example. The note tells us too, what we have abundant proof of in other scholia of Servius and Acron, that in their day the genitive of description was the usual idiom, and the ablative of description had become archaic. Probably it was becoming archaic even in Virgil's day, and this was his reason for preferring it. But it would be strange if it were always the ablative, and never the genitive, that we found in the transferred form-always turbidus caeno, and never turbidus caeni. Naturally Virgil offers no such example of the genitive, for he shuns the genitive of description. But I find in Tacitus: reductus in hiberna miles laetus animi (Ann. 2. 26. 1), and: principem longa experientia eundemque severitatis et munificentiae summum (1. 46. 2), and in Lucretius: longa diei infinita aetas (1. 557).

In: hamis auroque trilicem we have an expansion of our idiom, due apparently to hendiadys. But the most usual form of expansion is due to the addition of an adjective to the ablative of specification, giving it again the form of the ablative of description, from which it was derived; as in: per orbem aere cavum triplici (Aen. 10. 784), asperque inmani corpore Thybris (8. 330), caligine turbidus atra pulvis (11.876). To see the effect of this addition we may compare: Cacum ...timentem turbatumque oculis (8. 223) with: spelunca alta vastoque inmanis hiatu (6. 237), purpurei cristis iuvenes auroque corusci (9. 163) with turbidus imber aqua densisque nigerrimus Austris (5. 696), and: pietate insignis et armis (6. 403) with: sororem ... pedibus celerem et pernicibus alis (4. 180), aureus et foliis et lento vimine ramus (6. 137) with: terribili impexum saeta cum dentibus albis (7. 667).

Opposed to this tendency to expansion are the contractions we have in: prata recentia rivis (6. 674) for: prata virentia recentibus rivis, turris... opportuna loco (9. 531) for: turris opportuno loco sita, Messapus... altus equo (12. 295) for: Messapus alto equo vectus. Further examples are: laetantes agmine cycnos (1. 393) for: cycnos laetanti agmine volantes, aurea bullis cingula (9. 359) for: cingula aureis bullis ornata, galeam ... cristis... decoram (9. 365) for: galeam cristis decoris aptam, saeva sonoribus arma (9. 651) for: saevo sonitu sonantia arma, urbs Etrusca solo (10. 180) for: urbs Etrusco solo sita.

But in furiis accensas pectore matres (7. 392) we have an ablative of cause joined with this ablative of respect. To: atro... membra fluentia tabo (3. 626) Servius's note is: fluentia tabo pro fluenti tabo. Evidently Virgil's phrase is for: membra atra fluenti tabo, and is got by a double transference of atra from membra to tabo and of fluenti

from tabo to membra. But fluenti tabo is an ablative not of description, but of cause. Here then we have an extension of this transference; it arises not merely from ablatives of description, its usual source, but also from ablatives of cause.

Far more usual is the transference from ablatives of manner and means, as in: animis opibusque parati (2. 799), nos. . . effusi lacrimis (orabamus) (2. 651), sola fuga nautas comitabor ovantes (4. 543), proximus ingreditur donis (5. 543), furtivum partu... edidit (7. 660), ibat iam mollior undis (8. 726), praeceps saltu sese... in fluvium dedit (9. 815), animo spem turbidus hausit inanem (10. 648), equis aversi ad moenia tendunt (11.871), saltuque superbus emicat in currum (12. 326), nec numero inferior pugnae nec honore recedes (12. 630). We have this transference from ablatives of material in: crateres auro solidi (2. 765) and: argento clari delphines (8. 673).

