This passage distinguishes two classes of words, those which have a main accent and those which have only a subordinate accent. In the first class are included only the noun, verb, and participle; in the second class are included all other words. These dependent words as they occur in the sentence do not form a definite word-group with the more important words, but they resemble the enclitics of the word-group in being subordinate in accent to the words on which they depend. This is brought out by the expression “ut tanquam in unam partem orationis coalescant." To illustrate this relation of the accent of the subordinate word to the main word, we may take such types as the word-groups, bene rém gerit, ad illam hóram. The accent here indicated is the normal accent of Plautus (Trans. Am. Phil. Assoc. XXXIV, p. 73 ff.; Lindsay, Latin Lang. p. 170). With these types we may compare the union of the adjective with its noun, or of an adverb with its verb in such cases as Aen. xi, 664 postremum, aspera virgo; Juv. vi, 281 ipsa, olim convenerat. Aspera and olim would seem to be examples of such words as Audax classifies as subordinate words which do not have a sentence-accent.1 We may refer to classes of words as having a certain normal sentence-accent, though there may in certain cases be a variation from the normal. In a similar way we may speak of the normal pitch of the vowels, though their position in the word or sentence may cause a departure therefrom. It is apparently the normal sentence-accent of classes of words to which the passage in question refers. In examining the exceptional instances which have been quoted (p. 86) from Vergil, Horace, and Catullus, we see that has been omitted after the words quoted, and that two passages treating of two different classes of accents have been brought together.

1 We cannot draw a definite line between sentence-enclitics and subordinate words which have a secondary sentence-accent. For example, unemphatic pronouns are enclitics, but we cannot draw a definite line between their use as enclitics and as accented words. The sentence-accent of ante venit must have resembled ante Caesarem much more closely than it did Rómam venit, though the preposition is classed as a sentence-enclitic, while ante (adverb) and Romam are regarded as accented words. Such a combination as is formed by the union of a subordinate sentence-word (but one which is not an enclitic proper) with an accented word, we may call with Sweet a stress-group, or with Sievers a 'Sprechtakt' (Phonetik 5, § 621, 233).

the usage of these authors differed somewhat. In Vergil, besides the cases already considered (pp. 84 f., 87 f.), we find the following words in the second syllable of pause-elision: hic, hoc, hunc, hic (adv.), hinc, ille, illi, optime, aspera, una, ingens, udo, atro. The vocatives, xi, 353 optime regum and xi, 664 aspera virgo, may be compared with Aen. viii, 364 (p. 87). There is only a slight pause, if any, before the vocatives, and the phrase-accent is on regum and virgo. In the case of una and ingens the first syllable of the elision is -que. The vowel in -que was doubtless very lightly sounded, and in some cases was almost a negligible quantity; and in this as in many other respects, the stately hexameter reflected the usage of the spoken language. This light pronunciation would apply to the -que at the end of so-called hypermetrical lines and in pause-elision, especially if such a syllable was not found in the author in question in other combinations.1 The case of una differs from the others we have considered in that the vowel in the second syllable of the elision is long.

Only two cases in Vergil remain to be considered. If our texts containing udo and atro in pause-elision are correct, these two cases would stand by themselves as quite exceptional. The first of these is Aen. v, 681:

indomitas posuere ; udo sub robore vivit.

Ribbeck's critical edition2 has the following note on udo: "duro Macrobii w exc. Par. 1.” He does not, however, state that some of the best Mss of Servius imply a variation in the reading. Thilo and Hagen's edition of Servius contains the following note on this passage: "VVDO SVB ROBORE F VDO SVB RUN VDO SVB D. R. LM VDĒ SVBDER H. propter D. in Servianis libris additum moneo Macrobii codices Sat. vi, 6, 18 duro sub robore praebere." 2

1 Professor Lindsay (Capt. 26) shows that e was entirely silent in certain cases in Plautus, and he compares with this the hypermetrical use of -que and -ve in classical poetry. The fact that a vowel with m is similarly used by Vergil (Georg. i, 295; Aen. vii, 160) is opposed to the theory that the hypermetrical syllable was entirely silent. Again, the general character of epic verse as contrasted with dramatic would tend to keep the vowel sound from being entirely disregarded in pronunciation (cf. p. 101). We may even doubt whether the vowel in est was always silent in elision. Cf. Mar. Sacerd. K. VI, 493.

2 See footnote on p. 110.

The VVDO of F may be explained as arising from duro contained in a Ms written in the cursive hand, or as a perpetuation of an error arising from this source. d and rare often interchanged in Vergil (Ribbeck, Prol. p. 243). One example is also found of the confusion of d and v. In N the break is explained by reference to LM, where D. R. evidently stands for duro robore. In H these letters have given rise to SVBDER. There was probably a marginal correction indicating that duro was to be substituted for udo and to be placed after sub. The change from udo to duro in our texts involves the change from posuere to ponunt, and this is one of the strongest grounds for the change of udo. There is a definite unity of thought and expression from line 675 to 686, or even to 692. All the verbs of the passage are in the present tense with this one exception, and this variation of tense does not seem necessary, or even justified by the thought. With these changes the passage would read as follows (v, 680-682):

sed non idcirco flamma atque incendia vires

indomitas ponunt; sub duro1 robore vivit

stuppa vomens tardum fumum, lentusque carinas.

