page 9 note 1 The Neuter Plural in Vergil 53-4.

page 9 note 2 “A ‘poetic plural’ is a plural that is used instead of a singular normal to prose, that is, prose uninfluenced by poetry; or a plural that is used instead of a singular that is normal to poetry itself” (Gummere, NPV 5). Cf. Paul Maas, Studien zum poetischen Plural bei den Römern, in Archiv für lateinische Lexikographie und Grammatik 12.480 (1902). For modification and amplification of this definition, as given by Maas, see Wilhelm Schink, De Romanorum plurali poetico 17 (with further bibliography 7 and passim in footnotes).

page 9 note 3 The Additional Short Syllables in Ovid 6.

page 9 note 4 This was partly in consequence of the syncope that had eliminated many short vowels and brought together the consonants that had previously been separated by them (Roland G. Kent, The Sounds of Latin³ §123).

page 9 note 5 That these exigencies were real, and that the poets themselves were conscious of them, is proved by their own testimony, in several well-known passages. Lucilius cannot mention by name a servorum ...festus dies (perhaps the Sĭgillārĭa: see Warmington's note ad loc, in Remains of Old Latin 3.79) quern plane hexametro versu non dicere possis; and Horace, on his journey to Brundisium, is reduced to spending a night oppidulo quod versu dicere non est (Sat. 1.5.87: Porphyrion, to whose note on this line we owe the preservation of the fragment of Lucilius just quoted, tells us that the name of the town was Aequus Tūtĭcus). Toward the end of his Epistulae ex Ponto (4.12 and 14) Ovid apologizes to his friend Tütĭcānus for having so long deferred the dedication of an epistle to him, on the ground that it is impossible to bring his name into elegiac verse. After saying in so many words

he contrives nevertheless to introduce the recalcitrant cognomen twice in distorted form, by dint of suggesting divers violations to which it might have been subjected. Writing in the Sapphic strophe (which, apart from such an elision as he is evidently reluctant to admit, will not accommodate a word consisting of four long syllables), Prudentius finds it difficult to mention, renuente metro, the martyred Sāturnīnī (Peristephanon 4.161-72); and Sidonius Apollinaris, after him, struggles with the same name (9, epist. 16.77-84). Poets of a loftier style, like Lucretius and Vergil, who could not explicitly avow the mechanical difficulties under which they labored, were constrained to use a hybrid patronymic like Scīpĭădās to replace the unmanageable Latin Scīpĭō (Lucr. 3.1034 and, in the plural, Verg. Georg. 2.170, Aen. 6.843). Among words of prime importance, only Ītălĭa (with the corresponding adjective, in certain forms) and rĕlĭgĭo (with rĕlĭgĭosus), because of their equally unmanageable tribrachs, suffered modification in the opposite direction to Ītalia (e.g. Verg. Aen. 1.2; cf. Itala 7.643, but Ĭtalus 5.117) and relligio (e.g. Lucr. 1.101; relligiosa Verg. Aen. 2.365), in dactylic verse. But Italia, as is shown by the unweakened medial a (Kent, SL³ §122.11), had remained essentially a Greek form, and as such was subject to a licence already established by the Greeks (Sophocles, Antigone 1119; Callimachus, Hymn to Artemis 58), who were, in fact, very free in their treatment of vowel-quantities in the initial syllables of proper names especially (cf. Hom. Il. 24.629, with II. 2.816-7; Hom. Od. 24.211, with Theocr. 7.40; and, above all, Homer's juxtaposition Il. 5.31, which is cited by Lucilius, ed. Warmington, ROL 3.114, and by Martial, 9.11.15, as an example of this liberty). In relligio (with which relliquiae, and congeners of both words, must be classed) the l is long in dactylic verse, whether so spelled or not (Cyril Bailey, in his note on Lucretius 1.63, sustains an old inexactitude in regarding the e as long); if this lengthened l is not merely a metrical license, it may include an assimilated d from the original form of the prefix, red-, which would make relligio itself a genuinely Latin formation (Ernout and Meillet, Dictionnaire étymologique de la langue latine², s.v. re-, red-, allow this possibility, as Bailey himself, in his Prolegomena VI, §21, does for the word reliquus, among others; but see Walde-Hofmann, Lateinisches etymologisches Wörterbuch³, s.v. re-).

To the passages cited above from the Latin poets I have found only two purely Greek analogues, one in an epigram of Critias, preserved by Hephaestion (J. M. Edmonds, Elegy and Iambus 1.488), the other in a fragment of Archestratus, preserved by Athenaeus (7.284e). In the former, the poet, finding it impossible to fit the name into elegiac verse, has recourse to the novel expedient of substituting for the pentameter, in one couplet, an iambic trimeter:

In the latter, a fish is described (Gulick, in his note ad loc., tells us that it is the or sturgeon).—Martial can introduce the Greek name into his Phalaeceans only by lengthening the first syllable to Ei-, a liberty taken, as he says, by the Greeks, quibus est nihil negatum, but not permitted the Romans, qui Musas colimus severiores (9.11); his elegiac epigrams on the same subject (9.12, 13, 16, 17, 36), like the longer poem of Statius in hexameters (Silv. 3.4), either avoid the name entirely, or allude to it by ringing changes upon its significance.

It is perhaps indicative of the difference that we have noted between the two languages that, in the case of proper nouns at least, the difficulties encountered in dealing with the Latin in every instance arose (or, where the name in question is not given, are thought to have arisen), from an excess of long syllables, with the Greek, from an excess of short. Of common nouns generally, and of other parts of speech, we hear nothing, because synonyms or at least periphrases could always be found for them; but there were many that could be used in dactylic verse either not at all, or only in certain forms and under certain conditions. On this whole problem, with special reference to Catullus and Ovid, see Ernst Bednara, De sermone dactylicorum Latinorum quaestiones, in Archiv für lateinische Lexikographie und Grammatik 14.317-60 and 532-604 (1906). Cf. also ch. V, n. 102.

page 10 note 6 Variation in the Latin Dactylic Hexameter, in Philological Quarterly 5.212-25 (1926). For other statistics along similar lines see F. C. Hultgren, Die technik der römischen dichter im epischen und elegischen versmasz, in Neue Jahrbücher für Philologie und Paedagogik 107.745-72 (1873), also Bednara, ALLG 14.600-4 (with bibliographical notes). Cf. Lucian Mueller, De re metrica poetarum Latinorum² 149-56.

page 11 note 7 Ovid 54.6%, Cicero 37.4%, Catullus 35.6% (Steele, PQ 5.225).

page 11 note 8 Juvenal 10.122-4.

page 11 note 9 Catullus 42 in 794 hexameters (5.3%), Vergil 32 in 12814 (0.25%). Cf. Steele, PQ 5.213. According to Mueller, RM² 152, Catullus alone of the classic poets after Ennius has an hexameter consisting entirely of spondees (116.3).

page 11 note 10 The only neuter plural that ends in a short vowel other than a is duǒ, which is really a vestigial dual form (Roland G. Kent, The Forms of Latin §204.IV).

page 11 note 11 Greek forms apart, the only exception to this statement is again duǒ, masculine nominative (or accusative, later replaced by the analogical duōs: Kent, FL §204.IV).

page 11 note 12 The expression ‘initial vowel’, as used in this place, and henceforth, will be understood to include initial vowels preceded by h, the aspiration being of no metrical significance.

page 11 note 13 No distinction is here made between elision and ecthlipsis (cf. note 17). Aphaeresis of the e in est (or es), which allows the preceding word to remain intact, is a different figure, countenanced in positions where elision proper is avoided (see ch. IV, n. 15 ad fin.); but as it nevertheless deprives the verse of a short syllable it has in this dissertation been included in the general term ‘elision’, if not otherwise specified.

page 11 note 14 In all the pentameters examined (see §7) there is no instance of a fourth-declension genitive plural, and only six of rerum, in the fifth declension (Prop. 2.34.28 and 4.1.30; Ov. F 1.62 and 88, P 2.2.12 and 3.1.144). Dierum, the only other genitive plural of classical occurrence in the fifth declension (Kent, FL §297), is not found in these pentameters.

page 11 note 15 In certain nouns of the second declension (and, by analogy, in certain nouns of the first), the longer form could be avoided by using the older genitive plural in -urn (Kent, FL §§223 and 236). So virum (Cat. 68.90, Prop. 3.9.32), deum (Ov. H 15.130, F 2.132, 3.278, 3 334, 3.344, 5.130, T 1.5.70, P 2.8.58), divum (Cat. 76.4), humanuni (Tib. 3.4.26, for which, according to Postgate's critical note, Lachmann conjectured heroum), Danaum (Prop. 2.26.38, 3.9.40, 3.11.14, Ov. H 13.94), Aeneadum (Ov. T 2.261 and 262), and caelicolum (Cat. 63.168). (It may be noted here that macrons and breves are used in this dissertation only where there is reason to call attention to the length or shortness of a vowel.)

page 12 note 16 For the quantity of this final o, see Mueller, RM² 413-8. The increasing tendency to scan it as a short vowel in the off-beat may be exemplified by the treatment of the word nemo in the elegiac poets. In Catullus the o of nemo is always long (64.38, 67.17, 73.5, 81.1), even when the o does not receive the ictus (as is the last example). The word occurs only once in the Corpus Tibullianum, as nem, before the diaeresis (3.13.8). Of seven instances in Propertius (2.14.18, 2.16.19, 2.19.32, 2.30.25 and 31, 2.34.3, 3.16.14) the one that does not suffer elision or enclisis of est is used in the same manner (2.19.32); but the o of nemo in the second hemistich of 2.14.18 would hardly have been elided in this part of the verse if it had not been felt to be short (cf. ch. II, n. 9). In the pentameters of Ovid, nemō always occurs at the end of the first hemistich (A 1.8.100, T 1.1.112, 1.9.44, P 1.3.56, 1.7.40, 3.4.78, 3.6.58, 4.14.44; for quantity of final vowel and position of ictus cf. M 2.687, 3.137, 8.867, T 1.1.99, 5.7.15, P 2.3.23, in the hexameters), nemǒ always at the beginning of the second (F 6.324, T 2.348, P 2.3.16; cf. M 15.600, but also A 1.8.43). In the case of hǒmo, the change was probably hastened by the operation of iambic shortening (Kent, SL⁸ §128.II). Catullus has homǒ (115.8), as well as homō (81.2, 112.1) and hom, at the end of a pentameter (93.2); his contemporary, Lucretius, shows a similar variation (see Bailey's note on 1.66 ad fin.). In Ibis 408, the sole instance in an Ovidian pentameter, the word is scanned as a dibrach. The quantity of the final o in the only other example in Ovid (M 1.78) is obscured by enclisis of est; and the form does not occur at all in Tibullus and in Propertius. Vergil does not commit himself on the quantity of the terminal o in either word; in every instance he either elides it (as in A 5.305 and 383, 9.783, 10.720) or uses it in the syllaba anceps at the end of the hexameter (as in A 5.349 and 9.6).—Ovid scans his own cognomen Nasǒ (e.g. A 2.1.2; cf. Nasō in Cat. 112.1 and 2), and the name of his birthplace Sulmǒ (e.g. T 4.10.3).

page 12 note 17 As for m, “the probability is that when final it was not a real consonant, but merely the sign for nasal quality of the preceding vowel, like the nasalized vowels of French, and that this nasalized vowel was long” (Kent, SL³ §56).

page 12 note 18 Also in a few second-declension neuters, like virus and vulgus.

page 12 note 19 E.g. Quid dubitas quin omni' sit haec rationi' potestas? (2.53). In all the elegiacs under consideration (see §7) apocope of final s occurs only once, in the very last verse of Catullus, where it affects a verb-form: tu dabi' supplicium (116.8). The practice became obsolete during the lifetime of Cicero: Quin etiam, quod iam subrusticum videtur, olim autem politius, eorum verborum, quorum eaedem er ant postremae duae litterae, quae sunt in optimus, postremam litteram detrahebant, nisi vocalis insequebatur. Ita non erat ea offensio in versibus quam nunc fugiunt poetae novi. Sic enim loquebamur: qui est omnibu' princeps, non omnibus princeps, et: vita illa dignu' locoque, non dignus (Orator 48,161). See Mueller, RM² 427-9.

page 13 note 20 In the first book of Ovid's Fasti 72% of all words (including enclitics, as manipulable elements) begin with a consonant. In the same book this proportion remains constant, within a range of 1%, for hexameter and pentameter, and for both hemistichs of the pentameter, considered separately.

page 13 note 21 On this ‘poetic singular’ see Maas, ALLG 12.500-2.

page 13 note 22 In this résumé no attempt has been made, of course, to list exhaustively the brachysyllabic resources that stood at the command of a composer of dactylic verse. Not to mention verb-forms and other parts of speech that have not been considered, there were large numbers of nouns and adjectives that by reason of quantities inherent in their stems could, in one or in several forms, and under the proper metrical conditions, be fitted into a dactylic scheme. It is obvious that the poets found at least the required minimum of short syllables. We are here concerned to inquire to what extent they found them in the declensional terminations that the language placed at their disposal.—It might be w^Tell to quote at the outset Bednara's carefully qualified circumscription of his field, than which that covered by this dissertation is even narrower: “tantum abest, ut ex poeta versificatorem faciamus, qui anxie circumspiciat, unde dactylum promat, ut contra non ignoremus multa non magis necessitati metricae quam ingenio ac licentiae poetae deberi, multa quoque e sermone cotidiano vel volgari ab eo recepta esse, libereque profiteamur studia dactylicorum poetarum linguae Latinae non incommodo, immo summae utilitati magnoque incremento fuisse. Sed quamquam nos non fugit poetam saepe aliter dicere non minus noluisse quam non potuisse, tarnen huic rei plurimum momenti tribuimus in disputatione nostra; idque nobis, puto, concessum erit, quandoquidem nostrum est operam dare necessitati metricae” (De sermone dactylicorum Latinorum quaesti ones, ALLG 14.318-9).

page 14 note 1 With two listings under each head, viz.: under gender, masculine-feminine, as distinguished from neuter; under number, singular and plural; under case, nominativeaccusative-vocative, as distinguished from other cases. (See tables of chapter VII, with preliminary note.) As the nominative, accusative, and vocative cases of masculines and feminines singular (unlike the same cases of masculines and feminines plural, and of neuters both singular and plural) differ from each other in metrically significant respects (cf. §28), a separate subtabulation of these forms has been given. Other disparities of tabulation, like the inclusion, in a class the endings of which are otherwise always long, of forms the endings of which may be short (e.g. in the singular, genitives in -is and ablatives in -ĕ; in the plural, datives and ablatives in -bus) have been provided for partly by the figures given in the tables themselves for short terminations, and partly in the discussion, under the appropriate headings.

page 14 note 2 The final syllable of the casus rectus, whatever the etymology of that syllable, and even if the form itself be a monosyllable, is in this dissertation referred to as its ending (or termination). The -us of avus and the -us of opus are, of course, very different in their origin, but metrically the words are equivalent. Much confusion would be caused, and no purpose served, by an attempt to distinguish between them, or to exclude them entirely, on the ground that the latter is an uninflected stem, while the inflection of the former consists merely of the consonant s (Kent, FL §§226, 258). The i in the plurals of i-stems is taken to be, as in fact it is, part of the stem and not of the inflectional ending; only in the dative and ablative cases, where it appears universally in consonant- as well as i-stems, is it considered part of a disyllabic ending (Kent, FL §257). The only other ending that has been counted in the tabulations as a disyllable is the genitive -īus or -ĭus of certain pronominal adjectives. Whatever the origin of this i (Kent, FL §317.11), it became for all metrical purposes part of the ending, the poets varying its quantity according to the exigencies of their verse.—A syllable to which an enclitic is attached remains in this classification final, even if the suffixed element is inseparable, as in uterque, quisquam, idem, etc., the last syllable (or syllables) of the declinable part of the word being regarded as its ending, the enclitic as a following vocable, with initial consonant or vowel. Ī-dem and ĭ-em are respectively so divided (Kent, FL §315). Reduplicated pronouns like quisquís, quidquid (which in the verses studied occur only in these forms) and sese are counted as one word.

