* Note on citations: Hermogenes is cited by page and line number according to the Teubner edition of H. Rabe (1913); the Opuscula of Dionysius of Halicarnassus by volume, page and line number in the edition of H. Usener and L. Radermacher (Leipzig, 1899–1929; volumes v and vi in the complete Teubner Dionysius). ‘Sp.’ denotes a reference to L. Spengel, Rhetores Graeci (1853–6; vol. i partem ii iterum ed. C. Hammer, 1894); ‘Walz’ to C. Walz, Rhetores Graeci (1832–6; for authors not in Spengel); ‘Halm’ to C. Halm, Rhetores Latini Minores (1863).

¹ In practice rhythms and cohesive syntax both characterize the period (cf. below p. 92); the question here is what defines the period in theory.

² This definition is amplified below, pp. 90 f.

³ For the rhythmical definition: especially Zehetmeier, J., ‘Die Periodenlehre des Aristoteles’, Philol. 85 (1930), 192–208,Google Scholar 255–84, 414–36 (hereafter referred to by author's name); also Schmid, W., Über die klassische Theorie und Praxis des antiken Prosarhythmus (Hermes Einzel-schriften 12, 1959), 112 ff.;Google Scholar Schenkeveld, D. M., Studies in Demetrius on Style (1964), pp. 28 ff.;Google Scholar cf. Solmsen, F., AJP 62 (1941), 45 n. 36. For the ‘logical’ definition:Google Scholar Kaibel, G., Stil und Text der Поλιτεία Ἀθηναίων des Aristoteles (1893), pp. 64 ff.;Google Scholar Groot, A. W. de, Der Antike Prosarhythmus (1921);Google Scholar Wilkinson, L. P., Golden Latin Artistry (1963), pp. 167 ff.Google Scholar Primmer, A., Cicero Numerosus: Studien zum Antiken Prosarhythmus (1968), pp. 45 ff., thinks that neither logic nor metre defines a period, but the use of sound-figures, which he says constitutes a type of ⋯ριθμός.Google Scholar

⁴ Russell, translates (Ancient Literary Criticism, edd. Russell, D. A. and Winterbottom, M. (Oxford, 1972), p. 148): ‘a beginning and end determined by its own structure’.Google Scholar

⁵ This usually involves syntactic subordination, but not always, since a period can have only one colon (1409 b 17, on which see further below nn. 18, 22, 40).

⁶ The terms περ⋯οδος and κ***λον themselves come from metrical terminology; see Dion. Hal. Comp. 2. 84 ff. and O. Schroeder, Nomenclator Metricus (Bibliothek der klassischen Altertumswissenschaft, hrsg. v. J. Geffcken, Bd. v, 1929), p. 36; E. Norden, Die Antike Kunstprosa ³ 1. 42 n. 2; Zehetmeier, p. 274 n. 53.

⁷ Zehetmeier, pp. 257, 261; Schmid, op. cit. (n. 3), pp. 116, 118 f.

⁸ Zehetmeier, p. 263; cf. Primmer, op. cit. (n. 3), p. 47.

¹⁰ cf. Kaibel, op. cit. (n. 3), p. 89, who is right to maintain that the remarks at 1409a8 ff. do not mean that a paean should be placed at the beginning and end of every sentence.

¹¹ To this Zehetmeier replies (p. 272) that by the time Aristotle gets around to quotations, he is concerned to illustrate antithesis, and all of his examples do so at the expense of the earlier point. Moreover, they are taken from the Panegyricus of Isocrates, who is of course very fond of antithesis, but not so much of the paeanic rhythm. The practice of the orators on whom Aristotle must draw for examples is sometimes at variance with the precepts of his rhetorical theory.

¹² Hermogenes, Inv. 176. 23:. Cf. also Demetr. cc. 30 f., and the Auct. ad Her. 4. 19. 27: ‘Continuatio [= period] est et densa <et continens> [1] frequentatio verborum [2] cum absolutione sententiarum.’

An alternative explanation of this sentence might be to say that Aristotle conceives of discourse as a continuum which would be incomprehensible if it were not punctuated (or ‘limited’) at various intervals (so Theophrastus, cf. below p. 96). Punctuation marks are indicated by metrical clausulae. So far is pure theorizing; but then it takes only a moment's thought to realize that these clausulae cannot simply be sprinkled about indiscriminately. Their actual position is ‘also’ determined by the sense. The central importance of the latter is obvious; Aristotle's roundabout way of expressing the point takes nothing away from it.

¹³ On this cf. R. G. Bury, The Philebus of Plato (1897), appendices C-E; De Groot, op. cit. (n. 3), p. 18.

