¹ Bernhardi, K., ‘De incisionibus anapaesti in trimetro comico graecorum’, Acta Soc. Phil. Lipsiensis, i (1872), 245–86,Google Scholar of which pp. 245–76 = Dissert. Lipsiensis (1871), pp. 3–34.Google Scholar His rules may here be briefly stated: (i) Polysyllables are not terminated after the first or second short of an anapaest in the third and fifth feet, but may be in the second and fourth, provided there is ‘close connexion’ between the words involved in the split anapaest, and that there is penthemimeral caesura, (ii) Mono- and disyllabic enclitics, and the particles may not begin an anapaest, except in the second and fourth feet, under the same conditions as polysyllables. Other mono- and disyllables may occur in an anapaest with some freedom, but except for the article, prepositions, and oaths such as , their occurrence in feet other than the first is allowed only if they follow a punctuated pause, (iii) A strong pause does not occur after the first or second short, exqept in the first foot, (iv) Elision of a polysyllable is allowed freely after the first or second short of an anapaest in all feet except the last; but punctuation there is allowed only in the first, second, and fourth feet.

² Snell, B., Griechische Metrik (Göttingen, 1955) P. 11:Google Scholar ‘Ausnahmen etwa alle 700 Verse einmal’.

³ On Ar. Acharnians 178.

⁴ Perschinka, F., ‘De mediae et novae quae vocatur comoediae atticae trimetro iambico’, Dissertationes Philologae Vindobonenses, iii (1891), 355–60.Google Scholar

⁵ White, J. W., The Verse of Greek Comedy (London, 1912), pp. 45–48;Google Scholar the quotation is p. 47, n. 1.

¹ See especially his note, p. 140 of his ed. of Aristophanes, Acharnians, (on. 178). His great contemporary editor of Aristophanes, W. J. M. Starkie, tended unfortunately to make the same mistakes as Bernhardi. The rules he gave in his edition of the Wasps, pp. xxxvii–ix, require the postulation of corruption in Vesp. 25, 967, 1369 and Lys. 927 (where all the manuscripts agree, providing texts diat it would be difficult to ‘improve’); special pleading is needed to defend Av. 20 and Thesm. 637; in Nub. 62 and Ran. 1307 it seems difficult to explain how the readings of R and U respectively could have become corrupted; Elmsley's in Ach. 1078 would give us a spelling unparalleled in Aristophanes for the plural of ; in Nub. 876 without a word intervening is not impossible: see Denniston, , Greek Particles, ed. 2, p. 564;Google Scholar only at Av. 93 (intrusion of article into the manuscripts: see p. 197, n. 1) and at Nub. 238 ( is difficilior lectio: is over twice as frequent in Aristophanes) is Starkie justified.

² Descroix, J.-M., La Trimtire Iambique, Mâcon, 1931), pp. 210–11.Google Scholar I find that there is another paper whose title suggests that it may have touched on this subject: Sacht-schal, , De com. gr. metro accommodato (Vratislava, 1908).Google Scholar I have, however, not come across a copy.

³ This does not appear to have been sufficiently emphasized. A fair sample even of punctuated pauses after the second short can be given; Bernhardi himself admitted Sotades 2. 447–8, 1 v. 23 (2nd foot), and Ar. Aves 1226 (4th foot); Perschinka noted Dionysius 2. 423–4, 2 v. 19 (2nd foot); further examples are Ar. Lys. 731 (2nd foot), Plut. 417 (3rd foot), and Menander, Pk. 184 (3rd foot). It would have been attractive to add also Menander fr. 11 v. 2 (Demiańczuk), but the editor appears to misinterpret the fragment: see Körte-Thierfelder, , Menander ii fr. 368: yet cf. p. 293.Google Scholar