page 14 note 3 In giving references, these five works of Ovid will be designated by the letters A, H, F, T, and P respectively, and the Epistula Sapphus (considered in all generalizations as one of the last third of the Heroides) by the letter S.

page 15 note 4 The more recent Teubner editions of Ovid's Fasti, and of his Tristia and Epistulae ex Ponto, by Ehwald and Levy, listed in the bibliography, were unfortunately not available to me for extended use in the circumstances under which this work was begun. All specific references have been checked against them; and this comparison has borne out what reasonable conjecture would assume, that with a text so extensive and in the main so well-established as that of Ovid no significant differences in the statistics would have resulted from the use of these editions.

page 15 note 5 Gummere, The Neuter Plural in Vergil 11; Herr, The Additional Short Syllables in Ovid 8; Glenn, The Neuter Plural in Latin Iambic and Trochaic Verse 7-8; Zenn, The Neuter Plural in Latin Lyric Verse 7-8.

page 15 note 6 According to the Thesaurus Linguae Latinae s. v. (1.1164), “invenitur pluralis inde ab Ov.” But there are only two examples in the whole of Ovid (M 2.649, P 1.3.83), both in the form aevis, modified by omnibus. So also Plin. N. H. 16.238: omnibus ex eo aevis.

page 15 note 7 Gummere, NPV 25.

page 16 note 8 Mueller, RM² 156: “in pentametri posteriorem partem spondeum admittere ne christianorum quidem poetarum ultimi, quorum ego libros legerim, sustinuere. quamquam et hoc licentiae et quid non? invenitur in lapidibus inscriptis.” Cf. F. Vollmer, Römische Metrik, in Gercke-Norden, Einleitung in die Altertumswissenschaft³ 1.8.14: “Spondeus im 2. Kolon findet sich nur in kunstlosen Grabinschriften.”

page 16 note 9 Mueller, RM² 362: “pentametri ... posteriorem partem ... longe cultiorem altera pierique, si non omni no segregarunt a synhaeresi, tantum brevium, non longarum aut me-diarum voluere perviam elisioni.” From the time of Ovid, “elisionem brevis si modo ullam admisere—nam et plane abstinuerunt plerique—, adhibuerunt perraram et ferme in e vel i” (ibid.).

page 16 note 10 In the narrowest sense, the construction within a clause is in the Augustan authors rarely allowed to run from couplet to couplet. Of almost 12000 distiche examined in the works of Tibullus, Propertius, and Ovid, there are only three (all in the fourth book of Propertius: 5.14, 8.68, and 10.40) that are not followed either by some mark of punctuation (in the texts named in §8), or by a distich beginning with a conjunction or with a word to which a conjunctive enclitic is attached. In the 322 distiche of Catullus, on the other hand, there are 14 examples of syntactical involvement too close to allow of either punctuation or conjunction. (If the parenthesis in Prop. 4.8.6, coming as it does in the midst of a clause, not be allowed as punctuation in the strictest sense, we have here another example in the Augustan poets—again in the fourth book of Propertius. In Prop. 4.1.123, whether we read quam or qua, with or without punctuation after the preceding line, the relative is of course conjunctive in force.—In all exactness, Ov. F 3.524 should have been cited above, as a fourth example; but the line is differently taken, by Merkel and by Ehwald and Levy as going with 525, in the next couplet, by Frazer and by Landi as going with 523, in the same couplet. The latter punctuation is corroborated by Ovid's unvarying practice elsewhere, not to mention other considerations, grammatical and stylistic.)

page 16 note 11 Cf. ch. I, n. 5.

page 16 note 12 The distribution of spondees and dactyls between the two hemistichs of the pentameter (or, perhaps more correctly, between the second hemistich of the pentameter and the rest of the distich, including the hexameter) seems to have been of greater importance metrically than any analogous distribution between the pentameter as a whole and the hexameter. Actually, the proportion of dactyls in the first four feet of the alternate hexameters of the Fasti is 54.2%, a figure not significantly lower than the 54.6% given by Steele (PQ 5.225) for the wholly hexametric Metamorphoses—with which the Fasti, of all Ovid's works in the elegiac meter, may best be compared in point of length, date of composition, and subject-matter. The proportion of dactyls in the first four feet of the alternate hexameters of Catullus' elegiacs (318 verses, with 65.9, 66.91, 68.157, and 107.3 omitted, as defective in relevant respects) is 36.4%, which likewise differs by less than 1% from the 35.6% given by Steele (PQ 5.225) for 398 successive hexameters in Catullus 64. It is of course impossible to determine what the pace of the hexameters would have been if the same poets under similar conditions had been writing of the same subjects in hexameters exclusively; but it can be said with reference to Catullus and Ovid—the only two of whom we have both heroic and elegiac verse for comparison—that their hexameters hardly vary in this regard, whether they follow one upon the other or alternate with pentameters.—For proportions of dactyls and“ spondees in the first hemistich of the pentameter, see table XX.

page 17 note 13 E.g. 77.6% of all neuter plurals in the pentameters of Ovid's Fasti occur in the second hemistich (see §15 and table XV). If we disregard in both hemistichs neuter plural quae and haec, which, as long monosyllables, generally fall into the first hemistich (see §22 and table XXXIX), and consider only forms ending in ă, not elided and not final in the pentameter, the proportion for the second hemistich rises to 79.8% (see table XXXV). These differences have been accounted for partly in the tables, as indicated above, and partly in the discussion, in each instance under the appropriate head (cf. note 1 ad fin.).

page 17 note 14 The statistical differences between successive books of the Amores, Fasti, Tristia, and Epistulae ex Ponto did not seem great enough to warrant inclusion of the 36 tables (two to a book) which would have been required in addition to those already given.

page 17 note 15 For discussion of the authorship of the ‘Messallasammlung’, and bibliography, see Schanz-Hosius, Geschichte der römischen Literatur⁴ 2.177-9, 185-91.

page 17 note 16 For discussion of the authenticity of the last six ‘Epistulae’ (to three of which the title ‘Heroides’ cannot properly be applied) and of the Epistula Sapphus, with bibliography, see Schanz-Hosius, GRL⁴ 2.214-5, 220-3.

page 18 note 17 See §§14-5.

page 19 note 1 Henceforth to be abbreviated: NAV neuter plural. As no distinction was made among these cases in the tabulation, their separate totals are not at hand. The number of vocatives would in the nature of things be small. In all of Vergil there are only 30 occurrences of neuter vocatives, equally divided between singular and plural (Gummere, NPV 11).

page 19 note 2 In ALLG 12.479-550 (1902).

page 19 note 3 It is not part of our task to distinguish between NAV neuter plurals logical and ‘poetic’ (for a definition of the latter term see ch. I, n. 2). But it is not to be supposed that elegiac poets, handling a meter the dactylic requirements of which were more stringent than those of purely hexameter verse, would not have availed themselves of a means of facilitating dactylic composition already firmly established for the hexameter. Examples are so common that illustration is hardly in point; but see e.g. Ov. A 2.10.24.

page 19 note 4 See ch. II, n. 8.

page 19 note 5 It will be noted that although the proportion of NAV neuter plurals among all declinable nouns is in Vergil almost twice the corresponding proportion in ‘standard’ prose, neuter nouns altogether constitute less than a third of all declinable nouns in Vergil, as compared with almost a half in prose—neuter singulars in general and cases of the neuter plural other than NAV being much less frequently used by Vergil.

page 19 note 6 The only figures available for Vergil and for a sampling of ‘standard’ prose result from tabulation of declinable nouns only (with certain exceptions even among these: see Gummere, NPV 11), whereas the figures given here for the elegiac pentameter include declinable pronouns and adjectives. There is no obvious reason, however, why nouns alone should not be broadly representative of the whole class of words; that they are in fact so would seem (if the circular argument be admitted) the more probable in view of the practical identity of the proportions given in §14 for the hexameters of Vergil and for the pentameters of the Augustan elegists.

page 19 note 7 Henceforth, when a generalization is made without other qualification, the facts or figures adduced will be understood to apply to the pentameters of the poet in question (in the case of Ovid, to the pentameters of the five works listed in §7).

page 20 note 8 All these figures come from the tables in chapter VII. As the tables have been arranged for ready consultation, it seems unnecessary, here and in what follows, to give the specific reference in every instance.

page 21 note 9 Between the first and second books of TibulluS; in fact, the preponderance of NAV neuter plurals in the second hemistich of the pentameter drops from 79.1% to 69.8%.— Duff says of this second book that it “is not much over half the size of its predecessor: there are manifest gaps in the third elegy, and likely gaps elsewhere, which have produced the impression of inadequate finish” (A Literary History of Rome from the Origins to the Close of the Golden Age² 556).

page 21 note 10 Ov. T 4.10.24.

page 21 note 11 Tib. 2.1.90.

page 21 note 12 Ov. P 4.15.30. For the reservations with which the term ‘rhyme’ is used in referring to homoeoteleuton, see §51.

page 21 note 13 See §§45-54.

page 22 note 14 The statistics for NV masculine-feminine singulars are to be obtained by adding the separate figures given for nominatives and for vocatives singular, masculine and feminine, at the foot of the tables in chapters VII and VIII. See ch. II, n. 1, and §28.

page 22 note 15 This proportion cannot be obtained directly from the tables. 50.3% is the ratio of 152 (129 + 23: the number of MF singulars, nominative and vocative, respectively, the endings of which, as used in the pentameters of Catullus, actually provide metrically short syllables) to 302 (268 + 34: the total number of MF singulars, nominative and vocative, in the same pentameters). The figures cited in parentheses are to be found at the foot of table I.

page 22 note 16 The expression ‘percentage of utilization’ will be employed to indicate the percentage of forms in any category the endings of which, as they are actually used in each instance, provide metrically short syllables in the scansion. See preliminary note to chapter VII.

page 22 note 17 This proportion, like the one with which it is compared (see note 15 above), cannot be obtained directly from the tables. 76.5% is the ratio of 417 (395 + 22: the number of MF singulars, nominative and vocative, respectively, the endings of which, as used in the pentameters of Propertius, book IV, actually provide metrically short syllables) to 545 (516 + 29: the total number of MF singulars, nominative and vocative, in the same pentameters). The figures cited in parentheses are to be found at the foot of table IX.

page 23 note 18 In this connection it should be borne in mind that in Catullus there are roughly three times as many short syllables in the second hemistichs as in the first, in the fourth book of Propertius roughly twice as many. The number of metrically significant short syllables in the second hemistich was fixed by reason of the fact that dactyls could there not be replaced by spondees; in the first hemistich, where substitution was permitted, the later poets in general showed a greater ability to obtain, or a greater inclination to use, dactyls. See table XX, and cf. §§80-1, with notes.

page 23 note 19 Neuter duo, the one apparent exception to this statement (see ch. I, n. 10), has been classified in the tables w^rith NAV neuter plurals ending in a.

page 23 note 20 Except in Tibullus I, where haec, neuter plural, occurs once in the first hemistich (5.6) and once in the second (4.74).—It need hardly be added that from a metrical standpoint haec and quae, neuter plural, differ not at all from haec and quae, feminine singular. Attention is here confined to the former only for the reason that table XXXIX, in which the relevant figures appear, was designed, as we have said, “to account for those NAV neuter plurals that do not yield metrically significant final short syllables.” Haec and quae, feminine singular, are subsumed under type III in table XL (see §32).

page 23 note 21 See in particular §87.

page 24 note 22 Factors not pertinent to all long monosyllables will be operative in connection with relatives and interrogatives especially. As such words are ordinarily not far postponed, their incidence in one hemistich or the other will depend in no small degree upon the poet's practice with regard to length of clauses and enjambement of thought from distich to distich (see ch. II, n. 10). It may suffice here to point out that although in the first hemistich the 318 pentameters of Catullus considered in this dissertation (for excluded lines see §8) contain exactly the same number of occurrences (37) of the interrogative and relative quis, qui, quisquis and quicumque (in any of their forms) as the first 318 pentameters of Ovid's Fasti (1.2–636), in the second hemistich of these same pentameters Ovid has three and a half times as many as Catullus (14, against 4). This is to be sure a small enough sampling of Ovid's voluminous work; but the results accord with stylistic mannerisms that we have noted, and shall note, elsewhere. In his elegiac verse Ovid tends toward a less elaborate sentence-structure than Catullus (cf. ch. II, n. 10); his work abounds in clauses so short that they can begin and end within the compass of a hemistich. Furthermore, this facile brevity betrays him into jingling repetitions (cf. §43, with notes): within these relatively few pentameters he has quo tuealur habel (1.32), quad tueatur habet (1.86); quidquid in orbe fuit (1.284), quidquid in orbe patet (1.494); and, in one line, sitque quad a tergo, sitque quad ante, vides (1.92). This last may be well enough in a description of the two-headed Janus; but the pentameter is nevertheless peculiarly Ovidian in the brittleness with which it falls apart. A third instance of quidquid, postponed to third place in its clause, occurs in a distich (1.635-6) in which there is a similar exploitation of balance:

Catullus draws a longer breath: the one instance of quisquis in his elegies (68.28: see Eilis's note ad Joe.) occurs in the rising rhythm just before the diaeresis, and provides the subject of a clause that continues not only through the second hemistich, but through the whole of the succeeding hexameter. (The indefinite qui, qua, quad has not been considered in this note. Because of its enclitic character it cannot begin a clause; and several of its forms differ in quantity from their relative and interrogative counterparts.)

page 24 note 23 See ch. II, n. 9.

page 24 note 24 Ov. T 3.1.72 is an example (unique in these pentameters, in either hemistich) of est attached to a neuter plural form. Such enclisis occurs more commonly, as one would expect, with nominative singulars, which serve as, or agree with, the subject of est (cf. ch. IV, n. 15 ad fin.).

page 24 note 25 Ov. A 3.6.46, H 4.144, P 4.12.32; Cat. 76.6, 94.2 (note the successive elisions in this hemistich); Prop. 2.3.50, 2.26.28 and 52, 2.34.6, 4.1.82.

page 25 note 26 Catullus has 24 such verse-endings in 320 pentameters (7.5%), Tibullus 26 in 618 (4.2%), Propertius 86 in 2003 (4.3%), Ovid in his Fasti only 22 in 2485 (0.9%). Cf. Ernst Bickel, Antike Metrik, in Gercke-Norden EAW² 1.578: “Doch hat unter den Lateinern Ovid auch hier” [i.e. not only at the end of the first hemistich, but also at the end of the second] “sich gescheut, durch kurzvokalischen Auslaut den überhaupt nur aus der Hebung bestehenden Daktylos zu bilden”; also Isidor Hilberg, Die Gesetze der Wortstellung im Pentameter des Ovid 378: “Gesetz E. Kurzvokalischer Ausgang des Pentameters wird womöglich vermieden.” S. Hornstein, Die Wortstellung im Pentameter des Tibull und Ps.-Tibull 87, cites as a deviation from Ovid's practice the fact that “Hilbergs Gesetz E ... im tibullischen Korpus keine Anwendung findet.”—Variation among the books of Tibullus and of Propertius is confined within 1% (Tibullus I 18 in 403, or 4.5%, II 8 in 215, or 3.7%; Propertius I 16 in 353, or 4.5%, II 32 in 681, or 4.7%, III 20 in 495, or 4.0%, IV 18 in 474, or 3.8%). The proportion in pseudo-Tibullus (17 in 236, or 7.2%) almost equals that in Catullus.

