* A Distributed Proofreaders Canada eBook *

This ebook is made available at no cost and with very few restrictions. These restrictions apply only if (1) you make a change in the ebook (other than alteration for different display devices), or (2) you are making commercial use of the ebook. If either of these conditions applies, please check with an FP administrator before proceeding.

This work is in the Canadian public domain, but may be under copyright in some countries. If you live outside Canada, check your country's copyright laws. If the book is under copyright in your country, do not download or redistribute this file.

Title: The Latin Dual and Poetic Diction

Date of first publication: 1923

Author: Andrew J. Bell (1856-1932)

Date first posted: August 28, 2013

Date last updated: August 28, 2013

Faded Page ebook #20130838

This ebook was produced by: Louise Hope

This e-text includes characters that require UTF-8 (Unicode) file encoding, primarily single Greek words or phrases:


There are also a few words of transliterated Sanskrit, and some Latin words printed with macrons or breves (used only when required by the subject):


If any of these characters do not display properly—in particular, if the diacritic does not appear directly above the letter—or if the apostrophes and quotation marks in this paragraph appear as garbage, you may have an incompatible browser or unavailable fonts. First, make sure that the browser’s “character set” or “file encoding” is set to Unicode (UTF-8). You may also need to change your browser’s default font.

Missing periods (full stops) at the end of Index entries have been silently supplied. Other typographical errors are shown with mouse-hover popups. The word “invisible” means that there is an appropriately sized blank space, but the expected text is not present. Asterisks are in the original.

The Latin Dual & Poetic Diction





Macdonald Professor of Latin in Victoria College, and Professor of
Comparative Philology in the University of Toronto





Printed in England.


London   Edinburgh   Glasgow   Copenhagen
New York   Toronto   Melbourne   Cape Town
Bombay   Calcutta   Madras   Shanghai
Publisher to the University



It will be objected that in our Grammars there stands no mention of a Latin dual. Nor does there stand there any mention of a Latin objective case, or of a Latin aorist. The Latin objective has been obscured by an unfortunate mistranslation of the Greek term αἰτιατική; but the dual did find a place in the oldest Latin Grammars we have, and the term aorist now plays a part in Latin grammatical terminology that seems to me far too great. But in the study of any language the part played by the words not expressed, but to be supplied in the mind of the reader or hearer, is very great, and I have endeavoured in the following pages to determine the origin and nature of some ellipses peculiarly Latin, that have not hitherto been adequately treated.

The ellipsis has in the past played a great part in investigations into Latin syntax; and the reason for this is evident the moment we try to construe into our own tongue a few consecutive verses of Virgil or Horace. But the older grammarians of modern times, such as Sanctius, applied to the solution of the problems thus presented the ellipses familiar to them in their own tongues, instead of endeavouring to discover the ellipses peculiarly Latin. This tendency wrought such mischief in their investigations into Latin syntax that there succeeded to it an opposite tendency to assume as few ellipses as possible in conducting such investigations. In many of the constructions which Sanctius attempted to solve by ellipses this tendency has justified itself; but it soon becomes evident to all students that Latin, like English or French or German, has its peculiar ellipses, and to understand the language they must be determined. In many cases we come to perceive that vi the words we do not see are of quite as much importance to the meaning as those we do.

For their constant support and sympathy I have especially to thank Chancellor Bowles, and my colleagues, Dean Robertson and Dean Wallace. My colleague in Latin, Professor De Witt, has helped me constantly with suggestions, and I have to thank him especially for his help in the question of Acestes’ arrow. My old friend and class-mate, Professor Keys, has always been ready with sympathy and help, especially in questions affecting the Romance languages. I have further to thank Principal Hutton, Professor Alexander, Mr. Langton, Professor W. P. Mustard, Professor W. Sherwood Fox, and many others, for the kind patience and sympathy with which they have endured my demands on their time and attention. And lastly I have to thank the Board of Regents of Victoria College, whose generous help enables me to publish this book.

A. J. B.

Victoria College.