¹⁴ See Solmsen, F., Hermes 66 (1931), 241–67.Google Scholar

¹⁵ Zehetmeier, pp. 428 f.

¹⁶ cf. Solmsen, loc. cit. p. 249 n. 2, p. 266 n. 0; Kappelmacher, A., Wiener Studien 24 (1902), 452 ff.Google Scholar

¹⁷ Zehetmeier, p. 428.

¹⁸ Some (e.g. Grube, G. M. A., A Greek Critic: Demetrius on Style (Phoenix supplementary volume 4, 1961), p. 35 n. 41Google Scholar (suppose that in the sentence κ***λον δω ⋯στ⋯ τ⋯ ἕτερον μ⋯ριον τα⋯της (1409b16), τα⋯της (sc. τ***ς περι⋯δου) refers to the period of ‘Sophocles’ just cited. If however ὥσπερ ⋯ εἰρημ⋯νη περ⋯οδος at b 15 (so A; ὥσπερ κα⋯ ⋯ περ⋯οδος, β, which is utterly senseless) is to be removed, this view becomes difficult. The phrase μ⋯ ⋯ν τ***ι διαιρ⋯σει κτλ. is an explanation of εὐαν⋯πνευστος; some later student, misunderstanding and relating διαιρ⋯σει in 1409b 15 to διαιρε***σθαι in b 11, inserted ὥσπερ κτλ. With the offending phrase removed, one understands τα⋯της to mean τ***ς ⋯ν κώλοις περι⋯δου.

Why then does Aristotle apparently restrict the number of cola to two? Perhaps because his is a highly theoretical analysis; he probably thought that two cola and no more most perfectly reflected the principles of balance and proportion. Also, he is shortly to pronounce his doctrine of the two types of periods in cola: διηιρημ⋯νη and ⋯ντικ‘03B5;ιμ⋯νη; of necessity these sorts of periods are built from two cola. Theory may be getting in the way of practice here, but one will hardly condemn Aristotle for it; the point is not central. Zehetmeier (p. 274) notes that in metrical theory of the time, periods had two cola; cf. O. Schroeder, Vorarbeiten zur griechischen Versgeschichte (1908), p. 12.

¹⁹ Zehetmeier, p. 424, cf. p. 430.

²⁰ ‘Parallelism’ is probably what Aristotle intends by the phrase διηιρημ⋯νη περ⋯οδος (1409b33), and such a description fits his one example. Other meanings have however been assigned (see e.g. Zehetmeier pp. 272 f.).

²¹ Zehetmeier, pp. 198 ff., 277 f.

²² There is some doubt as to what a monocolon period is (among the ancients as well as us, e.g. Aquila Romanus 28. 16 ff. Halm). The explanation given by Zehetmeier for the first sentence of Herodotus is the one accepted by most. In Demetrius’ other example, the phrase τα***ς τ***ν ⋯κου⋯ντων διανο⋯αις gives a snap at the end which may count as a καμπ⋯. See further J. D. Denniston, Greek Prose Style (1952), p. 7; Schmid, op. cit. (n. 3), pp. 123 f.; Wilkinson, op. cit. (n. 3), p. 171 n.; Schenkeveld, op. cit. (n. 3), p. 27. Generally, a single-colon period ought to meet the same requirement as a multicolon one: the parts should be logically arranged. Demetrius (c. 17) and Quintilian (9. 4. 124) also specify that it should have a respectable length.

²³ J. Vahlen, Sitz. Berlin (1902), 166–94 = Gesamm. Philol. Schr. 2. 619–51, shows that quotations in the Rhetoric which are given t o illustrate a point of style are always sufficient in themselves.

²⁴ The adaptation of the runner-illustration is not of the same order.

²⁵ Die Anfangsstadien der griechischen Kunstprosa in der Beurteilung Platons (Diss. Leipzig 1916), pp. 9 ff.Google Scholar

²⁶ Saupe is right of course to the extent that διηιρημ⋯νη has a different meaning in the two authors; Aristotle uses it for one type of the περ⋯οδος ⋯ν κώλοις How διηιρημ⋯νη and λελυμ⋯νη came to replace εἰρομ⋯νη is not clear, but they do perhaps express the contrast with the periodic style better. It is true that in Caecilius (apud Aq. Rom. 27. 12 Halm, cf. [Aristid.] 2. 507. 2ff. Sp.) the λ⋯ξις λελυμ⋯νη and the λ⋯ξις εἰρομ⋯νη are distinguished. This I take to be an innovation, of no relevance to the problem here or the related one discussed above, p. 93 (no. 4).

