^{page 66 note 1} As, for example, when Kroll chooses emungêrê sôlēbât against emungi sôlēbât of other manuscripts of ad Her. 4. 67 on rhythmical grounds (Philologus, lxxix [1934], 74–75), obviously basing himself on Ciceronian practice since he admits (p. 80) that no adequate study of clausulae in the ad Her. was available to him. We shall in fact see that the Auctor did not favour , though it occurs quite frequently in the exempla as a whole, so that the choice of reading is an open question. Even if Cicero were the writer, we should have to remember that though he notoriously liked he uses far more frequently (and this not only because it naturally occurs more frequently in Latin prose, but because he was partial to it, to an extent which can be determined by methods discussed below). In any case he did not greatly care for the typology of with the word-break after the dactyl.

Similar objections may be made against Rose's, H. J. proposal (Handb. Lat. Lit. [London], 2nd ed., p. 106, n. 84)Google Scholar to emend Crassus', L.possumus ēt dēbēmûs to debēmûs ēt pôssûm^us. This imposes Ciceronian rhythm (I think a peculiarly Ciceronian rhythm originally) on an earlier author, and again, although it is true that Cicero normally avoided the double spondee except for special effect, he might well have used it in such a passage as this.Google Scholar

^{page 66 note 2} For some results see Shewring's, W. H. article ‘Prose-Rhythm' in O.C.D.Google Scholar

^{page 66 note 3} It is perhaps the fact that scholars of an earlier generation offered conflicting and sometimes extravagant solutions to possibly unanswerable questions about the general nature of prose-rhythm that has led some to regard the whole subject with suspicion. But I hope by the conclusions of this paper to show that there is a mean between the pointless amassing of statistics and wild theorizing. Only one of the main theories of prose-rhythm is totally incompatible with my approach, that which held that repetition of clausulae was the essential element (Blass, May, Zander: see Shewring's, bibliography, loc. cit.).Google Scholar The ‘French' school which stresses the word as unit (v.i.) and even those who see in accent the determinant of prose-rhythm (e.g. Broadhead) would probably reach the same results, though they would express them differently.

^{page 67 note 1} De numero oratorio Latino (Groningen, 1919)Google Scholar and Der antike Prosarhythiius (Groningen, 1921) are the most relevant of his works for my present purpose.Google Scholar

^{page 67 note 2} C.Q. xxiv (1930), 164–74,Google Scholar xxv (1931), 12–22, and an answer to criticisms by Broadhead, H. D. (C.Q. xxvi [1932], 35–44)CrossRefGoogle Scholar in C.Q. xxvii (1933), 46–50.Google Scholar

^{page 67 note 3} One of my own results seems to be abnormal, 7∣0 per cent, for and I have borrowed De Groot's 5∣4 per cent.

^{page 67 note 4} C.R. lii (1938), 148.Google Scholar

^{page 67 note 5} La Rhétorique à Hérennius, etc., in Méanges Boissier (Paris, 1903).Google Scholar The substance reappears in his Les Clausules métriques latines (Lille, 1907), pp. 542–6.Google Scholar

^{page 67 note 6} (i) Its emphasis on the word as the unit of investigation conflicts with ancient authority; (ii) by treating the last word of the sentence as ‘given’ and asking ‘What metrical group precedes final words of various metrical forms?’ it obscures the part played in Latin clausulae by the flexibility of Latin word-order; (iii) for a fault affecting the ad Herennium in particular cf. p. 69, n. 1. Seme modern French scholars, e.g. Cousin, reject the methods of their predecessors.

^{page 68 note 1} Sprachliche Beobachtmgen zum Auctor ad Herennium (Breslau, 1935), p. 16.Google Scholar

^{page 68 note 2} Because the Auctor claims originality in his exempla, Bornecque half apologizes for dividing the book up in this way with a view to settling questions of authorship. But even if the exempla were original, they might show interesting rhythmical peculiarities, and in fact the Auctor's claim is demonstrably false, and the dissection of his work justified by results.

