^{page 85 note 1} For a considerable, but by no means exhaustive, bibliography, see pp. 94–95. Unless it is obvious that the reference is to Griechische Verskunst, ‘Wilamowitz’ always means Isyllos Von Epidauros.

^{page 85 note 2} The only full-length discussion is that by Wilamowitz in Isyllos. Thomson has scanned a part of the fragment and pointed out its resemblance to Alcm. fr. 60. Wilamowitz in Gr. Versk. and Schroeder deal with the corresponding words οἷον … δο⋯ρατι and Zε⋯ … μο.

^{page 86 note 1} Bergk's later καταλεισ and Headlam's, W.τ' ⋯δε⋯ς (C.R. xiv (1900), p. 7)Google Scholar will not scan. It is possible that P's punctuation is right and that ταδ' εἰσ should go with αυλ⋯αν, but no emendation on these lines has ever been offered.

^{page 87 note 1} Nietzsche, following Hesychius's definition κγώσσειν ῥῥγχειν, explains κγώσσειν as ‘a kind of diminutive of “to snore,”’ connecting it with κν⋯ω, κν⋯ος, κν⋯ω, κν⋯μα, κνυζ⋯ω.

^{page 88 note 1} Ahrens's emendation κ⋯δ δ⋯ is usually rejected on the ground that κ⋯δ IS superfluous; but an exact parallel occurs in Alc. fr. 86. 3. Spengel's ⋯ δ⋯, usually accepted, is open to the serious objection that none of the other nouns in Sappho's list has the article.

^{page 88 note 2} It is shown on p. 91 that this hiatus is admissible, but it is not so common that we are entitled to create additional examples by emendation.

^{page 88 note 3} See ‘On the Road to Mandalay,’ v. 21: ‘With 'er arm upon my shoulder, and 'er cheek again my cheek.’ Barrack Room Ballads was first published in 1892.

^{page 89 note 1} The quotation from Pindar shows that the article is not indispensable.

^{page 89 note 2} The earliest reference for the absolute use of μεταβολ⋯ in this sense given by the new L. and S. is to (Philolaus), i.e. v/iv cent. The form μεταβολια (in the sense ‘barter’) is quoted from LXX.

^{page 89 note 3} This translation would more naturally require M's reading β⋯λε in the text; it will be shown that the aorist is also preferable on metrical grounds.

^{page 90 note 1} Radermacher bases assertions as to the metrical structure of the θρ⋯νος on the theory (first put forward by Schneidewin) that ‘Danae’ is part of a θρ⋯νος; but the theory is entirely without concrete foundation. The story of Danae is certainly suitable to a consolatory poem; but it is hard to understand how anyone can ever have thought that it could be addressed to a mother that had lost her child (Smyth, p. 321). The whole point of the story is that Danae, though deserted by everyone else, still had her son to console and later to avenge her. I ome the reference to Radermacher to Professor T. B. L. Webster.

^{page 90 note 2} I do not mean to assert that the tradition is not as faulty as Edmonds's text implies; all I maintain is, that in default of any proof that Edmonds is right the most dependable results are likely to be obtained by giving full weight to the available evidence.

^{page 91 note 1} References to Pindar are given throughout by the numeration of Schroeder's ed. min. of 1930; those to Bacchylides by that of Snell's edition (1934); those to other lyric poets by that of Diehl's, Anth. Lyr. (1925)Google Scholar (all published by Teubner).

^{page 91 note 2} 'Iσθμ⋯ς is perhaps not quite a fair example, as both Pindar and Bacchylides fluctuate in their treatment of it. To the instances given above add Isth. i. 9 and perhaps v. 5 (but see Farnell's note ad loc.); and contrast Ol. viii. 48 ε‘π’ 'Iσθμο⋯ and B. viii. 18 νεμ⋯αν τε καῐ 'Iσθμ⋯ν. Pindar never admits hiatus or lengthening before 'I⋯ρων (the relevant instances are Ol. i. 11, 107, Pyth. i. 56, ii. 5). For “εβρος, I can discover no relevant instance except Ar. Birds 774 παῤ “εβρῳ.

^{page 91 note 3} For a full explanation of this point see my ‘Double Scansion in Early Greek Lyric’ (C.Q. xxviii (1934), pp. 183 ff.)Google Scholar. Briefly, the theory is that the Pherecratean can be scanned either dactylically or as a Catalectic Dimeter

^{page 91 note 4} εἷπεν τ' ὦ τ⋯κος will not scan alone; but if the τ was not in the original text, it is hard to see why it should have crept in. The retention of the digamma is justified by the practice of simonides (cf. fr. 30. 3 ⋯πτ⋯ (F) ιοπλοκ⋯μων) and Bacchylides (examples in Snell, p.17*).

^{page 91 note 5} I use ‘Trochee’ and ‘Iamb’ instead of the clumsy ‘Ditrochaeus,’ ‘Diiambus’ or ‘Trochaic (Imabic) Dipody.’ The shorter froms have the advantage of corresponding with Greek metrical usage, which allows —∪ and ∪— no independent existence.

^{page 91 note 6} I have put forward reasons for believing that this papyrus should be ascribed to Simonides in a paper published in C. R. xlviii (1934), pp. 205–207.Google ScholarSnell's alteration of the accent on ιοδερκεῖ is quite unnecessary.

^{page 92 note 1} See ‘Double Scansion…“ p. 188, n. 1.

^{page 92 note 2} –⋯ν κον⋯ν βαθεῖαν and εὑδ⋯τω δ⋯ π⋯ντος also look like Itbyphallics, but are better explained otherwise (v. infra).

^{page 92 note 3} The line in question scans

∪ –∪– | ––∪∪|–∪–, i.e. Alcaic Hendecasyllable, the second metron being Ionic a majore, of which the first syllable is ‘anceps’ (see the examples given in ‘Double Scansion…’ pp. 187, 189).

^{page 92 note 4} Simonides and Bacchylides appear to admit synizesis only in words which are contracted in Attic. The best evidence for Simonides's general practice is in frr. 4, 5. Synizesis of κυαν⋯ος is possible in frr. 27, 46, and of πορφυρ⋯ος in fr. 28, 44. In Bacchylides, κναν⋯ος only occurs in the passage quoted above, but πορφυρ⋯ος occurs three times, always with synizesis (the references are xvii (xvi). 17, xviii (xvii). 52 and fr. 21. 2). Thus it is reasonable to assume synizesis in κυαν⋯ος and πορφυρ⋯ος in this fragment and to exclude it in δαιδαλ⋯ος and θαρσαλ⋯ος.

^{page 92 note 5} This reluctance seems to be due to the fact that one of the commonest forms of the Glyconic is (i.e. ‘light’ Trochee, Iamb). ButAesch, cf.. Agam. 412Google Scholar, which should apparently be scanned

^{page 93 note 1} For the synaloephe εὑδ⋯τ ⋯– cf. Bacchyl. iii 22 ⋯γλαιζ⋯τ.

^{page 93 note 2} The scansion –σαλ⋯ον (F ἓπος (∪ – ∪ –) is excluded by the argument of n. 4, p. 92. There is no need to assume that simonides was more inflexible in the use of the digamma than Bacchylides.