²⁷ ‘Aristotle on the Period’, Harv. Stud. 63 (1958), 283–8.Google Scholar

²⁸ Cope, E. M., Aristotle's Rhetoric with a commentary, ed. J. E. Sandys (1877); An introduction to Aristotle's Rhetoric (1867).Google Scholar

²⁹ In his commentary on Demetrius.

³⁰ G. M. A. Grube, op. cit. (n. 18), p. 65.

³¹ cf. 208 f. I should note that ‘common teaching’ refers strictly to the later Hellenistic schools (cf. De Inv. 1. 6); to be taught by Isocrates is something quite different (Oral. 191). Cicero's remarks here reflect the standard charge of philosophers that rhetoricians did not base their teaching on any sort of broader wisdom; rhetoric was no real art, therefore, but as Plato says, a mere ‘knack’. Cf. De Or. 1. 83 ff.; Kroll, W., RE Supp Bd VII (1940), 1080 ff.Google Scholar

³² It is perhaps in this illustration of Theophrastus’ that the recurring description of the λ⋯ξις εἰρομ⋯νη as fluctuans, vagans or the like finds its ultimate origin: e.g. Cic. Oral. 77,220,233; Auct. ad Her. 4. 11. 16.

³³ κ***λον is already in Aristotle, of course; κ⋯μμα was perhaps not in Theophrastus, if we may judge from its absence in this part of the De Oralore (cf. membra el pedes, 185; not membra et incisa). I find no evidence that the term was used by Stoic grammarians, a thought first encouraged by the (admittedly unlikely) suggestion that the Archedemus quoted by Demetrius (c. 34, see Roberts ad loc.) for his modification of Aristotle's definition of the colon is the Stoic Archedemus of Tarsus. A more likely source of this term is the Asiatic school of rhetoric: cf. Zehetmeier, pp. 434 f; Cicero, Oral. 212 ff. on speaking incisim; and κομματικ***ς λεγ⋯μενον in the Homeric scholia (e.g. schol. A in Il. 2. 112, Nicanor; further instances in Erbse's note ad loc.) If this is so, we have an example of the influence of schoolmen on philosophers, though only in a matter of detail and terminology (as often), so that the argument here is not affected.

³⁴ ‘…numerum…adhibere necesse est…, quod ait Aristoteles et Theophrastus, ne infinite feratur ut flumen oratio, quae non aut spiritu pronuntiantis (Rhet. 3. 9 1409b 15) aut interductu librari (ibid. 3. 8 1409 a 20), sed numero coacta debet insistere.’

³⁵ cf. also Hermog. Inv. 176. 23 ff., 180. 12–14.

³⁶ cf. in general Barczat, W., Defigurarum disciplina atque auctoribus (Diss. Göttingen, 1904).Google Scholar

³⁷ A. Primmer, op. cit. (n. 3), is a thorough attempt; reviewed by Drexler, H., Gnomon 41 (1969), 654–62 (esp. 658 f.). For examples of Cicero's varying use of ‘numerus’ see Orat. 170, 199, 204, 207 f., 228; Brut. 8. 34; cf. also Quintilian's remarks, 9. 4. 55–7. The casual reader is not much troubled by this; Cicero himself is not one to fuss over distinctions anyway, although this is not to say he is unaware of them (Orat. 219).Google Scholar

³⁸ Kroll, W., RE Supp. Bd vii (1940), 1089 f.Google Scholar

³⁹ Fragments, ed. E. Ofenloch (Teubner, 1907).

⁴⁰ See above nn. 18, 22, and the translation on p. 90. Primmer, A., Rh. Mus. 109 (1966), 73 ff., is the scholar who suggests emendation; considerable rewriting is necessary. Mr Russell points out to me the difficulty of supposing that λ⋯ξις in 1409 b 14 means ‘sentence’ (as it must), when it means ‘style’ almost everywhere else in this book; but it is protected by 1409a35 λ⋯γω δ⋯ περ⋯οδον λ⋯ξιν ἔχουσαν ⋯ρχ⋯ν κα⋯ τελευτ⋯ν κτλ.Google Scholar

⁴¹ Defined as abnormal word order, in which an essential idea is left suspended until the end of the sentence.

⁴²

⁴³ I wish to acknowledge the generous and acute criticism of earlier drafts by Dr D. C. Innes, Mr D. A. Russell and Professor Dr Rudolf Kassel; various suggestions of a non-technical nature from Professor J. C. Yardley; and the time afforded by a fellowship at the Calgary Institute for the Humanities, the University of Calgary.