^{page 174 note 2} For an enumeration of ancient remarks on sigmatism, see Wilkinson, L. P., Golden Latin Artistry (Cambridge, 1966), 13Google Scholar, and Stanford, W. B., The Sound of Greek (Berkeley, California, 1967), 8, 53–5.Google Scholar

^{page 174 note 3} Apart from Tibullus, Propertius, Ovid, and Martial the samples are hexameters.

^{page 174 note 4} My statistics were originally based on the text of Postgate, J. P., Corpus Poetarum Latinorum (London, 1905)Google Scholar, to ensure uniformity in trivialities of orthography. (E. Hiller contributed the text of Tibullus to the Corpus). The accuracy of my figures has been checked by sample comparisons with the most reliable modern texts, Lenz, F. W., Albii Tibulli Aliorumque Carminum Libri Tres (Leiden, 1959)Google Scholar and Barber, E. A., Propertius (Oxford, 1960)Google Scholar. My colleague, Dr. P. G. O'Regan, kindly analysed for me by computer the first two books of Propertius. His results gratifyingly agree with mine.

^{page 174 note 5} Arbitrarily defined as lines having 5 s's or more. Lines of 8 s's and 9 s's are not included: these occur so infrequently as to be useless for purposes of comparison.

^{page 175 note 1} It is a curious fact that Scott, J. A. (AFPh xxxix [1908], 69–77)Google Scholar and Todd, O. J. (CQ xxxvi [1941], 29–39)Google Scholar arrived at different conclusions when, in discussing sigmatism in the Greek dramatists, the former based his argument on lines of marked

Scott (unjustifiably, in my view) concludes that ‘the sigmatism of all four of these poets is essentially the same’ (he does not take into account the variations in the lengths of sigmatism, the latter on the average number of sigmas per 100 lines. (Todd used samples of trimeters, chosen at random in blocks of Too, totalling 600 each for Aeschylus and Sophocles, and a little over 700 for Euripides). Their statistics are worth comparing:

plays). Todd on the other hand, is led by his figures to state that ‘the comic poets... definitely reduced the amount of hissing sounds by some 40 to 50 per 100 lines’.

^{page 176 note 1} Perhaps this arises from a tendency in Tibullus to place the perfect infinitive active in the pentameter.

^{page 176 note 2} Cicero, Or. 149, in discussing the proper collocation of words in accordance with euphonic principles, urges that special care be devoted to the relations between the last syllable of one word and the first syllable of the next, so as to avoid the clash of harshly sounding consonants. Quintilian, Inst. 9. 4. 37, condemns the hissing sound produced by the combination ars studiorum. He tells us that Servius, being influenced by Lucilius, dropped final s whenever the next word began with a consonant and that Luranius disapproved of this practice while Messalla defended it. (Servius and Luranius cannot be identified.) It is noticeable that in the sole instance of elision of final s before a consonant in Catullus (tu dabi' supplicium, 116. 8) the elided s precedes another s.

^{page 176 note 3} For evidence as to date of composition, see H. E. Butler and Barber, E. A., The Elegies of Propertius, pp. xxvii–viii.Google Scholar

^{page 177 note 1} He gives statistics only for the combination s + n.

^{page 177 note 2} For our present purpose a level of at least go per cent. has been deemed necessary. For a discussion of the statistical method employed, see Hod, P. G., Introduction to Mathematical Statistics (1954), 109.Google Scholar

^{page 178 note 1} The percentages per book are: Tib. I 6·3, Tib. II: 4·7, Prop. I: 2·4, Prop. II: 2·3 Prop. III: 2·1, Prop. IV: i 1·9.

^{page 178 note 2} See Hod, loc. cit.

^{page 179 note 1} Quintilian, Inst. 1. 7. 35, records that Messalla, patron of Tibullus, wrote a book on the letter s. Unfortunately we cannot know what the attitude of Messalla was but the reference in Quintilian suggests there was an interest in the subject among the literary group of which he was the head.

Tibullus and Propertius differed in other aspects of poetic technique. The following are a few:

(a) Tibullus never permits hexameter to end with a double spondee. Propertius occasionally does. Cf. Platnauer, M., Latin Elegiac Verse (Cambridge, 1951), 39.Google Scholar

(b) Tibullus shows a tendency to use a weak caesura in the 3rd foot of hexameter, the percentages being: Tibullus, 20; Propertius, 5·4. Cf. Platnauer, op. cit. 9.

(c) Tibullus has 3 per cent trisyllabic endings in pentameter, Propertius has 1·5. Cf. Platnauer, op. cit. 15.

^{page 179 note 2} Ancient poetry was intended primarily to be read aloud. The importance of this for the appreciation of sound effects needs to be borne in mind.

^{page 179 note 3} Cf. Cic. Or. 150 and Auct. ad Her. 4. 32.

^{page 179 note 4} For the notion of the poet's omnipresence in his work, see Eliot, T. S., The Use of Poetry and the Use of Criticism (London, 1933), 118–19Google Scholar: ‘What I call the “auditory imagination” is the feeling for syllable and rhythm, penetrating far below the conscious levels of thought and feeling, invigorating every word; sinking to the most primitive and forgotten, returning to the origin and bringing something back, seeking the beginning and the end.’ I am indebted for this reference to my colleague, Professor S. Lucy.

^{page 180 note 1} ‘Tibullus’ IV denotes the poems of (a) Amicus Sulpiciae; (b) Sulpicia; (c) Tibullus or Pseudo-Tibullus. The Panegyricus Messallae is not included.