page 24 note 27 Non-observance of what came to be the ‘disyllabic law’ (see §§87 and 91, also table XLVIII) did, to be sure, make it possible to use in this final position NAV neuter plurals ending in tribrachs, which could not have been placed elsewhere in the verse without elision. Three of the five examples in Catullus of NAV neuter plurals ending a pentameter are of this type : flāgitia (67.42), supercilia (67.46), and maria (115.2); the other two are ‘regular’ disyllables, bona in both instances (68.158 and 77.4). But the first two books of Propertius, although they show 19 NAV neuter plurals ending a pentameter, and 197 polysyllabic verse-endings (see tables XXXIX and XLVIII), have only three polysyllabic NAV neuter plurals ending a pentameter (the word ‘polysyllabic’ being understood to refer, here and henceforward, to words of three or more syllables): auxilia (1.1.26), folia (2.3.12), and maria (2.18.38). Of the three final NAV neuter plurals in Tibullus (1.4.84, 1.9.18 and 78) only the first is polysyllabic: magisteria. In all of Ovid there is only one example: adulteria (T 2.514). This consideration therefore seems not to have been of great practical importance in furthering the use of such NAV neuter plural forms.

page 26 note 1 See Kent, SL³ §128.11, FL §308.I. On the quantity of the final o in the nominative and vocative singular of certain nouns of the third declension, see ch. I, n. 16.

page 26 note 2 See ch. I, n. 20.

page 26 note 3 See table XIX.

page 26 note 4 See ch. III, n. 6.

page 26 note 5 These proportions, which for lack of space have not been entered in the tables, can readily be computed from the figures there given.

page 27 note 6 NAV neuter plurals show no corresponding decrease (see §14); but they are often used as ‘poetic plurals’ even in personal connections. Cf. ch. III, n. 3.

page 27 note 7 See ch. II, n. 10.

page 27 note 8 Note Ovid's use of the paratactic construction:

With the former of these distiche cf. Verg. A 1.159-61.

page 27 note 9 See ch. VI. n. 2.

page 28 note 10 See §18.

page 28 note 11 See §§14, 15, 16, 19, 20, and cf. §§31, 46, 49, 52, 53, 81.

page 28 note 12 For definition of this term, see ch. III, n. 16.

page 29 note 13 It happens rarely, after the fashion of Greek poetry, that the ictus falls on a final short vowel preceding an initial double consonant, as in Cat. 67.32:

So also Tib. 1.5.28, 1.6.34, and (in an hexameter) Cat. 64.186.

page 29 note 14 It happens even more rarely that a syllable of this type remains closed, and therefore long, before an initial vowel. I have noted only one example, in Propertius (2.24.4). and here the word in question immediately precedes the diaeresis :

page 29 note 15 But Catullus shows a relatively low percentage of utilization in both hemistichs (53.1% and 80.4% respectively). In the first hemistich he loses almost half (23/49) of his short vowel terminations, NV masculine and feminine singular, by elision (the elided o of leno, 103.4, being regarded as potentially short; but cf. ch. I, n. 16), in the second hemistich 11 of 92 by elision, six more in the syllaba anceps (cf. ch. III, n. 26), and one in the Greek rhythm referred to in note 13 above. As compared with Catullus' 12.0%, Propertius in all his second hemistichs loses only 2.9% of these short vowel terminations, NV masculine and feminine singular, by elision, Ovid only 2.3%, and Tibullus only 1.6%. Pseudo-Tibullus, and Propertius in his last book, lose none at all. (See Table XLI for further figures.) Moreover, all but one of the 93 examples of ‘elision’ of these endings in the second hemistichs of Ovid are really examples of enclisis of est or es, final in the pentameter (the exception being H 10.86), as are all four examples in Tibullus (1.4.4, 1.10.4, 2.6.10 and 44). In Propertius 8 of 29 examples do not involve such enclisis (2.9.42, 2.12.10, 2.14.18, 2.26.30, 3.13.6, 3.16.4, 3.17.8, 3.21.26), in Catullus 5 of 11 (68.100 and 152, 87.2, 94.2, 100.6). Only Catullus has the enclitic est falling at the fifth ictus, followed by a trisyllabic verse-ending: multivola est mulier (68.128). For the scope of the term ‘elision’ in this dissertation, see ch. I, n. 13.

page 30 note 16 Pseudo-Tibullus agrees with Ovid, in that the proportion in the second hemistich is much higher than that in the first. But both proportions are far below Ovid's, and in any case the figures are too small to bear exhaustive analysis.

page 30 note 17 In Catullus, 28 of 68 forms of type II in the second hemistich (41.2%) are final in the pentameter; in Tibullus 75 of 167 (44.9%), in pseudo-Tibullus 23 of 60 (38.3%), in Propertius 155 of 461 (33.6%), in Ovid 504 of 2250 (22.4%). The Fasti show only 113 in 601 (18.8%). See table XLI.—All these figures are for NV masculine and feminine singulars only.

page 30 note 18 19 of the 34 vocatives singular in the pentameters of Catullus (55.9%) are found in the second hemistich. See table I.

page 30 note 19 Cat. 68.20 and 92, 101.6; Ov. F 4.852. Cf. Ov. H 9.166.

page 30 note 20 See §15, and tables in chapter VII.

page 31 note 21 F 6.722.

page 31 note 22 F 2.392.

page 31 note 23 F 3.140.

page 31 note 24 “Saepe a vocativo petunt auxilium poetae dactylici, ita ut eorum nominum quorum casus obliqui metro non apti sunt vocativum adhibeant, cum personas vel etiam res appellent libenter usi hac figura rhetorica iam a Graecis, praecipue a poetis Alexandrinis, usurpata” (Bednara, ALLG 14.568). Cf. Erwin Hampel, De apostrophae apud Romanorum poetas usu, especially 33-4 and 42.

page 31 note 25 Elision of the final vowel of cretic words was generally avoided (Mueller, RM² 342-3).

page 31 note 26 In the Fasti of Ovid there are 52 instances of vocatives immediately preceding the final disyllable; in 33 of these the final disyllable itself is a form of tuus, in three more a form of meus (3.472, 3.482, 6.652). Tibi would fit equally well (Ov. A 3.13.2), or mihi, as in the thrice-repeated hemistich of Catullus referred to above, in §36. The formula is used by Catullus only once more (79.2; cf. Prop. 2.25.4). In Tibullus we do not find it; in pseudo-Tibullus it occurs only twice, in one poem (3.14 and 34;. Propertius, who does not use it at all in his first, and only three times in his second and longest book (15.2, 25.4, 34.16) has twelve examples in books III and IV (3.3.30, 3.5.6, 3.9.44, 3.13.10, 3.15.42, 3.17.6, 3.21.26, 4.1.62, 4.4.18, 4.5.76, 4.7.86, 4.9.2). We may say then that this usage is characteristic of the later Propertius and of Ovid. (In this note, as generally in this dissertation, specific references are given for all the occurrences of a phenomenon under discussion only when its relative rarity makes such citation at once more significant and more feasible.)

page 32 note 27 A 3.13.2.

page 32 note 28 6.183-4.

page 32 note 29 5.52.7.

page 32 note 30 4.10.27-30. In Vergil there is an interesting example of a sudden shift to the second person at the end of a period, with no metrical advantage:

(A 2.56: on the reading, see Conington's note ad loc). In Aeneid 7.685, on the other hand, a similar shift to the vocative, again at the end of a period, is used with metrical advantage. Homer introduces speeches of Eumaeus fifteen times in the Odyssey with the words (14.55 and passim in books 14-17, and again in 22.194), although might have scanned as well (cf. 14.401, and, for the hiatus, 17.477).—In the Odyssey only Eumaeus is so apostrophized. For a discussion of this phenomenon, and a comparison with the Iliad, where the names of Patroclus, Menelaus, and a few others are so treated, generally with metrical advantage, see Ameis-Hentze, Anhang zu Homers Odyssee, on ξ 55; also G. W. Nitsch, Die apostrophe in Ilias und Odyssee, in Philologus 16.151-4 (1860).

page 32 note 31 On this whole subject, see, in addition to the works of Bednara and Hampel already cited in note 24, Gaetano Curcio, L'apostrofe nella poesia latina, and Walther Abel, Die Anredeformen bei den römischen Elegikern.

page 33 note 32 Cf. tables XLII and XLVIII.

page 33 note 33 In the pentameters of all except Catullus (see table XLII).

page 33 note 34 For the figures and proportions given in this paragraph see table XLII.

page 33 note 35 68.34-5.

page 34 note 36 These figures do not include ilia, ipsa, and ista, NAV neuter plural, which are found less frequently in all three positions. None of these forms occurs in the pentameters of Catullus or of pseudo-Tibullus; in those of Tibullus there are three instances, against 46 of the singular forms, masculine and feminine, in those of Propertius eight, against 121, in those of the five works of Ovid examined 64, against 727. The one notable thing about these NAV neuter plurals, as compared with the NV masculine and feminine singular forms, is perhaps the fact that ista—at least in Ovid—is much more frequent among them, especially in the third position. In the Ovidian pentameters under consideration, ille, ipse, and their feminines outnumber iste and ista, NV masculine and feminine singular, more than eleven to one in the last position (482/42); whereas ista, NAV neuter plural, occurs as often in this same position as ilia and ipsa together (illa 16 times, ipsa twice, ista 18 times). Ista is the only one of the three neuter plurals that Propertius there uses (3.6.8, 3.18.20). The sole example of this same ista in the pentameters of Tibullus falls in the third position (2.4.20).

page 34 note 37 2.24.34. Cf. Cic. Tusc. Disp. 1.49,117: supremus ille dies-, ibid. 118: eo tarnen simus animo, ut horribilem ilium diem aliis, nobis faustum putemus; Sen. Ep. 4.9: ultimam illam horam; Plin. Ep. 6.20.15: aeternam ... illam et novissimam noctem.

page 34 note 38 1.8.47-8 and 10.33-4. For non tardo pede cf. 3.5.16, and see note 78 init.

page 34 note 39 3.1091.

page 35 note 40 Cf. Cic. Tusc. Disp. 1.38,91: Itaque non deterrei sapientem mors quae propter incertos casus cotidie imminet, propter brevitatevi vitae numquam potest longe abesse, quominus in omne tempus rei publicae suisque consulat ... In one period of eight lines (3.1068-75) Lucretius has quern, quia, quam, quoniam, quo, quae ... cumque.—This is not to imply that the elegiac poets do not themselves make abundant use of the syndetic relative. We may compare Ovid's expression of a correspondent thought:

page 35 note 41 The numerous examples of ille (or illa) so placed (as in Verg. G 1.331, A 4.331, 5.439) refer invariably, it would seem, to a person standing in some contrast with the subject of the clause immediately preceding. In the first distich quoted from Tibullus, ilia of course stands over against tu, the subject of utere; in the second, it serves to distinguish the subject of imminet and of venit from the unspecified subject of accersere. But its postponement in its own clause lessens the part it plays in pointing the contrast, which has already been effected by the asyndeton.

page 35 note 42 T 5.2.1-2.

page 35 note 43 P 1.6.5-6.

page 35 note 44 H 2.20.

page 35 note 45 H 7.32, 34, 36. Cf. H 6.150, 152, 154, 156.

page 35 note 46 Cf. the use of ipse three times in one line: A 2.12.14 and 3.11.18.

page 35 note 47 In the Fasti: ille 19 times out of 104 (18.3%), and twice in addition by mihi (4.228 and 5.278); ipse 12 times out of 33 (36.4%). (Here, as elsewhere in this discussion, figures given for ille, ipse, and iste include those for illa, ipsa, and ista, NV feminine singular.)

page 36 note 48 Quintilian 10.1.88.

page 36 note 49 F 5.188, T 1.10.50. See also A 1.8.84, S 26, 206. Cf. alter et alter (A 2.1.30, 2.10.10), unus et alter (A 2.5.22). One cannot refrain from quoting the elder Seneca's comment: “non ignorabat vitia sua, sed amabat” (Contr. 2.2.12).

page 36 note 50 H 12.18; also A 1.7.26, H 9.96, 13.116, P 3.3.46. Moritz Haupt, Opuscula 2.186 (in a discussion of Aeschylus, Prometheus 762: cites from other works of Ovid AA 3.668, Ibis 404 [406] and 584 [586]; to which Johannes Vahlen, Gesammelte Philologische Schriften 2.106, adds F 5.551 (the only example that does not occur in the second hemistich of a pentameter), beside Verg. A 4.233 and 12.639 [638]. Vahlen comments, ibid. 106-7: “darin kann ich Haupt nicht beistimmen, wenn er meint, der oder die Dichter seien nur durch die Bequemlichkeit des Versmasses zu dieser Wortstellung veranlasst worden: sie dient vielmehr ersichtlich demselben Zweck, wie die entsprechenden griechischen Beispiele, durch die Aneinanderrückung der verschiedenen Pronomina den in ihnen liegenden Gegensatz zu schärfen.”—On the causal ablative with ab, see Loewe, Über die Praepositionen a, de, ex bei Ovid 11-3; Karl Guttmann, Sogenanntes Instrumentales ab bei Ovid 15-20; John C. Rolfe, A, ab, abs, in ALLG 10.504-5 (1898); Thesaurus Linguae Latinae s. v. a, ab (1.28-34). In Ovid's free use of the preposition ab, and even more strikingly perhaps in the greater frequency of such locutions as umida de guttis (A 3.5.6; cf. T 3.3.82), de rapidis immansuetissime ventis (H 17.37), de qua pariens arbore nixa deast (H 20.100), ad nostros non leve crimen avos (T 2.472), we may note a stage in the analytical development that rendered case-endings otiose in the Romance languages. Cf. Stolz-Schmalz, Lateinische Grammatik⁴ 395, 407-8. Ovid may frequently have availed himself of the preposition ab for no better reason than that it provided a short syllable where he needed one. In the line et tunicas lacrimis sicut ab imbre gravis (H 10.138), ab imbre in the second hemistich is clearly coördinate with lacrimis in the first; even Guttmann's elaborate analysis, arguing from the presence or the absence of ab, could hardly read a shade of meaning into the one that was not to be felt in the other. So also, in an hexameter, hic iaculo pisces, illic capiuntur ab hamis (AA 1.763); cf. Verg. G 1.234: semper sole rubens et torrida semper ab igni.

page 36 note 51 2.3.74 and 4.2.

page 37 note 52 See §17.

page 37 note 53 Prop. 1.11.26.

page 37 note 54 Prop. 2.28.56.

page 37 note 55 Prop. 1.7.24.

page 37 note 56 Tib. 2.5.96.

page 37 note 57 Tib. 1.10.38.

page 37 note 58 Ov. T 5.12.62.