^{page 68 note 3} No trace will be found in my list of Zieliński's procrustean ‘cretic-base’. It has been justly dealt with by earlier writers, who have also called attention to what is possibly its most flagrant defect of detail, its exclusion of the double trochee, which, if ancient testimony is worth anything at all, was certainly a clausula in its own right, whether pceded by a cretic or not.

^{page 68 note 4} Detailed comparison of such passages as ad Her. 1. 11, Inv. 1. 26, Quintil. 4. 1. 71 eveals a clear probability that the common ground is due to indebtedness to a common radition recorded in non-rhythmical Latin ather than to direct borrowing.

The elaboration of the prologues is not in itself remarkable. Ancient legal and mathematical treatises show the same phenomenon (cf. Wilamowitz, in Hermes, xxxv [1900], 33Google Scholar and Schulz, , Roman Legal Science [Oxford, 1946], p. 187, n. 3).Google Scholar

^{page 69 note 1} I am sure that the French method is wrong which by its rules of procedure excludes all words of five or more syllables from consideration. This distorts the Auctor's practice, and even in Cicero, according to Shewring (C.Q. xxiv. 167)Google Scholar, the type archipiratas is commoner relatively (i.e. in comparison with non-rhythmical prose) than esse constabat. Quintilian (9. 4. 66) warns against polysyllabic endings as etiam in carminibus … permolle, but in censuring the ending archipiratae clearly recognizes it as a clausula.

^{page 70 note 1} The Auctor achieved favoured rhythms by this means in 29 of Golla's 65 instances, but he liked it on its own account, for 27 instances are not rhythmically affected, 16 of the 27 giving favoured rhythms with either word-order. In 4 only does the hyperbaton produce a less-favoured rhythm. In 5 it improves the clausula from ‘disliked’ to ‘indifferent’. On the 9 further instances in the exempla see below.

^{page 70 note 2} Henceforward ‘rhythmical’ means ‘showing the Auctor's liking for and ’. It does not include or a cretic in the last place. ‘Non-rhythmical’ implies the absence of the favoured clausulae, and the presence of , or , or of clausulae so rare that even if there is some evidence that the Auctor liked or did not object to a particular ending of this kind, e.g. , it seems better to treat them as non-rhythmical.

^{page 71 note 1} References are by page and line of Marx's smaller (1923) edition, in which also will be found the parallel passages in Greek and Roman authors for the borrowed exempla.

^{page 72 note 1} Philologus, lxxxix (1934), 215.Google Scholar

^{page 72 note 1} Roman Essays and Interpretations (Oxford, 1920), pp. 91–99.Google Scholar

^{page 72 note 3} The other occurrences are 52. 3, 63. 16, 133. 6, 135. 26, 160. 10, 169. 4, 171. 17 (the first, fourth, and fifth not before main stops). The rhythmical effects are: in one no improvement of a ‘bad’ clausula, in five a less-favoured rhythm is replaced by the double trochee, in three a double trochee results either way.

^{page 73 note 1} Thus we cannot plead as Caplan does (Loeb, ed., p. xxxi) that his boasts of originality refer to producing his own translations from Greek.Google Scholar

^{page 73 note 2} In the Auctor's main work this clausula is but little sought, though this typology still occurs in a high, though smaller, proportion (16 out of 50). Its proportion to other typologies diminishes as the work proceeds: Book 1, 5 in 9; Book 2, 5 in 14; Book 3, 5 in 18; Book 4, 1 in 9. By nature the typology is extremely rare, since words of the form are few.

^{page 74 note 1} The only exemplum from Greek which is also paralleled in Latin is 150. 18, where the Auctor is more rhythmical than Quintilian— but also nearer the Greek.