¹ Hence the text of a fragment cited by Athenaeus, for all that he is a careful excerptor (see Zepernick, K., Philologus, lxxvii (1921), pp. 311–63),Google Scholar is always more liable to suspicion than a passage of Aristophanes with a unanimous textual reading, or, for that matter, than die papyrus text of a con tinuous passage of Menander. For a neat example of diis see Maas, P., Textkritik, 2nd ed. (Leipzig, 1950), p. 24, para. 40.Google Scholar

² For my computation I include in fr. 107 v. 1, since Leo's conjecture )Hermes, xliii (1908), 308),Google Scholar which completes the line, seems most plausible; in fr. 135 v. 6 I read , vide infra; fr. 155 v. 3 is to be scanned , see p. 194, n. 3; in frs. 186 v. 1 and 189 v. 3 there is slight corruption, but it seems likely that we have split anapaests in the fourth and first feet respectively; I have included also fr. 340, where there is doubt over the attribution. Excluded are fr. 25 (see below, II. 2, and p. 192, n. 1), fr. 77 v. 3, where are the words of Adienaeus, not of Alexis; frs. 16v. 11, 187 V. 6, and 279 v. 4, which are discussed below with fr. 25 v. 5, are excluded, as well as the papyri fragments (Pack, nos. 1291 and 1293) which have been attributed to Alexis on unconvincing grounds.

³ All references to comic fragments, unless it is otherwise stated, are to Kock, C.A.F., except tiiose of Menander (Korte– Thierfelder).

⁴ It will be seen that fr. 210 v. 3, cited as an exception by Descroix, pp. 220–1, is allowed under Bernhardi's third rule, and hence does not require discussion: Perschinka, p. 358, cites as parallels Antiphanes 2. 124, 275 v. 1 and Epicrates 2. 282–3, 3 v. 9; add Ar. Nub. 3: all these, instances of elision in the third foot. Fr. 222 v. 7 is cited by Perschinka, p. 355, as an exception; but similar first-foot interpunctuated breaks occur even in Menander (Epitr. 384); and cf. Ar. Vesp. 998, Lys. 731, Ecc. 1056, Ran. 1462 (all these cases before ).

⁵ Kock in fact prints but the common orthography is : see L.S. J., s.v. .

¹ Athenaeus 6. 325 a: ed. G. Kaibel (Leipzig, 1887–go), ii. 6; S. P. Peppink (Epitome) (Leyden, 1937), i. 85.

² But rightly printed by Coulon, and defended by Fraenkel, E., Aeschylus' Agamemnon, iii. 579 (on w. 1256 f.).Google Scholar

³ 2nd ed. revised by Kraus, W. (Vienna, 1954), pp. 312 f.,Google Scholar on v. 1220. Cf. also the remarks of Elliott, , loc. cit., p. 189, n. 1.Google Scholar

⁴ Or, for that matter, Goraes' (in Schweighaeuser's ed. of Athenaeus) , which also splits the anapaest, while giving less acceptable sense than Dobree's.

⁵ Op. cit., p. 256.

¹ C.Q. xlix (1955), 210 ff.Google Scholar

² See Descroix, , op. cit., p. 31,Google Scholar and Kühner-Blass, i. 227–8.

³ Athenaeus iii. 104 d: Kaibel, i. 239; Peppink Epit. i. 21.

⁴ The trouble is mainly that Bernhardi nowhere defines what he means by this ‘close connexion’; and so leaves room only for a subjective interpretation of his words. Most probably we must assume the reference is to a degree of closeness less strong than that of enclitic, proclitic, and equivalent words, but at least as strong as the relationship between noun and adjective (, II. 4 of my group of passages) or between verb and adverb (Menander Samia 97, ’).

⁵ Certain fourth-foot examples are Ar. Ach. 107, Cratinus 1. 89, 250 v. 2, Philemon 2. 536, 228 v. 2, Menander fr. 397 v. 3 (KÖ.-Th., p. 141), without punctuation; Ar. Aves 1226 and Eubulus 2. 166, 7 v. 9 with punctuation; and Ar. Pax 233 with punctuation and change of speaker.