page 38 note 59 These figures, striking as they are, may be regarded as minimal. Only pairs of words in agreement with each other or in coördinate construction within a clause were considered as examples of such a pattern—and then only if the ending of each word accounted for a short syllable in the meter. These restrictions excluded pairs of words the same in gender, number, and case, but standing in separate clauses:

pairs of words the same in gender and number, but differing in case:

pairs of words similar in ending, but differing in gender, number, and case:

as well as others in which the inflectional ending of one word for whatever reason failed to provide a short syllable:

Nor were the relatively infrequent pairs of forms other than NAV neuter plurals and NV masculine-feminine singulars included; for these see §§60 and 63, with notes.

page 38 note 60 The only other hemistich-ending that would leave room for such a combination would be the monosyllabic, which is extremely rare in the second hemistich (see ch. VI, n. 42), and far from common in the first (see note 61). When found in the first hemistich, they are ordinarily preceded either by another long monosyllable, or by a pyrrhic form (which must in turn be preceded by a word ending in a long syllable, if not itself a long monosyllable); in either case, a pair of declinable forms ending in short syllables can scarcely find room. Cf. Vollmer, Römische Metrik, in Gercke-Norden EAW³ 1.8.14: “Das erste, meist durch Spondeen retardierte Kolon geht auf Monosyllabon aus nur so, dass Monosyllabon oder pyrrichisches Wort vorhergeht.” (By ‘Monosyllabon’ Vollmer seems to mean, in the second instance as necessarily in the first, a long monosyllable.) Cf. ch. VI, n. 35 ad fin.

page 38 note 61 Of a hundred pentameters, chosen in each instance as specified, the following numbers have the first hemistich ending in words of the rhythm designated (in the case of monosyllables the preceding word is included) :

	–/	—/	—-/	—–/
Cat. 65.2-68.36	53	14	0	0	1	2	4	1	10	11	4
Tib. 1.1.2-3.24	51	13	0	0	0	0	1	0	1	27	7
Tib. 3.1.2-5.8	48	20	0	0	0	0	1	0	7	17	7
Prop. 1.1.2-6.24	34	26	0	0	4	0	2	0	14	16	4
Prop. 4.9.26-11.102	43	22	2	0	3	0	1	0	4	21	4
Ov. A 1, Ep. 2-5.18	42	15	0	0	1	0	2	2	5	20	13
Ov. F 1.2-200	41	9	0	0	1	0	4	2	10	25	8
Ov. P 4.13.8-16.52	37	11	2	1	4	0	2	1	6	22	14

(Catullus 67.12 was omitted as corrupt.) In all these samplings, the first hemistich ends most commonly in a spondaic word, with molossics and anapaestics following either second or third (except in the selections from the Fasti and from the Epistulae ex Ponto, in which molossics are slightly outnumbered by iambics and by choriambics respectively).

page 39 note 62 I have noted only one in Catullus (66.38), one in pseudo-Tibullus (2.10), nine in Propertius (1.12.20, 1.17.12, 2.9.36, 2.12.22, 2.16.2, 2.20.34, 2.23.2, 4.3,30, 4.8.30), and 23 in the five works of Ovid here examined (A 1.4.2, 1.7.2, 2.13.12, 3.9.32; H 1.90, 2.110, 4.64, 5.98, 16.108, 17.128, 20.242, S 40; F 1.392, 2.644, 3.386, 5.140, 5.608; T 3.10.24, 4.2.42, 4.10.58, 5.4.2, 5.11.24; P 1.5.2).

page 39 note 63 Ov. S 39-40; cf. H 4.144. Among the pentameters cited in note 62 such parallelism is found in Prop. 1.12.20, 2.9.36, 2.16.2, Ov. F 5.140, T 5.4.2, and to a somewhat lesser extent in Tib. 3.2.10, Prop. 2.12.22, Ov. A 3.9.32, H 2.110, F 2.644, 3.386. Cf. also Tib. 1.4.4, 1.5.64, Prop. 2.5.28, and the line from the anecdote told by the elder Seneca (Contr. 2.2.12; see ch. VI, n. 15): semibovemque vir urn semivirumque bovem (AA 2.24).

page 39 note 64 See ch. III, n. 26.

page 39 note 65 According to Vollmer, Römische Metrik, in Gercke-Norden EAW³ 1.8.14, only the Christians are guilty of this: “diese bringen in 3” [i.e. at the third ictus of the pentameter] “nicht selten sogar syllaba brevis und Hiat.”

page 39 note 66 See §52. As Norden observes (Antike Kunstprosa 2.840, n. 2): “Dass sich übrigens gerade bei den Elegikern im Pentameter so viele Reime finden, erklärt sich ganz einfach aus der bekannten Manier, Substantiva von ihren gleichauslautenden Attributen zu trennen, cf. Eichner 1. c. 35f.” (For the works here cited see note 69.)

page 40 note 67 Transactions of the American Philological Association 40.31-62 (1909).

page 40 note 68 1.20.35-8 (the italics are Foster's).

page 40 note 69 TAPA 40.37. On the whole subject see Wilhelm Grimm, Zur Geschichte des Reims, and Ernst Eichner, Bemerkungen über den metrischen und rhythmischen Bau, sowie über den Gebrauch der Homoeoteleuta in den Distichen des Catull, Tibull, Properz und Ovid; for more conservative views of the place of this phenomenon in Latin poetry, see Pietro Rasi, De Elegiae Latinae Compositione et Forma, and Eduard Norden, Die Antike Kunstprosa, Anhang I (vol. 2, pp. 810-908): Über die Geschichte des Reims, with extensive bibliography in footnotes.

page 40 note 70 Even those writers who are most concerned to find rhyme in Latin verse are of course aware of the difference (Grimm, Zur Geschichte des Reims 159-60, Eichner, Bemerkungen etc. 42, Foster, TAPA 40.32, n. 3); whether they make enough of it in their argument is another question. Foster, for instance, treating under the head of rhyme even such pairs of words in agreement with each other as we have discussed in §§17 and 45-7, says of the line munera quanta dedi vel guaita carmina feci (Prop. 2.8.11) that it “shows a rimed pair in the first half, as well as the second” (TAPA 40.37). Here he carries even further than Grimm and Eichner a tendency to consider as a kind of rhyme any consonance of inflections, regardless of their proximity to each other, and of their relation to the main, or indeed to any, caesura; but this is surely overplaying an unavoidable feature of Latin accidence.

page 41 note 71 See Roland G. Kent, The Educated Roman and his Accent, in TAPA 53.63-72 (1922); La lecture à haute voix des vers latins, in Revue Internationale de l'Enseignement 79.321-35 (1925); L'accentuation latine, problèmes et solutions, in Revue des Études Latines 3.204-14 (1925). Cf. ch. VI, n. 36.

page 41 note 72 E.g. by writing externis virtus incomitata bonis (P 2.3.36) instead of virtus externis etc., Ovid obtains a chiastic word-order, and differentiates the first hemistich by one conflict the more from prose (see §§93-4). The latter consideration applies also to the line per medias Histri plaustra bubulcus aquas (P 4.7.10); more important perhaps in this instance is the fact that the sequence dactyl-spondee is preferred to the sequence spondee-dactyl (see table XX).

page 41 note 73 The proportions given in these paragraphs (with the exception of those that have been especially computed for the last elegy of Propertius) are to be found in table XLIV.

page 41 note 74 In the last seven of the Heroides the proportion of hemistichs with final words in agreement falls to 29.9% (2.7% below his average); in the first fourteen the proportion with final words both agreeing and rhyming reaches 21.9% (1.8% above his average).

page 42 note 75 66% and 44% respectively. This same elegy has five successive pentameters (26-34), another (2.34) six consecutive verses, both hexameter and pentameter (85-90), with internal rhyme of this sort.

page 42 note 76 But the number of agreements, and of agreements accompanied by rhyme, is proportionately much higher in book III than in book II. Compare the first and second columns of table XLIV with the third.

page 42 note 77 This proportion is true of the last book as a whole, and of the last elegy in particular.

page 42 note 78 It is evident that agreements of the first type will rarely be found in the work of a poet who avoids terminating the pentameter in a short vowel (see ch. III, n. 26), and that agreements of the second type will virtually be eliminated by the establishment of the disyllabic law (see §§91-6 and table XLVIII). The former type is notably frequent in Tibullus: obvia nudato, Delia, curre pede (1.3.92); there are 21 instances in the two books (one third of them terminating, like the example here given, in the word pede). In pseudo-Tibullus there are seven instances (four with prece, which Tibullus does not use in this final position), constituting roughly the same proportion in relation to the total number of pentameters (Tibullus 3.4%, pseudo-Tibullus 3.0%); in Catullus six (1.9%), in Propertius 12 (0.6%, with two in the first and last books, four in the second and third), in the Fasti of Ovid 14 (0.6%, rather evenly distributed among the books, none having fewer than two or more than three). The second type is especially characteristic of the earlier Propertius: contactum nullis ante cupidinibus (1.1.2). In the first book the hemistichs of forty pentameters (11.3%) are marked by such agreements; in seven passages the pentameters occur in adjacent distiche (1.2-4, 3.22-4, 3.36-8,11.2-4,16.4-6,18.22-4, 20.36-8), and one passage has three in succession (16.42-6). In the second book, which is almost twice as long as the first, there are only nine instances (1.3%), two of these in adjacent distiche (34.68-70); in the third book only one (3.28). (The only verse-ending in -ibus that occurs in the fourth book belongs to the disyllabic tribus, which modifies and rhymes with ducibus at the end of the first hemistich: 10.2.) Catullus has six examples (1.9%), two of which are in adjacent distiche (66.58-60); Tibullus 15 (2.4%), pseudo-Tibullus two (0.8%). Agreement of this latter type is shown by the two pentameters of the Fasti (5.582 and 6.660) which, alone of 2485, end in words of more than two syllables.

page 43 note 79 The examples of rhyme apart from inflectional agreement given by Eichner and by Foster (see notes 67 and 69) are few and unimpressive; those given by Grimm, who fails to take account of differing vowel-quantities, include “eine ausserordentlich grosse Zahl notorisch falscher Beispiele” (Norden, AKP 2.829). According to Lachmann (who in a note on Prop. 1.5.20 gives as further examples 2.1.48 and 72, 4.7.24, 4.9.16, Cat. 66.38, Ov. H 4.162, 7.102, 8.24), Tibullus and Lygdamus have no such verses, “quales caeteri sibi permiserunt, similitudinem soni minus apparituram sperantes, sensus et interpunctionis beneficio” (Propertii Carmina 23).

page 43 note 80 Paradise Lost, Author's Preface on the Verse.

page 44 note 1 In considering these other cases, no distinctions of gender will be made, as there are no corresponding distinctions of form. However, some of the differences in the proportions given in the tables under MF 2s and N 2s (see preliminary note to chapter VII) may be explained by the fact that neuters tend to constitute a larger proportion of all ablatives in -ĕ than of genitives in -s preceded by a short vowel (in the pentameters of Ovid here examined 26.9%, against 20.2%). This might prove to be true of ablatives in general, as compared with genitives, in view of the somewhat greater likelihood that genitive constructions will have reference to persons, ablative constructions to things (it being remembered that while in Latin words designating things may be masculine and feminine, words designating persons may not be neuter); but time and space impose limits upon the extent to which such analysis can be carried.

page 44 note 2 There is a third specific exception : the slight preponderance in the second hemistich of masculine and feminine accusatives singular, peculiar to Tibullus. See ch. VI, n. 2.

page 44 note 3 See §32. The types are here so numbered as to preserve the analogy.

page 44 note 4 See ch. II, n. 9. In Catullus I have noted eight examples of ecthlipsis affecting declensional forms in the second hemistich of the pentameter (68.90, 73.6 bis, 76.4, 90.2, 91.4, 105.2, 107.2), in Propertius seven (2.24.40, 3.2.20, 3.7.12, 3.15.40, 3.22.12, 4.1.30, 4.7.48), in the Corpus Tibullianum and in Ovid none. (Verum, in the Epistula Sapphus 96, is in effect an adverbial conjunction; and Amores 1.13.32 an putat ignotam nequitiam esse suam is with its hexameter bracketed and asterisked by Ehwald, as not found by the first hand in the codices Puteanus and Sangallensis: see critical preface to his edition, p. ix.)

page 45 note 5 Hōc is here so marked because it is always scanned as a long syllable, even before vowels (e.g. Cat. 76.15, Hor. S 2.6.1); but see Kent, SL³ §185.II.D, FL §316.I, also Conington's note on Verg. A 2.664. For an account of the other words in this parenthesis, consult index to Kent, Forms of Latin.—The word ‘ending’, as used with reference to these monosyllables, is to be understood as defined in ch. II, n. 2.

page 45 note 6 Actually, in all these pentameters genu occurs only once (in Ov. F 3.722, where it is final), other neuters of the fourth declension {cornu, veru) not at all, in the nominative, accusative, or vocative singular.

page 45 note 7 For definition of this term, see ch. III, n. 16; for the percentages and proportions given in this paragraph, see italicized figures after N 1s in the tables of chapter VII.

page 46 note 8 In Catullus also—in so far as the small figures admit analysis—less than a third (6/19) come at the end of the line. But in Catullus the endings of a like proportion (6/19) are closed by following consonants; whereas in the other poets (except pseudo-Tibullus, who has only four such NAV neuter singulars in the second hemistich, one of them so treated, and two concluding the verse) this happens about once in ten times (Tibullus 3/33, Propertius 12/125, Ovid 83/809). Cf. §33 ad fin., with notes.

page 46 note 9 463 in 809 (57.2%). Propertius has 49 so treated, in 125 (39.2%). Tibullus, with 10 in 33, and Catullus, with 7 in 19, may be said to have one third, more or less. Pseudo-Tibullus has one in a total of four (cf. note 8).—For the totals in this and the preceding notes, see table XLV, under type II, column b. The prior figures given in this note, for forms the endings of which are actually short in the scansion, will be found in the same column, italicized.

page 46 note 10 In the pentameters of the five works examined, Ovid has 1014 NV singulars, masculine and feminine, of type I in the first hemistich, 3984 in the second; 64 NAV neuter singulars of the same type in the first hemistich, 177 in the second.—For these and analogous figures, see tables XL and XLV.

page 46 note 11 241, of 2307. The proportions for the other poets (except pseudo-Tibullus, who has only one in 32) are remarkably close: Catullus 12/115, Tibullus 8/76, Propertius 32/351.

page 46 note 12 Catullus has 141 NV singulars, masculine and feminine, of type I against 161 of types II and III; Tibullus 312, against 317; pseudo-Tibullus 139, against 126; Propertius 1309, against 1037; Ovid 4998, against 4837.

page 47 note 13 In the pentameters of Ovid here considered, of 2307 NAV neuter singulars 843 (36.5%) are of type III; of 9835 NV singulars, masculine and feminine, only 1570 (16.0%).—The figures given in this note, as in the two preceding, may be obtained by the addition of pertinent data in tables XL and XLV.

page 47 note 14 Ov. H 18.170. Catullus and Tibullus show no example of this pattern (the restrictions being here the same as those stated in ch. IV, n. 59). In Propertius I have noted six, all in the last two books (3.15.26, 3.18.32, 4.2.36, 4.6.32 and 40, 4.8.6), in Ovid 54 (A 1.8.14 and 64, 1.9.22, 1.10.40, 3.2.62, 3.8.24, 3.9.6, H 3.60 and 62, 6.128, 9.100, 10.116, 12.82, 13.6, S 116, F 1.202, 2.98. 2.228. 2.292, 2.346, 2.554, 3.18, 3.28, 3.230, 3.396. 3.746, 4.104, 4.460, 4.532, 4.734, 4.760, 4.912, 6.416, T 1.3.34, 1.6.36, 1.7.22, 1.8.34, 2.310, 2.472, 3.3.50 and 58, 3.4.36, 3.8.2, 3.10.64, 3.12.12, 4.2.64, 4.8.24, 5.5.16, 5.6.36, P 1.7.50, 1.10.22, 3.6.2, 4.5.40, 4.9.70), in addition to the line given, which would have been excluded from tabulation by reason of the fact that nihil is indeclinable (cf. §6), also because nihil and Mud are grammatically in separate clauses (cf. ch. IV, n. 59). It may be noted that in this line exiguum and plüs, of type III, are both found in the first hemistich, illud (and nihil), of type II, in the second. Cf. §§57 and 59.

