explanation; though we may not now be able to discover it. And indeed the explanation, though transparently simple, accords so little with an English conception of language and metre, that I should fear to propound it, if I were not able to cite authorities which will not be contemned. Catullus, in Poem 76, is credited by the MSS. with the pentameter quare cur te iam amplius excrucies?

Mr Munro' restores

quare cur te iam a! amplius excrucies?

which, considering the tendency of Latin to these interjections (as in Latin they really were), and "the fondness of Catullus" for this one in particular, seems, when done, obvious enough. Horace would assuredly not have written iam a!, not being so well content as Catullus with the genuine language of the Romans. But he learnt from Catullus, as from every one; and nothing could be more in his manner than to use with 'art' what Catullus used with nature. If the last syllables of invida and amica be lengthened into sighs (which is the effect in sound of writing invida a! and amica a!), the metre is saved. Whether the rhetoric loses, the reader must decide. Catullus would not have disliked it, for he wrote

num te leaena montibus Libystinis

aut Scylla latrans infima inguinum parte
tam mente dura procreavit ac tetra

ut supplicis vocem in novissimo casu

contemptam haberes a! nimis fero corde??

bringing in his a! precisely as Horace, to give emphasis to the close. Nor did Horace dislike such an end to the line, for he wrote,

ad non amicos heu mihi postes et heu
limina dura3.

1 Elucidations p. 207, citing for the unelided interjection Hor. Epod. 5. 71, Tib. (Lygdamus) III. 4, 82, (Sulpicia) 11. 3.

IV.

2 Cat. 60.

Epod. xI. 21. A hiatus, which, to

judge from Horace's general practice, he would have found highly objectionable, may be avoided in a similar way in 1. 2. 47, neve te nostris vitiis iniquum O! ocior aura tollat. The close juncture of the two last lines of the

The remaining examples are I. 17. 13 dis pietas meă | et Musa cordi est: and II. 13. 7, 8, penetralia sparsisse nocturno cruore hospitis; ille venena Colcha | et quidquid usquam concipitur nefas etc. To these may be applied, with at least equal strength, the argument by which Bentley disposed of the exceptions to the law of continuity in the anapaests of Greek tragedy-they cannot be the effects of chance, because they are (1) too few and (2) too easily corrected. Bentley inferred that the exceptions were false readings; strictly he should have inferred that they must either be false readings or be justified by special circumstances; and his rule has since received a qualification accordingly. If it can be shown what was the object of Horace in thus surprising the ears of his audience, we can believe these exceptions genuine'. Till then, it must remain more probable that they are clerical errors. To correct such errors with certainty is of course impossible unless, which is not the case here, there is only one likely way. The originals may have been e.g. dis pietas mea et dis musa cordi est, or dis pietas meis et musa cordi est, and in the other place (where the alternative Colchica is actually given by the MSS.) penetralia sparsisse nocturnos cruores hospitis; ille venena Colchica et quidquid etc., which would be an imitation, quite in the author's manner, of a construction familiar in the Greek poets2.

Sapphic is very common. Similarly in III. 9. 22, tu levior cortice et improbo iracundior Hadria, the last syllable of improbo has the accent of an exclamation, which accounts for the single collision in the piece. Whether we write improbo O! or not makes no difference in recitation.

1 As II. 13. 5-12 exhibits three exceptions in eight lines (7, 8, 11), it might be argued that they are genuine and mark arbitrary pauses, the effect of excitement. But this is improbable in the face of the careful juncture of vv. 6 and 7, where the ǎ of penetralia is lengthened before the double consonant.

2 As a Greek grammar would say, 'the verbal phrase sparsisse-cruores (blood-besprinkle) governs the accusative object penetralia'. Compare the 'accusative of the part affected' in II. 7. 7, the 'accusative in apposition to the sentence' in 11. 20. 7, and the numerous Greek infinitives and genitives. For the plural cruores see the Dict. s. v. In Epod. v. 87, it seems the better view that the construction is venena non valent convertere-humanam-vicem magnum fas nefasque, 'Spells cannot make human change in the (divine) law of right and wrong', vicem being the quasi-cognate accusative to convertere. In г. 19. 15 tres (cyathos)

The case is scarcely less strong against the non-elision (except at strong stops) of the syllable -um. It is elided constantly, and at the end of the line as well as elsewhere. To neglect the elision for mere convenience is therefore a gross offence against euphony. The artist who appeals—and Horace makes this appeal, if ever poet did-to the close attention of the ear cannot afford to write at one time

and at another time, as if it were a matter of no consequence,

or

neve te nostris vitiis iniquum

ocior aura tollat (1. 2. 47)

te triste lignum, te caducum

in domini caput immerentis (11. 13. 11).

Both ways cannot have been natural or euphonious to the same ear. The apparent cases of non-elision are, I think, six-the two just given, and the following,

In I. 31. 13 the correction Atlanticon would be easy and legitimate; but upon reflexion it will be seen that there is really no collision at all. Whatever be read, a pause before impune is required for the point. The self-approving merchant used the word in the sense of unharmed, the poet clearly in the sense of unpunished (see 1. 3. 23). When the last word of a sentence has this sort of significant emphasis, the voice naturally pauses before it; and the verse is so written as to make this pause necessary.

Of 1. 2. 47 I have spoken already1; the principle applying to it extends also to II. 13. 11, a place obviously appropriate for an interjection (perhaps heu), answering to the English Fie! Two of the others are probably right, but are not to the point. In III. 27. 10 some perceptible pause is natural after the long clause antequam...imminentum, and moreover the final syllable of this genitive had possibly a long vowel u, and was therefore a different sound from that in apricum, iniquém, caducům. The same reason justifies leonu' arida. In I. 8. 3 I have myself not much doubt that the true reading is apricus. The meaning is, 'Why has Sybaris, who (in summer) could stand any extreme, now (in winter) turned delicate, shrinking from the cold of the Campus and the river?' For the sense of apricus, and the use of the word in the Odes, see the note on p. 1432.

But the reader, though he be patiens pulveris as Sybaris himself, must by this time have had enough of this critical dust. As an excuse for raising it we may plead that, if Horace deserved the wreath of Melpomene, it was not till he had toiled for the condescension of Euterpe. And therefore let us join them together.

1 See note 3 on p. 191.

2 The inflexions -em and -am are on a somewhat different footing. They are not so often elided, never at the juncture, and seem to stand like long vowels at the boundary between what must be elided and what cannot be. The instances of non-elision in 1. 15. 2,

III. 4. 76 and III. 24. 61 (-am), I. 18. 15, II. 5. 9, III. 24, 60, III. 27. 33 (-em) may therefore be genuine. In 111. 24. 60 however there is a variant hospites adopted by some editors; and some of the others, especially III. 27. 33, where there is no stop at all, appear suspicious.