In: Gallica ora (Od. 1. 8. 6) we have clearly a shortened form for : equorum Gallicorum ora, where the epithet has been transferred to ora from its dependent genitive, giving us: (equorum) Gallica ora. We have parallel shortenings in: sacer paries (1. 5. 14) for: sacrae aedis paries, in medias fraudes (3. 27. 27) for: medii ponti fraudes, in: crudeles terras... litus avarum (Aen. 3. 44) for: terras crudelis regis ... litus avari regis. More difficult is: cavae aedes (2. 487), which Servius explains as: camerata tecta. Probably we may resolve it into: aedes cameris cavis constructae, and: tela Typhoia (1. 665) into: tela velut ea in Typhonem coniecta. Domum ambiguam (1. 661) seems for: domum feminae mutabilis, as: pallida Mors (Od. 1. 4. 13) is for: Mors quae corpora pallida reddit, and: inaequali tonsore (Ep. 1. 1. 94) for: tonsore qui capillos inaequaliter secat. But: commissa piacula (Aen. 6. 569) for: scelera pianda, and: in aperta pericula (9. 663) 'into the perilous breach' seem to be based on double or perfechypallages such as we have repeatedly examined in this chapter.

XXXVIII

OMISSION OF THE PREFIX

No student of Latin poetry can have failed to notice the frequent omission of the prefix in composition with Latin verbs, as we see it in tum Thetidi pater ipse iugandum Pelea sensit (Catull. 64. 21), where it is plain that iugandum is for coniugandum, and this leads the reader to the conclusion that sensit is for consensit, where the omission decidedly affects the sense. So in: denique saepe hominem paulatim cernimus ire (Lucr. 3. 526), where: vitalem deperdere sensum in the next verse shows the reader that ire is for perire, 'pass' for 'pass away. The figure is met at times in English poetry. We see it in Hamlet's: This bodes some strange eruption to our state', in Gray's: 'Graved on the stone beneath yon aged thorn', in Marvell's:

'And does in the pomegranate close,

Jewels more rich than Ormus shews'

It is not characteristic of Latin classical prose; indeed Cicero seems to avoid it; the student will remember the way in which he piles prefix on prefix in: abiit, excessit, evasit, erupit (Cat. 2. 1. I).

Often the meaning of the verb of itself suggests the prefix omitted, as in modo me Thebis, modo ponit Athenis (Ep. 2. 1. 213) for deponit, or frontis ad urbanae descendi praemia (1. 9. 11) for condescendi, or: quaeque aliae nationes usque ad Albim colunt (Ann. 2. 41. 2) for incolunt. Indeed at times in later writers the form of the verb at once suggests the loss of the prefix, as in: quo cingi cludique terrarum orbem hinc fides (Tac. Germ. 45. 1), where cludi for claudi is evidently short for concludi. In: si numerus militum potius quam legionum putetur (Hist. 3. 2. 7) the use of putare in its old sense of 'to sum up' straightway suggests computare. But when in orare Iovem qui ponit et aufert (Ep. 1. 18. 111) ' sets before and takes away', Lucian Müller tells us that ponit is for proponit, he fails to see the etymology of pono (= por-sino 'I set before'). Often the preposition omitted becomes clear from the context, as in bracchia candidae cervici iuvenis dabat (Od. 3. 9. 3)

or circumdabat, acrem militiam paras (1. 29. 2) for comparas, concursu ad ianuam facto moliuntur fores (Ann. 1. 39. 4) for demoliuntur. In: cessatum ducere somnum (Ep. 1. 2. 31) that ducere is for reducere is readily deduced from: cui pulcrum fuit in medios dormire dies (v. 30) whose ideal it was to sleep till midday. In: multa inter sese vario sermone serebant (Aen. 6. 160) the vario suggests the missing prefix dis-, and in: pueris... beata creandis uxor (Ep. 1. 2. 44) the missing pro- will be suggested by uxor and pueris. In: infelix operis summa quia ponere totum nesciet (A. P. 34) that ponere is for componere is suggested by: si quid componere curem in the next verse; the importance of this for the figure we shall presently see. But in pectus praeceptis format amicis (Ep. 2. 1. 128) the presence of invidiae in the next verse does not at first help me much, and I feel grateful to Acron for his explanation that format here is for informat' instructs'. Servius's note to: nec vim tela ferunt (Aen. 6. 400): non obferunt, tantum repellunt, is very welcome too. In: cui nomen Superiori, sub C. Silio legato; inferiorem A. Caecina curabat (Ann. 1. 31. 2) legato suggests that curabat is for procurabat. So in: it pectore summo. per collum circulus auri (Aen. 5. 558) it seems to be for circuit, and the prefix i implied in circulus.