page 47 note 15 Greek genitives in -ǒs (and -eǒs) have been included in the tabulations with Latin genitives in -is. Catullus has one such form, in the second hemistich (66.54), the Corpus Tibullianum none. Propertius has four in the first hemistich, all in book III (9.14, 11.44, 15.12, 17.24), six in the second hemistich (1.1.10, 1.3.20, 1.20.18, 2.31.14, 3.22.8, 4.7.72); Ovid only four that I have noted, two in the first hemistich (F 5.168, P 3.4.84) and two in the second (F 4.232, T 5.3.40). Cf. note 96 ad fin.

page 47 note 16 See §32.

page 47 note 17 Only in book III does Propertius have fewer such genitives in the second hemistich than in the first.

page 48 note 18 In Tibullus book II more than sixteen times, in pseudo-Tibullus and in Propertius book I roughly twenty.

page 48 note 19 See §59. On the substitution of -ĕ for -ī in the ablative, to gain a short syllable, see Walther Abel, Die Anredeformen bei den römischen Elegikern 55, n. 7 (with bibliography on the question whether there is a dative in -ĕ). All four instances of the ablative marĕ (cf. Lucr. 1.161, with Cyril Bailey's note ad loc.) cited for Ovid by Bednara, ALLG 14.342, occur in the body of the second hemistich of the pentameter (T 5.2.20, P 4.6.46, AA 3.94, Ibis 196 [198]); contrast marī at the end of the first hemistich (T 5.4.2) and at the end of the second (P 4.10.10). Of the adjectival examples cited by the same author, in ALLG 14.343 (Cat. 68.99, Ov. A 1.8.95, 1.9.18, 3.5.38, P 1.5.8,1.10.14, AA 3.359, RA 791), all but one occur either as the penultimate word of an hexameter, with the short termination falling in the off-beat of the fifth foot, or in the second hemistich of a pentameter. The one exception is rivalĕ (Ov. A 1.9.18 : cf. A1.8.95, RA 791), which occurs in the first hemistich of a pentameter, with the -ĕ elided, but is itself grammatically parallel to hoste immediately preceding the disyllabic ending of the second. To these we may add two instances of perennĕ (Ov. H 8.64 and F 3.654), the former in the second hemistich, the latter in the first, where it exactly balances the penultimate word of the second hemistich, Perenna (cf. note 24 ad fin.). To touch upon a more ambiguous matter, participial adjectives may be formed in -ĕ instead of -ī (as in H 10.140, 11.26, 14.44); conversely, Ovid uses stantī, metri gratia, in what is clearly an ablative absolute (F 5.571). Cf. Bednara, ibid.

page 48 note 20 See ch. II, n. 10.

page 48 note 21 1.1.59-60

page 48 note 22 A 3.9.58.

page 48 note 23 See §§17, 45–7, 60.

page 48 note 24 Tib. 1.1.48. For other examples see Tib. 1.5.16 and 22, 1.10.12 and 30, 3.4.32 and 40; Prop. 3.10.2, 3.17.32, 3.22.32. In Ovid I have noted the following: A 1.1.12, 1.4.60, 1.7.22, 2.14.22, 2.15.20, 2.19.40, 3.6.72, 3.7.6, 3.10.46, 3.13.24; H 3.88, 5.68 and 72, 6.96, 7.116, 8.64 and 96, 9.94, 11.26, 14.122, 15.168 and 284, 19.212, 20.88, S 140; F 1.100, 1.572, 2.80, 2.164, 2.364, 2.692, 2.786, 2.840, 3.236, 3.880, 4.302, 4.346, 5.184, 5.730, 6.338, 6.400; T 1.3.44, 3.3.68, 3.10.12, 3.11.54, 3.13.18, 4.2.50; P 2.3.88, 2.5.20, 3.1.154, 4.6.10, 4.8.22. For reasons already given (see §47, also §62 ad fin.), the same pattern hardly occurs in the first hemistich. In the one example I have found, the whole point of the line is in the echoing of sound from hemistich to hemistich: Amne perenne latens Anna Perenna vocor (Ov. F 3.654).

page 49 note 25 See §91, and table XLVIII.

page 49 note 26 3.9.59-60. Cf. also Ov. T 3.7.36 (quoted in ch. IV, n. 40 ad fin.).

page 49 note 27 Ov. F 2.684. For other examples, all from Ovid, see A 2.1.24, F 3.724, 4.860, 5.266, 6.12, T 1.1.4 (in all but the first of these the second word is huius).

page 49 note 28 H. 13.52.

page 49 note 29 See §95 ad fin., with notes.

page 49 note 30 There are noteworthy examples also of pairs of words in different cases, both ending in -ĕ: O penitus toto corde recepte mihi (Ov. H 18.156), or both ending in -ĭfitque novus nostri finis amoris amor (Ov. H 18.104; cf. A 3.11.20).

page 49 note 31 Kent, FL §§317.II, 318.II, 319.II, 321.I and III.

page 49 note 32 In the pentameters of Catullus I have noted four instances of -ĭus (68.44 and 146, 78.2 bis, at the beginning and at the end of the same pentameter), none of -īus; in Tibullus two of -ĭus (2.4.52, 2.6.36), one of -īus (1.8.66); in pseudo-Tibullus two of -ĭus (4.58, 6.44), none of -īus; in Propertius four of -ĭus (1.3.36, 1.5.30, 2.13.36, 3.6.10), one of -īus (4.5.58)— all of these (except the second instance in Catullus 78.2) in the first hemistich; in Ovid seventeen of -ĭus, of which ten are in the first hemistich (A 1.13.20, 2.7.10, 3.3.16, H 19.144 and 232, F 4.28, T 2.208, 3.3.2, 4.4.76, P 4.6.38), seven in the second (A 1.4.34, H 20.40, F 1.40, T 5.2.22, 5.6.34, P 1.1.36, 4.6.6), and nine of -īus (A 1.13.44, 1.14.14, H 11.60 and 62, 15.312, F 2.454, 5.222, T 2.112 and 564), all in the first hemistich.

page 50 note 33 See ch. IV, n. 60.

page 50 note 34 Catullus 68.44 and 146, Tibullus 2.4.52, 2.6.36, Propertius 2.13.36, 3.6.10—all at the beginning of the first hemistich; Ovid A 1.13.20 and P 4.6.38 at the beginning of the first, A 1.4.34, H 20.40, F 1.40, T 5.2.22, 5.6.34, P 1.1.36, at the beginning of the second. All the other references to forms in -ĭus in note 32 are to instances of alterĭus.

page 50 note 35 Kent, FL 314.III, 316.III, 320.I.B.

page 50 note 36 There are two examples in Catullus (91.6, 102.2), both in the first hemistich before consonants; one in pseudo-Tibullus (18.4), similarly treated; four in Propertius (2.5.26, 2.6.2, 2.14.6, 4.11.102), all in the first hemistich, and all except the last before a consonant. Ovid has seven examples in the first hemistich, two before consonants (F 1.194, P 1.7.14), five before vowels (H 2.74, F 4.492, T 2.158, 4.2.16, P 2.7.36); and he alone has one in the second hemistich (H 12.88), where it must of course be followed by a vowel.

page 50 note 37 Huius occurs not at all in the pentameters of Catullus and of the Corpus Tibullianum, and only three times in Propertius : in the first and second hemistichs of the same pentameter (2.15.36), and again in the first hemistich (4.10.6), always before a vowel. Ovid has twenty examples: four in the first hemistich followed by a consonant (P 1.2.102, 1.5.32, 1.7.8, 2.8.14), two in the first hemistich followed by a vowel (A 3.8.12, F 2.180), fourteen in the second hemistich (H 15.326, F 1.554, 3.180, 3.630, 3.724, 4.860, 5.266, 6.12, T 1.1.4, 1.5.10, 3.1.48, 3.12.40, 5.1.36, P 4.4.22), all except the last immediately preceding the disyllabic verse-ending, and all of course followed by a vowel (cf. note 27 ad fin.). It is noteworthy that all four examples of huius followed by an initial consonant occur in the Epistulae ex Ponto: in 1.2.102 and 1.5.32 immediately before the diaeresis, rather awkwardly.

page 50 note 38 Eius 4.15.6 (like huius 1.2.102 and 1.5.32, immediately before the diaeresis); eiusdem 2.9.64. Neither form occurs in Vergil.

page 50 note 39 The figures for the two hemistichs respectively are: Catullus 5/0, Tibullus 17/16, pseudo-Tibullus 6/6, Propertius 54/28, Ovid 216/204. In Ovid, however, the proportion of these pyrrhic forms in the second hemistich rises steadily from work to work: Amores 34.6%, Heroides 43.5%, Fasti 49.1%, Tristia 54.7%, Epistulae ex Ponto 60.0%. The proportion for the five works together is 48.6%.—The figures cited above, and those upon which these proportions are based, are the same as those given in italics under type III in table XLVI, with the exceptions there noted.

page 51 note 40 See §91, and table XLVIII.

page 51 note 41 To confine attention for the moment to the more voluminous Propertius and Ovid : of 47 instances of mihi, tibi, and sibi in the second hemistich of Propertius' pentameters, one-third (16, including one with enclisis of est) are final, more than four-sevenths (28— cf. note 39) occur as dibrachs in the first dactyl, while three have elided endings (2.20.16, 3.8.40, 4.7.74); of 354 instances in Ovid, over two-fifths (146, including eleven with enclisis of est) are final, less than three-fifths (204—roughly the same proportion as in Propertius) dibrachs in the first dactyl, while only one has its ending elided (T 5.9.12), and three occur as iambs in the body of the hemistich: two before -que (T 4.10.28, P 2.11.12—see §87 ad fin.), one before a trisyllabic verse-ending, for which the poet seems to apologize (P 3.6.46: vix excusari posse mihi videor). Tibullus has only three examples (one each: 1.3.90, 1.8.8, 2.4.24) of mihi, tibi, sibi final in 19 instances of these words in the second hemistich, pseudo-Tibullus three (10.16 and 22, 11.4—the first with enclisis of est) in nine; in neither is the final syllable elided in any part of the pentameter. In Catullus, on the other hand, all seven instances are final (68.20, 92 and 160—this last with enclisis of est, 76.6 and 20, 91.8, 101.6). Catullus is peculiar in that he alone uses the monosyllabic mī in the second hemistich (107.4); he has it twice in the first (72.6, in the form m, and 76.26, elided), Propertius three times (2.18.30, 2.22.18, 4.1.62), Tibullus and Ovid not at all. With regard to Tibullus in particular, it is interesting to note that in his pentameters, if we except the three final occurrences mentioned above, mihi, tibi, and sibi are always dibrachs; pseudo-Tibullus, in addition to the three final (or quasi-final) instances already cited, has tibī (11.14) and sibī (1.26) in the first hemistich, the former immediately before the diaeresis.

page 51 note 42 In Catullus these forms occur in the iambic rhythm seven times as verse-endings, twice in the first hemistich; in Tibullus three times as verse-endings, not at all in the first hemistich; in pseudo-Tibullus three times, against two; in Propertius 16 times, against 19. In Ovid they occur twice as frequently in the final position as in the first hemistich (146 against 73). Ovid alone has non-final examples in the second hemistich (see note 41).

page 51 note 43 Pseudo-Tibullus 19.11-2.

page 52 note 44 See ch. VI, n. 42.

page 52 note 45 See ch. IV, nn. 60-1.

page 52 note 46 Catullus has one of these dibrach forms in the second dactyl of the first hemistich, out of five (104.2), Propertius four, out of 54 (2.18.34, 2.32.8, 2.34.58, 4.5.58), Ovid four, out of 216 (F 1.146, T 3.12.52, 4.6.26, P 3.6.12), Tibullus and pseudo-Tibullus none, out of seventeen and six respectively.

page 52 note 47 See table XX.

page 52 note 48 For examples of mihi, tibi in both hemistichs of the same pentameter, see Tib. 1.4.4, 2.6.10, Prop. 2.16.46, 2.28.14, Ov. A 3.6.92, H 4.148, 11.114, 12.106, 13.104 and 166, 20.204, F 4.8, 6.214, T 5.1.66, 5.2.56, 5.6.2. Ovid A 3.6.90, 92, 94 begins three successive pentameters with si tibi, nec tibi, quas tibi; the second hemistich of 92 repeats nec tibi. Sometimes a pyrrhic form of the possessive adjective in the second hemistich balances mihi, tibi in the first (Prop. 4.9.18, Ov. A 2.9.36, 3.2.48, H 15.354, F 2.308, 3.170, T 2.148, 2.552, 3.3.52, 5.1.66, P 4.6.4), vice versa almost as often (Tib. 2.5.122, Prop. 2.24.36, Ov. H 5.6, 6.60, 12.2 and 202, 17.128, P 4.8.2). Ovid allows mihi, tibi, sibi in this position to come between prepositions and their case (H 6.20, 17.76, 18.148, 19.146, P 2.7.18, all in the first hemistich; P 1.2.150, in the second). Cf. §43 ad fin.

page 52 note 49 See note 39.

page 52 note 50 The figures for the two hemistichs are: Catullus 7/7, Tibullus 17/19 (I 11/12, II 6/7), pseudo-Tibullus 8/9, Propertius 75/47 (I 22/4, II 30/25, III 14/14, IV 9/4), Ovid 293/354 (A 40/32, H 112/105, F 32/46, T 59/90, P 50/81). A tendency similar to Ovid's may be noted in Propertius, if we allow chance to have played its usual part in the smaller totals of book IV.

page 53 note 51 See §91, and table XLVIII.

page 53 note 52 In the pentameters of Propertius rebus occurs once (2.34.16), bubus once (4.1.8); in those of Ovid, rebus twelve times (H 1.94, 4.114, 16.130, 18.2, F 2.734, T 2.216, 2.434, 3.11.70, 5.14.32, P 3.4.22, 4.5.22, 4.8.26), including three in the first hemistich (T 2.434, 3.11.70, P 3.4.22), bubus twice (H 5.116, F 1.244), duobus and duabus ten times (A 1 Epigramma 4, 1.10.2, H 8.90, 13.142, 19.234 and 238, F 1.258, 2.294, 2.470, 6.496), diebus once (F 6.144). All of these, except the three instances of rebus noted, occur in the second hemistich. Ambōbus is of course restricted to the first hemistich (Cat. 104.2, Ov. A 2.9.54, T 5.6.30, P 2.5.68, 2.7.64).

page 53 note 53 For examples of quibus in this position, see Ov. H 19.104, T 4.7.6, P 3.4.72, 4.6.2; for examples of tribus in the same position, Prop. 2.13.44, Ov. H 9.92, T 2.246, 4.10.34.