The rhythm of the verse in this form has not the lightness and rapidity of the version in our texts, but it is more in harmony with Vergil's method in such passages as the one before us. The rhythm of the texts is quite out of keeping with the thought as well as with the rhythm of the preceding and the following line. It seems entirely inappropriate to use the rhythm to describe a fire eating its way into wet wood which would be in place in describing a fire crackling in the dry leaves.2 In the line as given above we have the slow, stubborn progress of the fire working its way into the hard oak, described in a form which is full of energy and

1 Macrobius's reading is duro sub robore, and this more nearly preserves the rhythm of our editions.

2 This light rapid rhythm seems entirely in harmony with i, 175:

succepitque ignem foliis atque arida circum
nutrimenta dedit rapuitque in fomite flammam.

Cf. ii, 705 and 706.

force, and one which is in harmony with the preceding and following line. This rhythm is frequently used by Vergil to express that which is impressive and awe-inspiring. If the Roman critics with their tendencies to realism in art, although these tendencies were utterly at variance with the ideals of Vergil, were confronted with the two readings of this verse, they might be expected to prefer udo as being in harmony with the attempts to extinguish the fire. Duro is, however, more in keeping with the description of the fire as making irresistible headway in the hard oak. Whatever be our view of the origin of these two versions, or of their relative merits, it would seem highly probable that at an early date, certainly before the time of Macrobius, there were two readings of this verse, and that the retention of udo involves a form of elision which is most exceptional in Vergil.

I shall not dwell at length on the last exceptional cases in Vergil, ix, 333:

tum caput ipsi aufert domino truncumque relinquit

sanguine singultantem; atro tepefacta cruore
terra torique madent.

Ribbeck's note on this is as follows: "ATRO potest et superioribus adplicari, ut sit intellectus: sanguine singultantem atro, potest et sequentibus iungi, sed melius sequentibus. Servius Dan. Post singultantem interp. M."

Little dependence is to be placed on the punctuation of our early Mss. It is often from a later hand. In a passage like the one before us, sanguis is often modified by ater. Here it seems more natural with sanguis than with cruor, which is practically synonymous with sanguis ater. The emphatic position at the beginning of the second clause does not seem appropriate. The hephthemimeral caesura is here more natural and effective than the penthemimeral. L. Müller (de Re Met.2 204) regards this pause as frequently 1 We may compare the line in the form given with ii, 218:

bis medium amplexi, bis collo squamea circum.

The latter part of the line may be compared with such lines as ii, 85, 107, 133, 278.

employed "ubi asperitatem cruditatemque rerum versus congrua imagine quasi depingit." This line is practically a runon-line, and is more easily read as such if the caesura comes after atro, and this is more in harmony with Vergil's usage when there is no pause in the following line before the penthemimeral.

In Horace the most exceptional cases of pause-elision are in connection with the words addens, accipe, and aufer. It is important to notice that these are found in Sat. ii, 3. The language of this satire resembles that of comedy and in its use of elision is influenced by the drama. The percentage of elisions in this satire is twice as great as it is in the other satires of the second book. If we take into account only the lines in which these elisions occur and in each case the four preceding and following lines, we find for these twenty-seven lines a higher percentage of elisions than the average of Lucilius.1

With reference to Catullus we have already referred to the use of Hymen (p. 89) and arva (p. 88). In 8, 9 tu quoque, impotens, noli, impotens may be compared with the vocatives in Aen. viii, 364 (p. 87), and xi, 353, 664 (p. 92 f.). The stronger accent on the vocative in this case, owing to its emphasis and its use as a noun, is counterbalanced by the light syllable -que in the first syllable of the elision (for the sound of -que see p. 93, footnote 1). The most noteworthy cases which remain are the following: 9, 1 Verani, omnibus; 14, 19 Suffenum, omnia; 75, 4 amare, omnia; 114, 3 auspicium, omne. The language of these poems reflects the freedom of the conversational style, and in the use of elision shows the influence of the drama. The percentage in these four poems is nearly as high as the normal percentage of the Latin language,2 and

1 The accent of addens, accipe, and even aufer was doubtless comparatively weak, owing to the fact that it fell on the prefix. This view is favored by the usage of the Vulgar Latin, which transferred the accent in compound verbs from the prefix to the stem vowel of the verb. Compare demorat with Ital. dimora, Fr. demeure (Class Rev. V, 407; Lindsay, Latin Lang. 164). It is noticeable that the accented prefix of the verb occurs in pause-elision in a few instances, though the accented stem vowel of the verb is not thus admitted.

2 Maurenbrecher, Hiatus u. Verschleifung im alten Latein, 251.