^{page 155 note 1} This clause introduces a reference to the structure of the verse, as does the more explicit clause in Labarbe's definition, no. (iii). It seems to me an adumbration of Parry's theory of improvisation, the organiza tion of formulae into ‘types’ capable of equivalent substitution. As such it ought to be in the conclusion of the argument, not in the definition of the terms used in the pre misses. It allows us to say, for example, that and are the ‘same formula’ but would prevent us saying the same of and . I am not satisfied the poet saw such declen sions in entirely different lights.

^{page 155 note 2} Parry, M., L'épithète traditionelle dans Homère (Paris, 1928), p. 16.Google Scholar

^{page 155 note 3} Bowra, C. M., Heroic Poetry (London, 1952), p. 222.Google Scholar

^{page 155 note 4} Labarbe, J., L'omère de Platon (Liège, 1949), P. 19.Google Scholar Labarbe's statement may be taken as that of Severyns also, on whom he greatly depends. Severyns included no formal definition in his model description of Homeric formulaic usages in Homère ii, Le poète et son æuvre (Brussels, 1949), pp. 49–61.Google Scholar

^{page 155 note 5} Parry, M., ‘Studies in the Epic Tech nique of Oral Verse-making: I. Homer and the Homeric Style’, HSCPh xli (HSCPh xli), 73–147. The words quoted are taken from p. 117.Google Scholar

^{page 155 note 6} HSCPh xli (HSCPh), 77–78, cf. 137–8. Parry thus rejects the suggestion of P. Chantraine (RPh iii [1929], 299) that formulae are more typical of the nominal than the verbal diction.Google Scholar

^{page 156 note 1} See the Tables, HSCPh xli (1930), 118–21Google Scholar, repeated with slight variations by Lord, A. B., Singer of Tales (Cambridge [Mass.], 1960), p. 143.Google Scholar

^{page 156 note 2} HSCPh xli (1930), 117.Google Scholar

^{page 156 note 3} Notopoulos, J. A., AJPh lxxxiii (1962), 360.Google Scholar

^{page 156 note 4} HSCPh xli (1930), 128.Google Scholar

^{page 156 note 5} HSCPh xli (1930), 145; cf.Google ScholarLord, A. B., TAPA lxxxiv (1952), 127,Google Scholar and Singer of Tales, pp. 31 ff.Google Scholar

^{page 156 note 6} HSCPh xli (1930), 133.Google Scholar

^{page 156 note 7} In this he is followed by Lord, , Singer of Tales, p. 143, despite Lord's evident interest in phrase structure.Google Scholar

^{page 156 note 8} Notopoulos, J. A., ‘The Homeric Hymns as Oral Poetry’, AJPh lxxxiii (1962), 337–68:Google Scholar chart of Hymn to Apollo 1–18, pp. 356–7.

^{page 156 note 9} This point would be more impressive if confinement to a given position were as much a feature of formulae as it is often supposed to be. In fact the mobility of formulae is not materially different from that of words of the same length. In general the shorter a formula is the more likely it is to be found in two or more places: 64 per cent, of formulae scanned , a neglected group, are mobile. In the placing of words of a given shape in a given position the Homeric practice does not differ much from that of hexameter poets of all periods (O'Neill, E. G., ‘The Localisation of metrical word-types in the Greek hexameter’, YCIS viii [1942], 115).Google Scholar Yet a very frequent repetition at a given point deserves notice, cf. Kirk, G. S., Songs of Homer (Cambridge, 1962), p. 67.Google Scholar I am less happy with C.J. Ruijgh's application of this to certain connectives, L'élément achaéen dans la langue épique (Assen, 1957), pp. 55 ff.Google Scholar, on which see Kirk, G. S., Mus. Helv. xvii (1960), 199.Google Scholar

^{page 157 note 1} e.g. by the critics of the German neo- analytical school, Mette, H. T., Lustrum i (1956), 14;Google ScholarSchoeck, G., Mas und Aithiopis (Zürich, 1961), p. 13.Google Scholar

^{page 157 note 2} Cf. Basset, S. E., CF xxv (1930), 642Google Scholar and Poetry of Homer (Berkeley, 1938), pp. 14 ff.;Google ScholarShorey, P., CPh xxiii (1928), 305.Google Scholar

^{page 157 note 3} HSCPh xli (1930), 84–89 and xliii (1932), 7–8.Google Scholar

^{page 157 note 4} Cicero's description, Brutus 221.

^{page 158 note 1} Summarized in his epigram ‘die Sprache der Homerischen Gedichte [ist] ein Gebilde des epischen Verses’, R.-E. viii col. 2214.

^{page 158 note 2} HSCPh xli (1930), 127.Google Scholar

^{page 158 note 3} Cf. Rank, L. P., Etymologiseering en Ver- wante Verschijnselen bij Homerus (Assen, 1951), passim.Google Scholar

^{page 158 note 4} 31 times: a metrical equivalent occurring once only is . Parry, , Épithète traditionelle, p. 64,Google Scholar claims this as an expression with the equivalent of a double initial consonant, but the occurrence at Il. 23. 168 does not require it.

^{page 160 note 1} The unique Il. 19. 404 is from an altogether exceptional context.

^{page 160 note 2} AJPh lxxxiii (1962), 356 n. 59.Google Scholar

^{page 160 note 3} It is far from clear, however, which epithets are generic and which are the fixed epithets of particular place names, cf. Page, D. L., History and the Homeric Iliad (Berkeley, 1959), pp. 123–4. We can only speak of an order of preference if there is an adequate system of generic epithets on call.Google Scholar

^{page 161 note 1} Virtually every possible shape has its formulae. This should restrain us from laying too much stress upon shape. When the poet can make and use an expression of any shape, the fact that he does choose an expression of a given frequent shape does not prove that he is working within the confines of a methodically-organized system of interlocking shapes.

^{page 161 note 2} The use of the ordinal terms is not intended to imply any order of priorities.

^{page 162 note 1} It is generally assumed, but nowhere demonstrated, that proper names are typical of the diction as a whole. This seems to me true only of the oblique case forms: the nominative proper name has a very stereotyped usage. I have made no distinction between functional and ornamental epithets ; the problems of getting a phrase into the verse are indifferent to the force of the epithet.

^{page 162 note 2} Here are placed expressions which occur again once only in a different shape or case, the principle being that where derivatives exist the source must be established. I ignore at this point non-stylistic criteria and the possibility that the direction of adaptation varies from one example to another.

^{page 163 note 1} This expression actually appears in the text twice (Il. 9. 128 and 270), but the second instance is in the verbatim repetition of a speech. This is not formulaic repetition.

^{page 163 note 2} These may appear to be necessarily functional, yet the usage is often as formulaic as the most ornamental epithets, and they are manipulated in just the same way, cf. 8 times, and 22 times, 5 times, etc.: on numerals see Waltz, P., Rev. Ét. Hom, iii (1933). 1–38.Google Scholar

^{page 164 note 1} Cf. Lord's statement (of the Yugoslav tradition), ‘Each singer has a group of formulas which forms the basis of his style. These change but seldom; on them he patterns others’, TAPA Ixxxiv (1953), 127 (cf. Singir of Tales, p. 37).Google Scholar

^{page 164 note 2} I have sought to indicate something of the nature and extent of formular flexibility in a short paper ‘The Homeric formula and the problem of its transmission’, BICS ix (1962), 57–68.Google Scholar