page 53 note 54 For examples, in the first hemistich, of quibus as dibrach, see Cat. 67.36, Tib. 1.7.24, Prop. 3.9.18, Ov. A 1.12.26, H 15.362, F 1.136, 3.528, 4.422, 4.524, T 2.430, P 1.7.16, 2.1.46, 4.14.22; for examples of tribus so used, Ov. F 3.94, 4.614, 6.288, T 4.10.10. For quibus in iambic rhythm, in the first hemistich, see Prop. 1.4.12, 3.23.2, Ov. F 4.274, T 1.5.68, 1.6.16, P 1.1.74; for tribus similarly treated, Ov. H 3.48, T 3.12.24. The form quīs is always found in the first hemistich of the pentameter (Prop. 1.8.42, 2.34.88, Ov. H 5.96, F 2.22, 2.200, 4.134, T 5.5.44); cf. ch. III, n. 22.

page 53 note 55 As in Prop. 1.13.30, 4.10.2, Ov. H 3.38, 10.118, P 2.8.8 and 56, 4.3.26.

page 53 note 56 The only examples I have noted are of quibus, in Ovid: A 2.8.28, T 3.12.34, P 4.7.46. See §88, ad fin.

page 53 note 57 The term ‘polysyllable’ is used here, as in table XLVIII, to designate words of three or more syllables.

page 53 note 58 44, including four pentasyllables (1.2, 5.24, 20.38 and 42). We might add three rhythmically analogous Greek datives plural, all in one poem (20.12, 32, 34—see §78).

page 53 note 59 See §91, and table XLVIII. In the first book of Tibullus nine of 27 polysyllabic verse-endings are anapaests or choriambs in -ibus, in the second book seven of 18. Pseudo-Tibulius has only two in 13 (including one pentasyllable: 4.56).

page 53 note 60 Including two pentasyllables (16.14 and 34.68).

page 54 note 61 10.2. Cf. note 55.

page 54 note 62 196 in 320 (61.3%), as compared with 214 in 2003 (10.7%). Cf. table XLVIII.

page 54 note 63 Of these six verse-endings in -ibus, five occur in one poem (66.4, 58, 60, 80, 92), one in the poem preceding (65.2).

page 54 note 64 5.582 and 6.660.

page 54 note 65 See §§87-90.

page 54 note 66 66.90. Cf. note 63.

page 54 note 67 Tib. 2.3.80, 2.5.2 and 100, 3.19.24; Ov. A 2.13.26, 3.12.22, PI 2.50, 14.96, 15.54 and 130, F 1.552, 1.570, 1.628, 1.672, 2.556, 3.132, 3.184, 3.750, 3.776, 4.96, 4.200, 5.50, T 1.2.104, 1.8.16, 2.370, 2.434, 3.11.16, 4.1.58, 4.10.64, P 2.2.84 and 110. Note the double occurrence in Ov. F 2.556; cf. §47 ad fin.

page 54 note 68 The one non-final example in Catullus (66.90, cited above) and the one in pseudo-Tibullus (19.24) occur, to be sure, in the second hemistich; and in the second book of Tibullus more of the very few occurrences of these forms are found in the body of the second hemistich than anywhere in the first (3, against 2). But in the first, and longer, book of Tibullus the figures are 5/0, for the two hemistichs respectively; in Propertius: book I 0/0, book II 3/0, book III 4/0, book IV 5/0; in Ovid: A 12/2, H 15/4, F 21/12, T 20/7, P 14/2.

page 54 note 69 See §68.

page 54 note 70 T 1.1.34.

page 55 note 71 Tuse. Disp. 1.42,101. I have noted only one pentameter in which both hemistichs begin with such a form (unless we allow bis tribus and qui tribus respectively to balance mensibus in F 4.614 and T 4.10.10) :

page 55 note 72 The figures for the two hemistichs respectively are: Catullus 2/5, Tibullus 3/4 (I 3/2, II 0/2), pseudo-Tibullus 0/2, Propertius 12/24 (I 2/10, II 2/3, III 6/5, IV 2/6), for Ovid 88/131 (A 9/11, H 11/26, F 27/39, T 24/31, P 17/24).

page 55 note 73 Greek forms have everywhere been included in the tables without distinction from the Latin.

page 55 note 74 The MS tradition is garbled, but seems (partly by reason of that very fact) to point to a Greek form in both instances. See Ellis's critical note ad loc.

page 55 note 75 Cicero in fact uses the genitive Arsinoae (Nat. Deor. 3.22,57). In the preceding paragraph (56) he has a genitive Phoronidis.

page 55 note 76 E.g. Pēnelopē (Prop. 3.13.24), Ithacē (Ov. T 1.5.67).

page 55 note 77 E.g. Prop. 2.26.16, 2.28.52, 3.13.24. The homoeoteleuton with the Greek choriambs ending these pentameters was at least welcomed, if not sought (cf. §52 ad fin.). Ovid, having ended an hexameter with Ascrē (P 4.14.31), in the second hemistich of the pentameter three lines lower uses Ascră–-the first form as well as the second metri gratia, according to Hilberg, who posits a law that in the hexameter, as in the pentameter (cf. ch. III, n. 26), short-vowel verse-terminations are where possible avoided (Die Gesetze der Wortstellung im Pentameter des Ovid, footnote p. 846).

page 56 note 78 66.48.

page 56 note 79 3.12.38.

page 56 note 80 T 1.9.28. See also Prop. 2.1.38, 3.15.12, 4.6.84, Ov. A 2.1.12, 2.11.10, H 7.26, 9.6, F 2.318, 3.582, 4.76, T 4.3.64, 5.1.60, 5.4.26, P 1.5.80. In all of these examples the Greek form is followed by an initial vowel; in all except Ov. H 7.26 it immediately precedes the diaeresis. For a succession of Greek—and Latin—accusatives variously placed before both consonants and vowels, see Ov. F 4.172-4, in Frazer's text (Ehwald-Levy gives Greek forms throughout this passage).

page 56 note 81 2.5.21-2, the only example in Tibullus. Cf. Ov. H 16.240: Ilion arsurum . . . For other examples in the first hemistich, see Prop. 2.34.68, 3.13.8, 4.1.78, Ov. A 2.17.4, F 3.658, 4.174, T 1.10.28 and 30; in the second hemistich, Prop. 2.1.20, 2.34.68, Ov. A 2.1.14, H 18.30, F 2.192, 4.440, P 3,2.62. Note the double occurrence in Prop. 2.34.68.

page 56 note 82 For examples in the first hemistich, see Prop. 2.24.44, 3.1.28, Ov. H 9.154, 10.10, S 172, F 1.306, 2.82, 2.606, 3.460, T 2.402, 4.3.30, 5.3.22, P 3.3.42, 4.8.60; in the second hemistich, Prop. 2.1.62, 2.8.38, 2.9.12, 3.9.48, Ov. A 2.16.8, 3.2.54, H 2.98, 5.94, 10.34 and 110, 13.94, 18.176, F 2.42, 3.286, 3.416, 3.816, 3.848, 4.678, 5.160, 6.424, 6.430, T 2.36, 3.4.30, 4.5.30, 4.6.14, 5.2.26 and 74, 5.5.32, P 4.14.12. Propertius elides the final a of paeāna, in the first hemistich (3.15.42).

page 56 note 83 For examples in the first hemistich, see Ov. H 3.36, S 200, F 2.272, 2.290, 3.444, 4.210, 6.504, 6.514, T 1.11.8, 3.1.62; in the second hemistich, Tib. 3.8.6, Prop. 3.17.34, Ov. A 1.9.34, H 10.86, 12.138, 13.68, 15.350, F 4.210, 4.458. Note the double occurrence in Ov. F 4.210.

page 56 note 84 As in Ov. H 15.338. When the rhythm of its stem requires that a Greek nominative or vocative plural ending in -es be used before a consonant, the quantity of the e is of course obscured, so that without further evidence one cannot determine the extent to which Latinization of the ending has progressed. The fact that Ovid regularly puts the accusative of such words in -ăs, and that he never allows the -es to receive an ictus unless followed by a consonant, makes it seem probable that he still felt the ending to be Greek, and the vowel short.

page 57 note 85 As in Prop. 2.1.14.

page 57 note 86 As in Prop. 2.1.50. On the practice with regard to such short-vowel verse-terminations for the pentameter, see ch. III, n. 26.

page 57 note 87 Tēthyi, in the second hemistich (66.70). Cf. Tēthyos, Ov. F 5.168.

page 57 note 88 In 1.20: Ādryasin (12), Hamādryasin (32), Thȳyniasin (34). Ovid has hērōisin (T 5.5.43) and Trōasin (H 13.137), in hexameters.

page 57 note 89 E.g. in Catullus 66 (the so-called ‘Coma Berenices’, translated from Callimachus) we find Athōn (46), Chalybōn (48), Arsinoēs Locridos (54), Callistō (acc, 66), Boōtēn (67), Tēthyi (70), Hȳdrochoī (dat. , 94), in addition to the transliterated nominatives Conōn (7), Zephyrītis (57), and Ōariōn (94). Cf. note 88, for an example from Propertius.

page 57 note 90 Cilicas (1.7.16), Memphītēn (ibid. 28), and Ilion (2.5.22) about exhaust the list, at least of oblique cases used in the pentameters. Pseudo-Tibullus has lampadas (8.6).

page 57 note 91 Cf. notes 15, 76-7, 80-4, 87-8, 94.

page 57 note 92 See Kent, FL §262.

page 57 note 93 3.11.34 and 3.13.10.

page 57 note 94 For examples of such vocatives in the first hemistich, see H 2.98, F 5.610, 6.714, P 2.9.38; in the second hemistich, A 2.8.2 and 22, 2.18.22, H 2.138, 6.48, 9.50, 10.100, 11.34, 15.100, 15.128, 15.308, 16,254, 17.2, S 100, F 1.454, 1.634, 3.524, 4.468, 4.572, 6.238, P 4.10.52. Cf. §§35-8.

page 57 note 95 In Catullus Tēthyi (66.70), in Tibullus Īlion (2.5.22), in pseudo-Tibullus lampadas (8.6).

page 57 note 96 The distribution between the hemistichs is as follows, for the several books of Propertius: I 0/0, II 2/5, III 4/4, IV 1/0; for the several works of Ovid: Amores 1/7, Heroides 7/22, Fasti 14/19, Tristia 7/8, Epistulae ex Ponto 3/3. These figures comprise the instances cited for Propertius and for Ovid in notes 81-3 and 93-4, with the addition of three genitives in -eǒs, which Latinized would have become -eī (Prop. 3.15.12 and 3.17.24, both in the first hemistich, Ov. T 5.3.40, in the second). The other Greek genitives listed in note 15 do not differ metrically from their Latin counterparts in -is.

page 58 note 97 Cat. 68.112, in the second hemistich. Conversely, the spondaic Latin adpellandorum (Ov. P 4.16.40) constitutes the whole of a first hemistich. Cf. §§2-3, with notes.

page 58 note 98 Prop. 1.3.4.

page 58 note 99 Ov. T 4.9.16 et alibi.

page 58 note 100 Tib. 1.7.16.

page 58 note 101 See §91, and Table XLVIII.

page 58 note 102 E.g. the one example of Īlion in Tibullus, cited in §77. In the pentameter of Ovid: Pollucem pugiles, Castora placet eques (A 3.2.54), note the spondaic character of the Latin Pollücem, in contrast with the dactylic Greek Castǒră. Cf. Horn. II. 3.237: where the last two of four short syllables in must coalesce by synizesis at the end of the hexameter.

page 58 note 103 As these categories are defined in chapter II, §6 and note 1.

page 58 note 104 See §§18-20.

page 58 note 105 In Ovid the proportions for the pentameter as a whole from work to work describe an arc which, shallow as it is, is not inconsistent with what we know from other criteria of the waxing and waning of his skill: Amores 26.5%, Heroides 26.7%, Fasti 27.2%, Tristia 25.9%, Epistulae ex Ponto 25.1%. Cf. §§15-6.

page 59 note 106 The greatest difference between the two hemistichs in this regard is found in pseudo-Tibullus, where the proportion of short syllables derived from declensional inflections is almost two and a half times as large in the second hemistich as in the first (30.9% against 12.5%). Cf. §31.—It must be remembered that the number of metrically significant short syllables in the second hemistich is fixed at four, whereas in the first it may vary among four, two, and none. The total sum of metrically significant short syllables in twenty-five second hemistichs must therefore always be 100; in the same number of first hemistichs it may range widely, from an average of 36.3, as in Catullus, to an average of 62.1, as in the first fourteen of the Heroides (see table XX for other figures). It is evident that these Heroides, in the first hemistichs of which only 15.6% of all short syllables are declensional inflections, will nevertheless have a larger number of such inflections providing short syllables in a given number of first hemistichs than will be found in a like number of first hemistichs in Catullus; in these latter, 18.0% of all short syllables are declensional inflections, but there are less than three-fifths as many short syllables. Comparisons among the proportions given for the second hemistich, with its constant number of metrically significant short syllables, remain uncomplicated by such considerations, which in any case do not impair the argument of §81.

page 59 note 107 For a criticism, from a formal standpoint, of these extremes, and of Tibullus and Propertius as intermediate between them, see J. P. Postgate, Select Elegies of Propertius cxxvi.

page 59 note 108 See §91, and table XLVIII.

page 59 note 109 Actually, a short independent monosyllable is not found here. A short monosyllabic preposition, being proclitic to its case, would give in this position a trisyllabic verse-ending. I have noted no example. Rarely a conjunction is so used before the final disyllable: nis(i) amat, nis (i) amo (Cat. 92.2 and 4), nec ego (Mart. 4.72.4), but as these are likewise proclitics, such a verse-ending is not really disyllabic. (It is interesting to note that Martial's nec ego concludes the final pentameter of the epigram, with special effect. Cf. ch. VI, §91 and nn. 31 and 42.) Ovid uses sit twice in this position, in late or dubious works (H 19.62, P 1.1.14), both times after an emphatic monosyllable to which it may be regarded as enclitic. These examples—none of which constitutes a real exception to the general statement— exhaust the list, to my observation.

page 60 note 110 See §§87-90, and table XLVII.

page 60 note 111 The proportion of declensional inflections among the short syllables of the second hemistich is actually lower in the Amores than in the fourth book of Propertius (32.6% against 34.5%).

page 60 note 112 On the relative infrequency of the vocative, see ch. III, n. 1, also §35.

page 60 note 113 A 3.9.35-40.

page 60 note 114 T 4.10.25-6.

page 61 note 1 The dibraehs mĭhĭ, tĭbĭ, and sĭbĭ, which, where they are not as a group about equally divided between the hemistichs—as in the Corpus Tibullianum and the Fasti (and in the work of Ovid as a whole)—are found more frequently in the first hemistich than in the second, in all but the last two works of Ovid. See §§66-8, with notes.

page 61 note 2 The masculine and feminine accusative singular, in Tibullus alone. In the pentameters of his work as a whole 54.9% of all these forms are found in the second hemistich (the proportions for the two books severally being 54.0% and 57.1%). If we except the last seven of the Heroides, which have only 32.0% of the forms in question in the second hemistich, pseudo-Tibullus there has the smallest proportion: 35.1%. A partial explanation of this difference between Tibullus and the other elegists may be found in the frequency with which he uses disyllabic accusatives singular, masculine andj feminine, at the end of the verse— the only point in the second hemistich where a long termination is of advantage. In his pentameters, 61 of 142 accusatives singular, masculine and feminine (43.0%), are found in this final position (book I 40/100, book II 21/42). Pseudo-Tibullus, on the other hand, has only 15 in 57 (26.3%), including two trisyllables (6.10 and 48). The last seven of the Heroides have an even smaller percentage so placed: 47 in 219, or 21.5%.