I. Geminusque Pollux 3
II. Domitus Pollucis habenis 9
III. Pilumnus and Pitumnus 16
IV. Liberi 19

The Numbers in Greek and Latin, and their Relations

VI. The Dual in Latin 25
VII. Inflections of the Dual 29
VIII. Inter 36
IX. Alter and Alius 39



Use of the Dual for the Plural and of the Plural for the Dual


Use of the Singular for the Dual and of the Dual for the Singular

XIII. Use of the Singular for the Plural 67
XIV. Use of the Plural for the Singular 72

The Dual Pronouns σφώ, σφῶϊ, and σφωέ


The Schema Pindaricum and Allied Constructions


The Schema Alcmanicum, and further Syntax of the Dual

XVIII. Numerals, Cardinal and Distributive 102
viii XIX. Constructio ad Sensum 111
XX. Development of the Imperative in Latin 116
XXI. Use of the Infinitive for the Imperative 121
XXII. Ait fuisse navium celerrimus 131
XXIII. Poetic Diction 137
XXIV. Longe Lateque 161
XXV. Synecdoche and Metonymy 174
XXVI. Tellus Terra 192
XXVII. Prolepsis 204
XXVIII. Proximus ardet Ucalegon 215
XXIX. Ilicet Extemplo 225
XXX. Tollo and Puto 246
XXXI. Hendiadys 257
XXXII. The Ellipsis with Que 264
XXXIII. Synchysis or Distribution 272

Primus quisque—Iamdudum—Nequidquam

XXXV. The Amphibole 293
XXXVI. Zeugma 304
XXXVII. Hypallage 315
XXXVIII. Omission of the Prefix 330
XXXIX. Antallage 340
XL. Nisi and Donec 351
XLI. First and Last 358
XLII. The Oxymoron 386
XLIII. Quod minime reris 394
XLIV. Alia Quaedam 410




We read in Horace:

Tunc me biremis praesidio scaphae

Tutum per Aegaeos tumultus

Aura feret geminusque Pollux. (Od. 3. 29. 62-4.)

Lonsdale and Lee translate geminusque Pollux, ‘and Pollux with his twin brother’. Acron’s note is: 64. Geminusque P. Pollux cum Castore intellegendus est; amborum enim stellae simul nascuntur; and Porphyrio’s: geminusque Pollux, quia horum sidera cum se ostendunt laborantibus nautis praebent spem salutis. It is the constellation that is in question here, and the two stars rise together; so we feel sure that Pollux here is used for Castor and Pollux. But it is one thing to feel sure of this, quite another to see how this use is possible and what it involves.

Kiessling explains: ‘The one of the Dioscuri is named, the other merely signified by geminus’, and this explanation seems that commonly accepted. But if the Twins were commonly termed: gemini Castores, and we found geminus Castor used to designate both, the use of geminus would not explain the difficulty, but merely double it. In naming the Twins, besides: Castor et Pollux (Cic. N. D. 2. 6. 2), cum Castore Pollux (Hor. Ep. 2. 1. 5), gemelle Castor et gemelle Castoris (Catull. 4. 27), we find in use Castores (Tac. Hist. 2. 24 et saepius), Polluces (Serv. ad Geo. 3. 89), Gemini (Varro, R. R. 2. 1. 7), Castoras geminos (Pacat. Pan. 39). To take a parallel use, we read in Juvenal: geminos sub rupe Quirinos (11. 105), of Romulus and Remus; if we followed Horace’s idiom, we should write: geminum sub rupe Remum. But the presence of geminum in this phrase would hardly account for the use of Remum for Romulum et Remum. Horace’s idiom does not stand alone, though Kiessling cites no parallels. But we read: potaque Pollucis lympha salubris equo (Prop. 3 (4). 22. 26). In Statius we read: donec ab Elysiis prospexit sedibus alter Castor (Silv. 4. 6. 15-16), where alter Castor is also of the constellation, and must stand for Castor and Pollux; and for Virgil’s Haedi pluviales (Aen. 9. 668) we find Horace using the singular in: orientis Haedi (Od. 3. 1. 28), and Propertius in: purus et Haedus erit (2. 26. 56).


