce such an effect, as in praeruptus aquae mons, procumbit humi bos.

¹ Figures from Postgate, , Tibullus, p. liiGoogle Scholar, and Butler, and Barber, , op. cit., p. xvi.Google Scholar

² Not by Ovid, as has so often been said (e.g. by Plessis, F., Le Pentamètre Dactylique, p. 64).Google Scholar

³ Postgate, , Prosodia Latina, p. 85.Google Scholar

⁴ H. 9. 24; 36. Am.. iii. 7. 60Google Scholar. A.A. i. 736Google Scholar; iii. 150.

⁵ H. 3. 12;5. 98;6. 140;9. 106; 11. 96;13. 40; 15. 86, 206. Am. i. 10. 60Google Scholar; ii. 10. 14; iii. 11. 18. Rem. 30, 406Google Scholar. Med. 10Google Scholar. In later works, F. i. 28. 564Google Scholar. Tr. v. 10. 2.Google Scholar

⁶ Ovid, , Rem. 306Google Scholar, has a monosyllable in this position: institor, heu, nodes quas mihi non dat habet. But here the preceding monosyllable non enables coincidence to predominate slightly. Elision in the fourth foot producing conflict, as × / desiner (e) esse mea, occurs only five times in Ovid and never in Propertius or Tibullus. Whereas elision in this foot producing coincidence, as in / / poner (e) in aede morer, is allowed more often.

⁷ The hexameters of Nonnus and Musaeus are usually accented on the penultimate syllable. It has been held that they were imitating the Latin practice, the Greek accent having by then (4th cent. a.d.) become one of stress. This would be a strong argument for the ictus-accent theory for Latin verse, were it not that, as Otto Crusius pointed out, the phenomenon can be better explained by the growing desire to make the last syllable definitely long. (There is no lack of perispomenon endings.) See Rutherford, W. G., Babrius (1883), p. xviii n.Google Scholar

¹ Lindsay, , The Latin Language, p. 167.Google Scholar

² Lindsay, , The Captivi, p. 360Google Scholar; Harkness, , loc. cit., p. 43.Google Scholar

³ Catullus has 15 instances, Lygdamus 10. I have taken the pronouns to be emphatic when predicative, when contrasted, and in various other cases. Hardie, (op. cit., pp. 54–5)Google Scholar says: ‘It is a principle of symmetry and emphasis that a line should not end with an unimportant or otiose word, and this is even more clearly true of the ending of a couplet.’ That is so. But in the same paragraph he notes meus, tuus, suus as exceptions to the rule of not ending with an adjective. It looks as though Ovid thought the avoidance of conflict more important.

⁴ Select Elegies of Propertius, p. cxxviii.Google Scholar

⁵ Properties, p. viii.Google Scholar

⁶ Meyer, W. (Sitz. bayer. Akad., 1884, p. 1024 f.)Google Scholar says: ‘Wer möchte behaupten dass diese Regel nicht thöricht war?’!

¹ Characters and Commentaries, pp. 288–9.Google Scholar

¹ The figures are: Catullus 39 per cent., Propertius 23 per cent., Tibullus io per cent., Ovid 9 per cent. A long syllable was now rarely elided before a short, except in a few stock phrases like uidi ego.

² Schuster, M., Tibull-Studien (1930), p. 58.Google Scholar

³ Accentuation latine, p. 75.Google Scholar

⁴ Trans. Camb. Philos. Soc., 1864, p. 388 ff.Google Scholar, a detailed criticism.

⁵ Aussprache, Vocalismus und Betonung (1870), vol. ii, pp. 980–8.Google Scholar

⁶ Müller, , op. cit., p. 468Google Scholar; Langen, P., Philologus, xxxi (1872), p. 108Google Scholar; Shipley, F. W., loc. cit., pp. 31–2.Google Scholar

⁷ No doubt their rarity is due to the fact that in Virgil only words in -a are concerned; -que in the verse-position immediately preceding (as in diuisque uidebit) is extremely common (83 instances in Aen. vi alone). But it is a notable fact that -que is only attached to a trochaic word-end 34 times in all the speeches of Cicero (0·5 per cent, of its total occurrence), and that in 21 of these cases it is elided and has therefore probably no effect on accentuation; Shipley, , loc. cit., p. 29.Google Scholar

¹ Loc. cit., pp. 106–7.

² Aeneis vi, p. 437.Google Scholar

³ De Stati Siluis, p. 8.Google Scholar

⁴ Ibid.

⁵ Iktus und Akzent, p. 349 n.Google Scholar

⁶ Meyer, , Sitz, bayer. Akad., 1884, p. 1042Google Scholar; Müller, , op. cit., p. 247Google Scholar; Postgate, , op. cit., p. 86.Google Scholar

⁷ Zieliski, ‘Das Klauselgesetz in Ciceros Reden’ (Phüologus, Suppbd. ix, 1904), pp. 827 (242) ffGoogle Scholar. Smith, , A.J.P. xxv (1904), p. 461Google Scholar; Tibullus (1913), p. 99.Google Scholar

¹ Gercke-Norden, , loc. cit., p. 14.Google Scholar

² Metrik (1879), p. 209.Google Scholar