page 61 note 3 Also manŭs and (less acceptably) opĕ. On pentameters ending in a short open vowel, see ch. III, n. 26.

page 61 note 4 See §72.

page 62 note 5 The mora between the second ictus and the final disyllable may not be occupied by a short proclitic preposition, which would make the verse-ending really trisyllabic. Very rarely we find a short proclitic conjunction in this position, in authors who are not strict in their observance of the disyllabic law (Cat. 92.2 and 4, Mart. 4.72.4), and equally seldom, in late or dubious works of Ovid, a short monosyllable like sit (H 19.62, P 1.1.14), which in the phrase-accentuation is enclitic to the preceding word. Cf. ch. V, n. 109.

page 62 note 6 -que: 66.40, 73.4, 76.20, 109.2; -ve: 76.16. (It has seemed better to regard -ve, even in a word like sīve, as a separable enclitic, under the definition given in the footnote to table XLVII. In any case the instances here involved, as the same footnote indicates, are very few.) Dictaque factaque sunt (Cat. 76.8), interesting as it is in that it contains a unique example of -que before a monosyllabic verse-ending, is not a case in point.

page 62 note 7 The other works of Ovid here examined, especially the Amores, fall far below these percentages. Tibullus shows a great increase in his second book. See table XLVII, with note.

page 62 note 8 There are no examples in the first and last books, only four in book II and five in book III (-que: 2.8.14, 2.20.12, 2.27.4, 3.7.6 and 66, 3.11.28, 3.22.10; -ve: 2.9.28 and 3.13.46, on the latter of which see parenthesis in note 6) : in all, nine in 1784 pentameters ending in disyllables, or 0.5%.

page 62 note 9 2.8.14, 3.7.6 and 66, 3.11.28.

page 62 note 10 1.9.72, 2.6.46, 3.3.26.

page 62 note 11 P 2.3.10. Cf. note 14, for examples from the Fasti. H 15.96 has the prior -que elided.

page 62 note 12 F 2.782 (Ehwald-Levy here reads videris! audentes forsque deusque iuvat). See also Tib. 3.11.20.

page 62 note 13 F 5.110. See also F 6.274, T 3.12.34, 4.10.58, 5.4.36, P 4.9.96.

page 62 note 14 21 of -que: 1.18, 1.486, 2.234, 2.732, 2.758, 2.780, 2.822, 2.836, 3.86, 3.334, 3.360, 3.362, 3.428, 3.656, 4.304, 5.438, 5.584, 5.652, 6.308, 6.334, 6.678; two of -ve: 5.110 and 6.274; one of -ne: 2.782 (in Merkel's text; but cf. note 12). Note that in Fasti III such a correlative pair is found in the second hemistichs of two successive pentameters (360 and 362).

page 63 note 15 AA 2.24. This line, and the one quoted below from the Amores (2.11.10), are the only two preserved to our knowledge of the three that Ovid—according to the story told by the elder Seneca (Contr. 2.2.12)—would not have expunged from his work, of all others; precisely the same three had been simultaneously chosen by his friends for expunction. Perhaps the third was the line: tergeminumque virum tergeminumque canem (T 4.7.16). On first hemistichs ending in an iambic word, see §47, with notes.

page 63 note 16 S 54. See also H 4.112, and Tib. 2.5.100.

page 63 note 17 A 2.11.10.

page 63 note 18 P 4.7.45-6.

page 63 note 19 Ov. T 4.1.33-4. See also Tib. 1.3.38, 2.5.72, Ov. A 3.7.10, 3.10.12, 3.13.30, 3.14.12, H 16.84, 17.214, 20.200, F 1.44, 2.756, T 2.532, 5.10.40, P 1.4.54, 1.8.56, 3.6.48, 3.9.18. For examples of less extreme postponement in Tibullus, see 1.1.40, 1.6.54 and 72, 1.7.62, 1.10.54, 2.3.12, 2.5.22, 86 and 90, 2.6.16 and 52; in Ovid they are too numerous to cite. Pseudo-Tibullus has a kind of involvement, rather than postponement, in the line: attuili in nostrum deposuitque sinum (13.4); per Geniumque (11.8) is hardly abnormal.—The only examples of -que postponed that I have noted in Propertius put the enclitic in the first dactyl of the second hemistich: 2.32.14 and 3.21.16. These happen to be in the two books which alone of the four show -que before the final disyllable (cf. note 8).

page 63 note 20 1.3.55-6. Ovid has -que still further in advance of the word it should normally follow: et tauri mixtaque forma viri (H 2.70). It will be noted that in this example also the displacement of -que, however it may be interpreted grammatically, has the effect of freeing, so to speak, an iambic word for use at the end of the pentameter.

page 64 note 21 In the preceding chapters inflected forms compounded with -que inseparable were not considered apart from the corresponding forms uncompounded, because those that are accented like uterque, utrumque, and utrāque (as compared with utrăque), owe their utility in the second hemistich not to their inflectional syllable, which is long, but to their enclitic element. See ch. II, n. 2 ad fin.

page 64 note 22 Catullus has, besides utrumque (71.4), three instances of atque in this position (68.48, 77.2, 101.10—this last his atque in perpetuum, frater, ave atque vale), one of usque (116.4); the Corpus Tibullianum eight instances of inflectional forms compounded with -que inseparable (1.2.96, 1.4.38, 1.6.86, 1.10.58, 2.1.84, 3.5.18, 3.10.24 and 26), one of atque (3.4.72), one of -cumque (2.6.54), and eight of usque, which occurs with notable frequency in Tibullus (1.2.88, 1.3.16, 1.5.74, 1.6.8, 1.8.36, 1.9.38, 2.4.14, 2.5.32); Ovid's Fasti thirty examples that I have noted of inflectional forms compounded with -que inseparable (1.226, 2.136, 2.188, 2.244, 2.306, 2.530, 2.816, 3.96, 3.272, 3.292, 3.402, 3.526, 3.774, 4.66, 4.98, 4.484, 4.486, 4.654, 4.864, 5.32, 5.48, 5.240, 5.366, 5.478, 5.510, 5.574, 5.702, 5.704, 5.720, 6.548), one of ubique (1.218), two of -cumque (3.572, 4.516), three of usque (3.686, 3.872, 5.230), one of adusque (4.30). The -ne of anne (Ov. F 6.28) has been included in the figures given for separable enclitics (see footnote to table XLVII).

page 64 note 23 21 inflectional forms compounded with -que inseparable (1.3.14, 1.4.16, 1.13.20, 2.7.2, 2.8.26, 2.12.10, 2.17.4 and 12, 2.20.16, 2.24.44, 2.25.42, 2.28.58, 2.34.4, 3.3.36, 3.9.20, 3.15.24, 3.17.12, 3.21.22, 4.8.36, 4.10.36, 4.11.46), in addition to ubique (1.4.24, 3.22.18), utrimque (2.27.8) and usque (4.5.48).

page 64 note 24 Nor apparently of an analogous ‘paroxytonic law’, to be inferred from a statement of Wilamowitz-Moellendorff (Griechische Verskunst 52) regarding the accentuation of the verse-ending: “Wenn Tibull und Ovid den Pentameter so bauen, dass die vorletzte Silbe betont wird, tun sie das nach dem Vorgange gleichzeitiger griechischer Epigrammatiker.” (This observation, according to the footnote on the same page, seems to have been suggested to him personally by Paul Maas: “Darauf hat mich einmal P. Maas hingewiesen, und ich habe es zu meiner Überraschung bestätigt gefunden.”) When the last syllable of the pentameter contains a long vowel, and does not itself receive the accent, it is true that Greek words ending the pentameter will regularly be accented on their short penult, which can happen in Latin only if the word in question is a disyllable. Of eleven pentameters in the one long fragment that has been preserved of Callinus, the earliest of Greek elegiste, all but the first and last end in words so accented (see Edmonds, Elegy and Iambus 1.44-6). On the other hand, Meleager, who flourished about 100 B.C., ends pentameters with and for example, indifferently (AG 4.1.58, 5.191.6); in the long poem which served as an introduction to his ‘Garland’ (4.1) only seven of 29 pentameters end in paroxytones. Crinagoras of Mytilene not only lived in the time of Augustus, but spent several years in Rome, as an intimate of the imperial court; yet so far was he from setting such an example for his Roman contemporaries as Wilamowitz' statement would lead us to expect, that of the 49 elegiac epigrams that have come down to us under his name in the Greek Anthology (varying in length from two to four couplets, except 9.513, which consists of only one) seven of three couplets each (6.242, 345; 7.376, 638; 9.81, 516; 16.61), and two more of four couplets each (9.562, 16.273), have only one pentameter ending in a paroxytone; in three epigrams of three couplets each (7.636, 9.283, 11.42), and in one of two couplets (9.542), all the pentameters end in words of other accentuation. In only 14 of the 48 epigrams that consist of two or more couplets are the pentameter-endings paroxytone throughout. Vollmer comments on this explanation of Maas and Wilamowitz: “m. E. nicht richtig” (Römische Metrik, in Gercke-Norden EAW³ 1.8.14).—On the subject of the Greek epigrams at Rome, see James Hutton, The Greek Anthology in Italy to the Year 1800 (Cornell Studies in English XXIII) 10-29.

page 65 note 25 68.112. Cf. ch. V, n. 97.

page 65 note 26 The sole example that I have noted of a Latin pentameter ending in a word of six syllables occurs in Martial (inimicitiae 5.50.2)—if we except the outright Greek of the same poet, who once uses the Homeric tag (1.45.2) and again the expression (14.201.2). It would seem that only in Greek could such long sequences of conveniently arranged syllabic quantities readily be found. (Ibis 508 ends in the hexasyllable Berecyntiades; but this word is also Greek, and in any case the authenticity of the line is questioned by Merkel. Cf. table XLVIII, n. 1.)

page 65 note 27 Cat. 96.4.

page 65 note 28 Prop. 1.22.4.

page 65 note 29 Tib. 2.6.34.

page 65 note 30 About one in twenty. For the statistics on these verse-endings, see table XLVIII.— It is interesting to note that Ovid's last work, the Epistulae ex Ponto, although it is somewhat shorter than the Tristia, has almost twice as many pentameters ending in non-disyllabic words (30, against 16), and a fourth again as many as all his other works together, including the Tristia (see table XLVIII, with note 1). Moreover, 16 of these 30 non-disyllabic verse-endings in the Epistulae ex Ponto occur in the very last book.

page 65 note 31 The proportion of non-disyllabic verse-endings in the work of Martial as a whole is 11.5% (381/3311). In the Liber Spectaculorum and books I-XII the proportion is 12.2% (360/2956); in the same books, the proportion of non-disyllabic verse-endings in non-final pentameters is 8.0% (165/2060), in final pentameters 21.8% (195/896). (These figures include the almost negligibly few monosyllabic verse-endings discussed in note 42.) The comparison cannot be made for the Xenia and Apophoreta of books XIII and XIV, as these consist almost entirely of single couplets; but even here, among the pentameters of the five longer poems 1-3 in book XIII and 1-2 in book XIV, the one that ends in a non-disyllabic word is itself the last line of the epigram (13.3.8). Something of the effect may be felt in Tib. 1.4.84: cum mea ridebunt vana magisteria (there is among the preceding pentameters of this rather long poem only one non-disyllabic verse-ending : Pieridas, in line 62, a Greek word, and as such subject to freer treatment), and again in Ovid: praemia si studio consequar ista, sat est (T 5.7.68), although the last two words may be regarded as constituting a virtually disyllabic ending (cf. note 42 init.); but these are sporadic examples, from which no inference can be drawn.—For another example of Martial's practice, see ch. V, n. 109.

page 66 note 32 Vollmer's explanation amounts to the same: “Sehr auffallend ist, dass der Pentameter bei Tibull wie bei Ovid und einem Teil ihrer Nachahmer ganz überwiegend auf iambisches Wort ausgeht (natürlich ist dann der gleiche Ausgang im ersten Kolon verpönt). Da der Forderung nach Gegensatz zum Ausgang des Hexameters auch ein ebenfalls akzentwidrig betontes anapästisches Wort wie ‘celerí‘ genügen würde, so hat wohl das Streben, am Ende des Distichons noch einmal akzentgemässen () und akzentwidrigen () Wortschluss zusammenklingen zu lassen, das iambische Ende begünstigt” (Römische Metrik, in Gercke-Norden EAW³ 1.8.14). Cf. Eichner, Bemerkungen etc., 24: “Mit choriambischen Wörtern die Reihe zu endigen, scheint Anfangs wenig Anstoss erregt zu haben. Dass man aber bald auch diesen Ausgang verschmähte, dafür weiss ich, wie ich offen bekenne, einen rhythmischen Grund nicht anzugeben, es müsste denn der sein, dass man die starke Betonung der Endsilbe allzusehr in Contrast mit dem natürlichen Wortaccent fand, den man in dieser Reihe von ursprünglich fallendem Charakter so wenig als möglich verletzte.” This explanation is good as far as it goes, except that the objection to choriambic verse-endings, as Eichner states it, is hardly less applicable to the accepted iambic. Wilamowitz suggests that the disyllabic verse-ending was employed with a view to securing a conflict at the end of the pentameter, over against the concord with which the hexameter ends (GVK 5); but an anapaestic trisyllable, as Vollmer points out in the passage quoted above, or indeed any verse-ending of more than two syllables, would do this just as well, and produce a conflict at the penultimate ictus besides—which is evidently what the Augustan elegiste were trying to avoid.

page 66 note 33 See §§87-8, and note 5.

page 66 note 34 For a more exact analysis, see note 43.

page 67 note 35 The self-evidence of this conflict, as a fact, may be observed in every line in which the hemistich does not end in an iambic disyllable; and such hemistich-endings are not common before the diaeresis (see note 32 init. and ch. IV, n. 61). If a monosyllable at the end of the first hemistich were preceded by a dactylic or spondaic word (or by a longer word terminating in dactylic or spondaic rhythm), there would be no conflict at the second ictus; but we have seen (ch. IV, nn. 60 and 61) that these semifinal monosyllables are regularly preceded either by another long monosyllable or by a pyrrhic word—an arrangement evidently calculated to produce just such a conflict.

page 67 note 36 There is, however, considerable disagreement regarding the nature of both ictus and accent: as to the ictus, whether it consisted of an energetic stress, or of a purely metrical predominance of length (in which latter case it could hardly be said to have existed as an ictus); as to the word-accent, whether it was one of tone or of energy. In a conference given at the Sorbonne in 1925, entitled “La lecture à haute voix des vers latins”, Dr. Kent maintains that the word-accent had become among the educated, under Greek influence, tonic in nature, while the ictus as such remained one largely of stress: “Le poète ne devait plus se préoccuper de l'accent verbal, puisque celui-ci ne le gênait plus. Il trouvait que la régularité vers la fin du vers, avec la convention qui y défendait les quadrisyllabes, y attirait les mots où les deux accents s'accordaient, tandis que les mots dans lesquels un tel accord n'était pas possible étaient refoulés vers le commencement. De là provient une plus grande variété de la mélodie dans la première partie du vers, suivie d'un schéma plus régulier et plus uniforme, arrangement très acceptable au point de vue musical” (in Revue Internationale de l'Enseignement 79.332; see also SL³ §66.1). The alternative thesis, to the effect that the meter was marked solely by the succession of long and short syllables, while the word-accent remained one of stress, is defended by Charles E. Bennett, American Journal of Philology 19.361-83 (1898).