The use of Quirini for Romulus and Remus, or of Castores for Castor and Pollux, is evidently connected with the usual likeness of twins: proles indiscreta suis gratusque parentibus error (Aen. 10. 392). Virgil seems to have extended this use to brothers in: Assaracique duo et senior cum Castore Thymbris (Aen. 10. 124), where the mention of Castor immediately after Assaraci suggests the analogy of Castores, Assaraci being probably for Assaracus and Ganymedes, descendants of an older pair of brothers of like names. The likeness that is not always found in twins is very common in brothers that are not twins. Lucan extends the use further in: Heroas lacrimoso litore turres (9. 955), where Heroas evidently means ‘of Hero and Leander’.

Cicero uses Castor for Castor and Pollux in: in aede Castoris (Verr. 2. 1. 129. 49), and: ad Castoris (Milo, 91. 33); with which we may compare: ad vigilem ponendi Castora nummi (Juv. 14. 260) or: vicinum Castora canae transibis Vestae (Mart. 1. 70. 3-4). This may primarily have been merely a short and convenient way of designating the shrine, and the longer form of the name of the second twin Pollūces (= Πολυδεύκης), still in use in Plautus’s day, adds probability to this view. If so, Horace’s use of the name Pollux for the two may be merely a poetic variation of the ordinary colloquial use. But the tendency to use the name of one of a pair to express both is so common in Latin both for proper and common names, as we shall see, that I hesitate to accept this account of its origin. We may believe, however, that the Romans in their adjurations: Ecastor, Mecastor, Pol, Edepol, were using the name of one of the twins for both; and Horace seems to have both in mind in: hac arte Pollux et vagus Hercules enisus arces attigit igneas (Od. 3. 3. 9-10), when we compare: dicam et Alciden puerosque Ledae (Od. 1. 12. 25).

The use of the plural we noticed in Castores, usually called the Elliptical Plural, in Assaraci Virgil has extended from twins to a pair of brothers, and Lucan in Heroas to a pair of lovers; we might expect a similar extension of the use of geminus Pollux for Castor and Pollux. I read in Keller-Holder’s Horace:

Videre †Raetis bella sub Alpibus

Drusum gerentem Vindelici. (Od. 4. 4. 17-18.)

But all the manuscripts give Raeti, not Raetis, which is Bentley’s emendation. This reading is confirmed by the Scholia of Acron and Porphyrio, more valuable than the manuscripts, as their readings are fortified by the explanations they add. Acron’s scholium runs: 5 17. Videre Rh. b. s. A. Per hyperbaton longum sensui superiori respondit. Qualem aquilam et leonem imbellis praeda, talem Drusum videre Rhaeti Vindelici barbarorum gentes; and Porphyrio in the introduction to his Scholia on this ode writes: (Haec ode) scripta est ergo in Neronem Drusum, privignum et successorem Augusti, qui Rhaetos Vindelicos bello vicit. It seems certain that both Acron and Porphyrio had before them the text:

Videre Rhaeti bella sub Alpibus

Drusum gerentem Vindelici;

and at present editors seem inclined to accept it. Very curious is Porphyrio’s: Drusum, privignum et successorem Augusti; it looks as though he were following Horace in making Drusum stand here for Tiberium, for surely Porphyrio knew that Tiberius, not Drusus, was the successor of Augustus.

While most of the older scholars accepted the text of Acron and Porphyrio, and followed their explanation, which seemed to make of the Rhaeti and Vindelici a single people, subdued by a single victor, Bentley saw that Horace in the fourteenth ode of his fourth book stated plainly that, while Drusus subdued the Vindelici, it was Tiberius, the maior Neronum, who defeated the Rhaeti. This is confirmed by other historians of the period; and Bentley rightly stresses the words of Velleius: divisis partibus Rhaetos Vindelicosque adgressi (2. 95). The Rhaeti and the Vindelici are separate peoples, and Tiberius and Drusus were leading separate armies, when they defeated them at different places and different times. So Bentley accepts an emendation already suggested by Heinsius, Raetis for Raeti. He afterwards heard of a manuscript in the library of Peter Francius which had here the reading Retis; and this he thought, wrongly, was confirmed by the readings Retii and Reti in some of his own manuscripts. We have no further knowledge of this manuscript of Francius; and our business is, not to change the text handed down to us, but to explain as we best can the only evidence we have of what Horace wrote.