¹ Athenaeus 10. 443 d: Kaibel, ii. 464; Peppink Epit. ii. 42.

² Athenaeus 4. 164c: Kaibel, i. 370.

³ As suggested by Naeke, , Choerili Samii quae supersunt (Leipzig, 1817), p. 5.Google Scholar

⁴ Peppink Epit. i. 53.

⁵ Hence we shall reject any conjectures that introduce new names in place of those in the manuscript, such as Erfurdt's for .

⁶ Add. Ath., pp. 103 f.

⁷ Two lines sometimes cited as parallels [Ar. Thesm. 1184, Ran. 1393] are not lists; on the former, in any case, representing as it does a Scythian's broken Greek, no arguments can safely be based, while the latter has been emended with some plausibility to : see Radermacher and Coulon, ad loc. Certainly Aristophanes seems normally to avoid amphibrach rhythm at the beginning of a trimeter, as Starkie, (op. cit., p. xxxviii, n. 3) shows.Google Scholar

¹ On such transpositions in our manuscripts see Headlam, , C.R. xvi (1902), 243 ff.Google Scholar

² Athenaeus 7. 302 f: Kaibel, ii. 167; Peppink Epit. i. 135.

³ The second syllable of is always long whenever the word occurs in a non-corrupt and checkable passage of comedy: AT. frs. 247, 318 ( at end of the line); Ephippus 2. 251–2, 3 v. 9 (a passage almost identical with Eubulus 2. 214, 150 v. 6: possibly Ephippus borrowed lines from Eubulus, or possibly Athenaeus made an error in one of the two places con cerning die attribution, citing Eubulus at 2. 65 c [Kaibel, i. 153], and Ephippus at g. 370 c, d [Kaibel, ii. 309, Peppink Epit. ii. 4, where the epitome’s text is incorrect]); Ephippus 2. 258–9, 15 v. 4 (where vv. 3 and 4 are again almost identical with Eubulus 2. 204, 110 w. 1 and 2, except that in die second line—the relevant one—the text of Eubulus is . If this is to scan we must allow the solecism of with a long antepenultimate; the line dierefore is probably corrupt, and ought to be restored, as L.S.J, suggests [s.v. ], to , which is the corresponding line of Ephippus).

Petersen, W., Greek Diminutives in -ION (Weimar, 1910), p. 217,Google Scholar suggested an interesting rule for die length of the iota in the antepenultimate of diminutives—that the iota is lengthened in comedy when it bears the ictus; this rule does not always work, however. For instance in Ar. fr. 204, the ictus is on die second syllable of , but it remains short; cf. also Ar. Pax 202, Ran. 60, Antiphanes 2. 23, 33 v. 4, and Menander fr. 793 KÖ.–Th. (= 765Kock), in all of which places we should be compelled to postulate both irregular broken anapaests and the anomaly of allowing the antepenultimate iota to be long here alone and nowhere else for die word in question, if we followed Petersen's canon. In fact the reasons which govern die lengdiening and shortening of the antepenultimate iota in diminutives are by no means as simple as Petersen would make them. Often with words such as , where there is an iota in the stem of the primitive, contraction to may have been felt to take place (Schwyzer, , Gr. Gr. i. 471),Google Scholar and once some diminutives had been formed on this principle, popular analogy might extend the lengdiening to other nouns without an iota in the primitive stem. Vice versa, when we meet such formations as , it is not impossible that common usage irrationally pronounced the antepenultimate iota short from a fallacious analogy of words such as , where the primitive has no iota in the stem.

¹ Op. cit, p. 218.