page 67 note 37 Servius in Verg. Aen. 11.601. The line is preserved in its entirety by Macrobius (Sat. 6.4.6).

page 67 note 38 In Verg. Aen. 6.868-9, Anchises begins his solemn reply to Aeneas with two such lines; but in the eight spondees there is only one example of agreement between ictus and accent (and this one in the fourth foot), as compared with four in the four spondees of Ennius.

page 68 note 39 In hos regit artus (Verg. Aen. 4.336), regit is in the phrase-rhythm enclitic to hos (cf. Kent, SL³ §§64.111 ad fin. and 67), which is in any event an emphatic monosyllable with an accent of its own.

page 68 note 40 With regard to conflict in the body of the verse, even Fränkel, whose thesis emphasizes in the main the agreement of ictus and accent “im lateinischen Sprechvers”, has the following to say of the hexameter, as the successor of the Saturnian: “Ihr Nachfolger im Epos, der lateinische Hexameter, ist zwar von vornherein ein Rezitationsvers gewesen, wurde jedoch gleich durch seinen Schöpfer hinsichtlich der Flexion, Wortstellung . . ., kurz der gesamten stilistischen Haltung so stark von der Umgangssprache abgerückt und auf ein erhöhtes Niveau versetzt, dass es nur folgerichtig war, wenn man hier auf einen getreuen Widerklang der Betonungen des Lebens verzichtete” (Iktus und Akzent im lateinischen Sprechvers 355, n. 1).

page 68 note 41 The alternative disyllabic ending permitted the hexameter would, with the same suppression, have given a monosyllable in the pentameter (cf. J. P. Postgate, Prosodia Latina §232).

page 68 note 42 A monosyllable at the end of the pentameter, if preceded by a dactylic word or by a short monosyllable, proclitic or enclitic, would have the same effect—but this is very rarely found. The only examples of monosyllabic verse-endings in all the pentameters tabulated are forms of the verb esse, used in close conjunction with the word preceding: Cat. 76.8 (dictaque factaque sunt), Prop. 1.2.26, 2.6.40, 2.10.6, 4.1.146, 4.9.36 (sat est in every instance), Ov. F 4.922, T 5.7.68 (both sat est), and P 1.6.26 (scelus est). This last is in effect a trisyllabic verse-ending, such as Ovid admits for the pentameter only in his Epistulae ex Ponto (see table XLVIII).—Martial employs these monosyllabic forms of the verb esse as verse-endings more frequently, five times in non-final pentameters (3.70.2, 3.95.4, 5.61.2, 7.51.6, 9.70.6), eleven times in the final pentameter of the epigram (3.81.6, 4.7.6, 5.62.8, 7.41.2, 7.81.2, 7.90.4, 8.29.2, 9.13.4, 9.47.8, 10.56.8, 11.2.8). There are in Martial also other monosyllabic verse-endings, used in conjunction with words that may be regarded as proclitic: et hoc (7.10.12 and 14, non-final pentameters), et hic (12.68.6, final), sine te (12.46.2, final), dare vis (7.75.2, final: for the infinitive as a proclitic cf. It. darò from LL dare habeo). That monosyllabic verse-endings of whatever type were felt by Martial to give a special character to the cadence may be indicated by the fact that two-thirds of the 21 examples noted occur in the final pentameter of the epigram, in which the point is turned, although only 30.3% (896/2956) of Martial's pentameters (in the Liber Spectaculorum and books I-XII) are final. (Cf. §91, and note 31, in which the statistics are given for Martial's non-disyllabic verse-endings in general.) This does not preclude the likelihood that even before the close of the epigram these endings had their effect, as is very clearly the case with the repeated et hoc of 7.10.12 and 14.—Where final est (or es) loses its vowel by aphaeresis and attaches itself to the word preceding, as happens very commonly, it cannot of course be regarded as a monosyllabic verse-ending.

page 69 note 43 Words of the type — or -∪∪- (which, terminating on the second ictus of the hexameter, cause a conflict in the first foot) are excluded from the second hemistich of the pentameter, the first by its spondaic character, the second, under the disyllabic law, unless it becomes a pentasyllable by the addition of an enclitic, which shifts its primary accent to the second ictus (cf. §§87 and 92). If the second hemistich begins with a monosyllable, a trochee, or a dactyl, the word-accent will coincide with the first ictus; if it begins with a pentasyllable like virginitāte, the first and fourth syllables will have secondary and primary accents respectively coinciding with the first and second ictuses. (Words of four syllables, like virginibus, as we have said, cannot be used at the beginning of the second hemistich unless followed by an enclitic, which gives them in effect a fifth.) Short monosyllables or pyrrhic disyllables falling in the first dactyl will ordinarily be either enclitic to a preceding, or proclitic to a following word.—A conflict in the first foot may be caused by elision at the second ictus, as in Catullus: cum pretto atque malo (77.2), in Propertius: Carpathium omne mare est (3.7.12), and in Ovid: destriere esse mea (A 2.19.48). (I have noted no example in the Corpus Tibullianum.) But elision is relatively rare in the second hemistich of the pentameter (cf. ch. II, n. 9), and when it does occur ordinarily affects short unnasalized vowels in the off-beat.

page 69 note 44 Only a monosyllabic verse-ending, preceded by a dactylic word, or by a short monosyllable, enclitic or proclitic, would accomplish this (cf. note 42). Such a verse-ending could obviously not be found, or used if found, in pentameter after pentameter.

page 69 note 45 “In der zweiten Hälfte des Hexameters ging der Rhythmus von dem steigenden vermittelt zu dem fallenden über: im Pentameter tritt der fallende Rhythmus in unvermittelten und bewussten Gegensatz zum ansteigenden der ersten Reihe. Der rhythmische Charakter des 4ten Gliedes ist demnach energisches Fallen ...” (Eichner, Bemerkungen etc., 8).

page 69 note 46 Cf. Ov. A 1.1.27: sex mihi surgat opus numeris, in quinque residat, T 3.1.11 : clauda quod alterno subsidunt carmina versu . . ., and the well-known epigram of Schiller:

(See Postgate, Prosodia Latina §231, for a criticism of Coleridge's translation of the German, as ending in a monosyllable, and therefore not fitted “to describe and exemplify the Ovidian Elegiac couplet.”) Catullus 105 is cited by Otto Weinreich, Zur Ästhetik des Distichons, in Neue Jahrbücher für das klassische Altertum 45.88 (1920), as a “drastische Parallele zu Schillers Distichon.” Cf. Schanz-Hosius, GRL⁴ 2.166.

page 70 note 47 With the exception discussed in note 2.

page 70 note 48 These principles will be found in fact to apply to line after line. Their application is especially interesting when the one hemistich is the converse of the other, as in Ov. F 5.394:

or when the two are identical in meaning, as ibid. 1.704:

page 72 note 1 For explanation of signs and abbreviations used in this and the following tables, see preliminary note to this chapter.

page 80 note 1 For statistics on the alternate hexameters of this book, see supplementary table at the end of this chapter.

page 90 note 1 The figures marked with an asterisk are those given by Gummere, NPV 16-7, the ‘selected parts of prose authors’ there cited being Caesar, BG 1.1-22, Cicero, Cat. 1, and Livy 1.1-10.

page 91 note 1 For a summation of the statistics with regard to short declensional inflections in general, see §§80-1.

page 91 note 2 See §§28 and 77. (Of the listings in ch. V, nn. 81-2, the first includes Greek neuters in -ǒn, the statistics for which have been subsumed under N 1s in the tables of this chapter, and under type II in table XLV: cf. ch. V, n. 73.) Similarly—with the exception of six instances of duǒ, in Ovid (H 8.90, F 4.788, T 4.4.72, P 2.3.30, in the first hemistich, H 10.56, T 3.5.10, in the body of the second: see ch. I, n. 11, and cf. ch. III, n. 19)—the numbers following MF lp, for masculine and feminine NAV plurals, represent occurrences of Greek forms in -ĕs and -ăs, preceding an initial vowel (see ch. V, n. 83).

page 91 note 3 See ch. II, n. 2.

page 92 note 1 This table is designed to give for each author and work the number and proportion of pentameters beginning with the sequences dactyl-dactyl, dactyl-spondee, spondee-dactyl, and spondee-spondee. Under vv is found the number of pentameters considered, under ss the number of short syllables in the first hemistichs of these pentameters. The number of metrically significant short syllables in the second hemistichs equals vv × 4—a figure that would also be the maximum number of short syllables possible in the first hemistichs. —For abbreviations of the names of Ovid's works see ch. II, n. 3.

page 93 note 1 For explanation of this and the following tables, see preliminary note to this chapter. Total ss = total number of short syllables, with the final syllaba anceps left out of account in the figure given for the second hemistich (see table XX and note). The key to other signs and abbreviations is given in the preliminary note to chapter VII.

page 111 note 1 Under quae in this table are included all the compound forms: quaedam (Prop. 3.23.6), quaeque (Ov. T 1.3.70, 1.9.22, P 2.3.24), quaecumque (Ov. H 10.80), quaelibet (Ov. T 5.7.54, P 4.5.12). Cf. ch. II, n. 2 ad fin.

page 111 note 2 Add vos, neuter plural (A 3.15.14)—if a pronoun referring to a personified neuter may itself be said to be neuter.

page 111 note 3 Add se, reflexive to a neuter plural (H 8.120).

page 111 note 4 See ch. III, n. 24, also ch. I, n. 13.

page 113 note 1 For the significance of the italicized figures, see preliminary note to chapter VII.

page 114 note 1 This table is designed as a supplement to table XL, to account for all NV masculinefeminine singulars of types I and II in the second hemistich, the endings of which are not scanned as short syllables in the off-beat of a dactyl. The symbol E — elided (but see ch. I, n. 13), C = followed by an initial consonant. —In the first hemistich endings of type I, in ordinary practice, fail to yield a short syllable only if elided, endings of type II only if followed by an initial consonant; the figures for these two phenomena in the first hemistich are therefore respectively equivalent to the difference between the italicized figures and the figures directly above them in columns Ia and IIa of table XL. [There is only one exception in the first hemistich, in Prop. 2.24.4, where an ending of type II receives the ictus before the diaeresis, followed by an initial vowel (cf. ch. IV, n. 14). In the second hemistich Catullus (67.32) loses one short ending of type I by stressing a final ά before initial sp (cf. ch. IV, n. 13).]

page 115 note 1 In this table occurrences of ille, ipse, iste and their feminine counterparts are totaled in three columns, according as they fall in the first hemistich, or in the first or second dactyl of the second hemistich. (Separate figures for the three words, in the order named, are given in parentheses.) IlIa, ipsa, ista, NAV neuter plural, are not included (see ch. IV, n. 36).

page 116 note 1 Symbols: .. = pair of like forms in body of second hemistich (see ch. IV, n. 59); a..a = pair of like forms, bracketed by words in agreement with each other at the ends of the two hemistichs; ar..ar = pair of like forms bracketed by words both agreeing and ‘rhyming’ with each other. The figures for .. include those for a..a and ar..ar; the figures for a..a include those for ar..ar. The percentages represent the number of occurrences in 100 pentameters.—For definition of the term ‘agreement’, as here used, see table XLIV, note 1.

page 117 note 1 The figures in the first column include those in the second. The percentages in these two columns represent the number of occurrences in 100 pentameters, those in the third column, under the sign %, the proportion of agreements that are accompanied by homoeoteleuton.—The term ‘agreement’ is here restricted to the modification of one declinable word by another as attribute, predicate, or appositive; instances in which like forms are coordinated with each other (as in Cat. 66.2, 68.106, Tib. 1.4.2, Prop. 2.3.44, 2.9.36, 2.16.2 2.34.78, 3.14.18, Ov. A 3.2.34, 3.6.48, F 2.132, 4.408, T 5.6.44), or otherwise rhetorically counterpoised while remaining grammatically: independent (as in Ov. P 4.8.12, and mediately in A 3.11.48), or used in successive subordination (as in Prop. 2.27.6), have not been included. The figures may therefore be regarded as minimal.

page 119 note 1 For the significance of the italicized figures, see preliminary note to chapter VII. Percentages of utilization have been omitted except for the five works of Ovid taken together; where the numbers are large enough to yield significant proportions, others can readily be computed for comparison with this average.

page 121 note 1 See table XLV, note 1. By a kind of anomaly, mihĭ, tibĭ, and sibĭ have been retained under type III, to which mihī, tibī, and sibī belong (cf. §32); the italicized figures here represent, with one exception (the Greek dative Tēthyĭ in the second hemistich of Cat. 66.70), occurrences of these dibrachs. In addition, Ovid has seven examples of sibĭ plural in the first hemistich (F 2.246, 2.420, 3.674, 6.654, T 5.10.16, P 2.5.24, 2.7.18), and one in the second (F 5.672), which as plurals could not be included in this table. [The Ehwald-Levy text of the Fasti has tibi in 2.246, where Merkel has sibi reflexive to a plural subject; in 5.672, where Merkel reads rogant, it has rogat, which would make the sibi singular. See ch. II, §8 and n. 4.]

page 122 note 1 In this table, ‘separable enclitics’ are those that do not modify the meaning of the words to which they are attached as conjunctions. Under ‘inseparable -que‘ are subsumed forms of quisque and uterque, the conjunction atque, such adverbs as usque, adusque, ubique, utrimque, and the suffix -cumque (see ch. VI, nn. 22-3). The -ve of sive and the -ne of anne have here been regarded as separable enclitics, by extension of the definition above given; in any event, the words rarely occur in this position (sive: Cat. 76.16, Tib. 3.4.12, Prop. 3.13.46, Ov. T 3.8.6; anne: Ov. F 6.28 and P 4.13.30).—Proportions based on the number of pentameters with disyllabic ending (see table XLVIII) are given in parentheses before those based on the total number of pentameters, wherever the two percentages differ appreciably.

page 122 note 2 If the reading of Ehwald-Levy is adopted in Ov. F 2.782 (see ch. VI, n. 12), this number will have to be reduced to 1, and the preceding 142 increased to 143.

page 123 note 1 To supplement these statistics: there are no non-disyllabic verse-endings in the pentameters of the Ars Amatoria, the Remedia Amoris, the Medicamina Faciei Femineae, and the Nux of Ovid; his Ibis contains one pentameter ending in a hexasyllable (508, bracketed by Merkel as suspect), one in a tetrasyllable (520). Of the 3311 pentameters of Martial, three end in hexasyllables (one Latin, 5.50.2; two Greek, 1.45.2 and 14.201.2), 38 in pentasyllables, 187 in tetrasyllables, 132 in trisyllables, 2930 (88.5%) in disyllables, 21 in monosyllables. For a special feature of Martial's practice see ch. VI, nn. 31 and 42. (In determining the number of syllables in the verse-ending, the final word has been taken as spelled, no attention being paid to prepositions and conjunctions which as proclitics form part of the final vocable. Cf. ch. VI, n. 42.)

page 123 note 2 These monosyllabic verse-endings are cited in ch. VI, n. 42.

page 123 note 3 Two pentameters (114.6 and 115.4) listed among the omitted lines in §8 have here been included, as having endings that are unimpaired.

page 124 note 1 See §8, with note 4, in chapter II.

page 124 note 2 The work of Maurice Platnauer, Latin Elegiac Verse; a Study of the Metrical Usages of Tibullus, Propertius & Ovid (Cambridge 1951), appeared when this dissertation was already in the hands of the printer.