^{page 74 note 2} The resemblance to Aeschines in Ctes. 133 observed by Marx is slight, as Hubbell comments in reviewing Caplan's Loeb edition of the ad Herennium (A.J.P. lxxvii [1956], 214).Google Scholar

^{page 75 note 1} It is usually assumed that both exempla are more or less accurate accounts of real events, though M¨nzer, (op. cit., p. 217) reports that Kroll had communicated doubts to him.Google Scholar

^{page 77 note 1} Well set out in Bonner's account in Roman Declamation (Liverpool, 1949),Google Scholar though the early dating of the ad Herennium is of course assumed. The root sense of the group of words derived from declamare is ‘loud utterance’, The first technical sense is to ‘rehearse’ a speech (for a real case) (Rose. Am. 82). The association with voice-training survives at least till 55–54 (De Oral. 1. 251). Declamator and declamatorius are related to practice-speeches from 55–54 (Planc. 83, De Orat. 1. 73, Q.F. 3. 3. 4), but declamatio still means ‘shouting’ (Mur. 44, Planc. 47) until 44 B.C. (Fam. 16. 21. 6), and declamare first appears in the full technical sense in 45–44 (Fin. 5. 5, Att. 14. 12. 2).

^{page 77 note 2} On these sections Marx observes (86. 24 with app. crit.): hinc usque ad finem libri III stilus peculiaris, a statement for which support can be found in the clausulae, with a figure of 49 per cent, for So this may genuinely be a ‘late’ passage.

^{page 77 note 3} e.g. in Gwynn, A., Roman Education (Oxford, 1926), pp. 68 and 162.Google Scholar His view about the political significance of the closing of the schools of the rhetores Latini by the censors of 92 has been questioned by Frank, Tenney, Life and Literature in the Roman Republic (Cambridge, 1930), p. 150, n. 28,Google ScholarClarke, M. L., Rhetoric at Rome (London, 1952). pp. 12–13.Google Scholar

^{page 77 note 4} The sceptical are invited to apply the criterion of the latest datable event to Cicero's Topica and Part. Orat. and to the two sets of declamations once attributed to Quintilian, and to consider, among much else, Quintilian's list of turbulenti in Inst. Or. 2. 16. 5. Suetonius (Rhet. 1) refers to de clamatory themes drawn ex veritate ac re si qua forte recens accidisset, but he is dealing with controversiae, which were on legal themes. So in Cicero's De Inventione of c. 86 B.C. a legal case of the late 90's is cited, but the latest important political event referred to belongs to 106 B.C.

^{page 77 note 5} For good arguments against Cornificius' authorship cf. Kohler, , De Rhetoricis ad C. Herennium (Berlin, 1909)Google Scholar and the admirable summary by Caplan, , op. cit., pp. ix-xiv.Google Scholar

^{page 78 note 1} The best defence of the Auctor's failure to keep his promise is that the proem of Book 4 deals with borrowings from creative writers, i.e. orators, poets, and historians. Borrowings from such sources can nowhere be proved of him in Book 4 apart from Greek oratory, and the exempla from these reached him through intermediate sources. But he makes unacknowledged borrowings from Latin rhetoricians, and in implying that the choice lies between taking exempla from orators and poets and inventing them he is clearly guilty at the least of falsa enumeratio.

^{page 78 note 2} This article was completed before the appearance of Schmid's, W.Über die klassische Theorie und Praxis des antiken Prosarhythmtis (Hermes Einzelschriften Heft 12 [1959]), in which the author, adhering closely to ancient sources, argues inter alia (i) that the word clausula strictly means only ‘end of a sentence’, not a rhythmical form of sentence-ending, (ii) that prose-rhythm runs throughout a sentence, and (iii) that the end-rhythms are in no way so distinctive as to merit the customary concentration of investigation on the (so-called) ‘clausulae’. While accepting that (i) and (ii) are undoubtedly true of ancient theory, I do not feel it necessary to abandon the established modern usage of clausula; and though I agree with Schmid that ‘clausulae’ are not the only element in prose-rhythm, I remain convinced that, as indeed much ancient evidence suggests, the rhythms of the sentence-endings were and are the most striking in effect, and the most susceptible of analysis; and that as ancient authors show undoubted idiosyncrasies in their end-rhythms, it is not improper to make special studies of these, at least for such purposes as the present one, as an ‘empirical and objective test’ (cf. p. 66, n. 3 above). For further comment on Schmid's novel and interesting ideas cf. my forthcoming notice in C.R.Google Scholar