² The example that might be quoted from tragedy ( in fourth foot, Eur. Iph. A. 1596) is probably a Byzantine forgery: see ProfessorPage, D., Actors' Interpolations in Greek Tragedy (Oxford, 1934), p. 196.Google Scholar Certainly, apart from the Iph. A. and the Cyclops, and excluding proper names, we do not find anapaests—let alone split ones—in Greek tragedy except in the first foot; for Eur. Heraclid. 223 is desperately corrupt (Murray in O.T., and Zuntz, G., Political Plays of Euripides [Manchester, 1954], p. 108).Google Scholar

³ Athenaeus 9. 386 b, c: Kaibel, ii. 342.

⁴ Athenaeus 7. 324 c: Kaibel, ii. 213; Peppink Epit. i. 150.

⁵ This must be the way to interpret ; it is the only word in the text as we have it which can govern , and cannot (the latter verb would require ).

¹ e.g. Ar. Aves 441, 1644, Ran. 652, 658, Eccl. 998; Plato Com. 1. 607, 28 v 1 (cited by White, J. W., op. cit., p. 48,Google Scholar or Descroix, , op. cit., pp. 220–1):Google Scholar passages sometimes wantonly altered by Bernhardi, against a unanimous manuscript tradition.

² In the Loeb edition of the passage (see his translation, and n. b on it).

³ Athenaeus 6. 223 e (citation only in A (and derivatives), not in the Epitome): Kaibel, ii. 3.

⁴ Diar. Jenens., Dec. 1817, pp. 393 ff. This article I have not seen.

¹ For we may not cite cases such as Ar. Nub. 1192 or Aves 90 (Descroix, pp. 214, 220, White, p. 47), where a change of speaker is involved, the first short of the anapaest consists of a verb ending with nu-ephelky-stikon, and the next word begins with a vowel: on these cases see Radermacher on Ar. Ran. 1220 (cited in my n. 3, p. 191). Ar. Aves 93, cited by White, p. 47, is probably to be read without the article before (so , followed by Coulon; RMVU, where the article may be the addition of a ‘prosifying’ scribe [see Headlam, article cited p. 194, n. 1]). In Ar. Aves 23 I see no objection to the ‘’ of , and is there of course to be considered as a quasi-proclitic (Bernhardi, pp. 261–2). Finally, the text of Antiphanes 2. 71, 148 v. 6 is either desperately corrupt, or else a paraphrase merely of the poet's words (cited by Clemens Alex. Paed. 3.2/7, Stählin i. 240).

² Athenaeus 10. 429 e: Kaibel, ii. 434; Peppink Epit. ii. 33.

³ These cited variously by Perschinka, p. 356, White, p. 46, Descroix, pp. 220–1.

⁴ Athenaeus 4. 164 f: Kaibel, i. 371; Peppink Epit. i. 53. Kaibel prints his own conjecture in the text, but gives the manu script reading in his apparatus.

⁵ Cited by White, p. 48, Descroix, p. 220. Cf. also possibly Strattis 1. 728, 60 v. 2 ( begins fourth foot), though the sources for this verse (Aid. Scholia of Ar. Ran. 303, and Scholia of Eur. Orestes 279) are not of the most reliable; and Ephippus 2. 260, 17 v. 1 ( begins third foot), though the end of the line as we have it is clearly not what Ephip pus himself wrote.

¹ Denniston, , Greek Particles, 2nd ed. (Oxford, 1954), p. 291;Google Scholar cf. also pp. 157 f.: ‘The effect of is to stress the addition made by .’ Among Denniston's citations is S. Ajax 1376,

¹ See on p. 197, n. 1 with reference to Ar. Aves 93. Quotation for such a common trait is superfluous; but one instance may be worth noting: the partiality of Triclinius' text for adding the article before nouns in the choruses of the Agamemnon.

³ 2. 63 e, f: Kaibel, i. 149 (both C and E).

⁴ pp. 26–29. The list contains all the defensible instances in comedy, with one possible exception: after the unaccented . That hiatus is allowed after I hope to show elsewhere.

⁵ In his edition of the Anth. Pal. viii. 234.Google Scholar