...

The prefix omitted is usually what we call a preposition, but it is evidently not confined to this class. It is quite plain that the ancient grammarians called any words prepositions that were put in composition with verbs. Potis or pote, which is compounded with est in potest, is one of the particles with which we have to do. In utputa and utpote we have ut apparently entering into such composition; and accordingly we read in Donatus (p. 389, K.): sunt etiam dictiones, quas incertum est utrum coniunctiones an praepositiones, an adverbia nominemus ut cum et ut. So we read eat for veneat in Claudian's verses:

Tot Galatae, tot Pontus eat, tot Lydia nummis,

Si Lyciam tenuisse velis, tot milia ponas (In Eutrop. 1. 203-4). We have the prefix omitted from nouns as well as verbs. Virgil has theatri for amphitheatri in: ad tumulum cuneosque theatri (Aen. 5. 664); when he first speaks of it, he uses for amphitheatrum the periphrasis: theatri circus (5.288). For Acroceraunia he uses Ceraunia (3. 506), as does Propertius too (1. 8. 19), very much as to-day we have Salonika for Thessalonica. So Adryasin (Prop. 1. 20. 12) seems a similarly shortened form for Hamadryasin (v. 32). In: cras vel

atra nube polum Pater occupato vel sole puro (Od. 3. 29. 44) Pater seems for Diespiter, and in return: Assyria . . . nardo (2. 11. 16) seems a lengthened form for Syria nardo.

For we have also the opposite of this omission in poetry. It is quite plain that in: ubi non Hymetto mella decedunt (2. 6. 15) decedunt is lengthened for cedunt; and so in: prospere decedentibus rebus (Suet. Iul. 24. 3). In: seu cum se Martia curru Penthesilea refert (Aen. 11. 661) Servius explains that refert is for fert or infert, and to: nec plura adludens (7. 117) Servius suggests: vacat ad et ludentem significat. So in: tuaque exspectata parenti vicit iter durum pietas? (6. 687) he explains that exspectata is for probata, as is spectata in: rebus (gestis) spectata iuventus (8. 151), which implies that ex in: exspectata pietas is superfluous. But it seems to me that: tua exspectata pietas is for: spectata pietas tui exspectati.

But the omission of the verb may be taken as the opposite of the omission of the prefix; does that ever occur? We read: neque erat Lydia post Chloen (Od. 3. 9. 6), where post seems short for postposita. When the verb is dropped and the preposition only remains, the dative must pass to an accusative, and this, no doubt, is why it is so rare with prepositions. But this difficulty is not felt with potis or pote. So we have: quam potis (est fieri) (Pl. Mil. 781), hoc facias, sive id non pote sive pote (Catull. 76. 16), nil pote supra (Ter. Ad. 264), quid pote simplicius? (Mart. 9. 15. 2). If potis or pote can be used for potest, what of the more usual ellipsis? Is not est used for potest? It would be strange if it were not. We read:

Ut caput in magnis ubi non est tangere signis

Ponitur hic imos ante corona pedes (Prop. 2. 10. 21-2), as when in tall statues one cannot touch the head, the wreath is laid here before the feet'. And so in: si quid usquam iustitia est (Aen. 1. 604), est quadam prodire tenus, si non datur ultra (Ep. 1. 1. 32), neque est te fallere quidquam (Geo. 4. 447), a use corresponding exactly to pote in the examples cited. Like pote in Pl. Mil. 781 it is often for potest fieri, as in nil erit ut distet (Lucr. 1. 620), non est ut copia maior ab Iove donari possit tibi (Ep. 1. 12. 2), where it is nearly equivalent to licet. When for the subjunctive with ut there is substituted the infinitive, as in: nec non et Tityon... cernere erat (Aen. 6. 596) it becomes equivalent to licet. The Greek use of ἔστι in: ἔστι μὲν εὕδειν, ἔστι δὲ τερπομévoiσiv åkoveiv (Od. 15. 392) is parallel, but does not account for the