When we meet the ellipse we know in: good men and true, it does not puzzle us long. More puzzling to some readers was Horace’s: fortes creantur fortibus et bonis (Od. 4. 4. 29); some scholars wrote a comma after fortibus, and joined et bonis with the following: est in iuvencis. But Bentley saw that fortes et boni was a usual pair, and supplied: fortes (et boni) creantur fortibus et bonis, where we have a union of two pairs, one of which is represented by a single term, as we 6 saw Castor and Pollux represented by Pollux. Now in the verses in question we have a union of two pairs, the Rhaeti and Vindelici and Tiberius and Drusus, the latter of which is represented by Drusus. In the examples cited above the abbreviated pairs are designated by the same word, ‘men’ and ‘men’, boni and boni, and in v. 28 the pair thus shortened is named by the one word Nerones. And, as we shall see, this shortening of four to three is very usual in Latin and lies at the base of such usual figures as hypallage and zeugma.

Horace’s motive in this use of the figure seems plain enough. The courtiers of Augustus thought and spoke of him as the Jove on earth; we find Horace drawing this parallel in:

Caelo tonantem credidimus lovem

Regnare; praesens divus habebitur

Augustus. (Od. 3. 5. 1-3.)

If he is Jove on earth, are not his boys, Tiberius and Drusus, the Dioscuri on earth? Horace nine years before had, in speaking of the Dioscuri, named them by the name of the second, Pollux; is he not paying the boys of Augustus a subtle compliment by naming them in like fashion by the name of the younger, Drusus? On this view our difficulties disappear; from Drusum gerentem in v. 18 we infer Tiberium gerentem for v. 17, and understand: Videre Rhaeti Tiberium gerentem bella sub Alpibus, Drusum bella gerentem videre Vindelici.

We have a further extension of this use in:

Non celeres fugae

Reiectaeque retrorsum Hannibalis minae,

Non incendia Karthaginis impiae,

Eius qui domita nomen ab Africa

Lucratus rediit, clarius indicant

Laudes quam Calabrae Pierides. (Od. 4. 8. 15-20.)

Here in describing the exploits of the Scipios, Horace seems to attribute them to the younger Scipio, making him defeat Hannibal at Zama as well as burn Carthage. And yet it is only of the elder Scipio that we can understand the praises of the Calabrae Pierides; for Ennius had been dead nearly a quarter of a century when the younger Scipio destroyed Carthage with fire. It seemed to older scholars that Horace was confusing the elder with the younger Scipio, but Bentley rejected this as incredible. He threw out v. 17, which seemed to him to cause all the trouble, as it did not show the usual caesura. Just compare:

Lucratus rediit ‖ clarius indicant,


which does show this caesural pause with:

Non incendia Kar‖thaginis impiae.

True, this caesura is at times obscured by synaloepha, as in:

Reiectaeque retrors’ ‖ Hannibalis minae,

and is even found after the prefix of the verb, as in:

Dum flagrantia de‖torquet ad oscula (Od. 2. 12. 25);

and in the Greek use of this caesura there is no such regularity as Bentley assumed for the Latin. Still he concludes: Agnosco enim monachalis plane genii et coloris. But Acron had this verse before him in his text of Horace; here is his note explaining impiae: 17. impiae. Quae ter gessit bellum contra Romanos. Kiessling brackets it as spurious, and thinks it is inserted later to prevent any reader from referring: eius qui . . . rediit to Hannibal, a risk which seems remote. As we read on in the ode, we meet: clarum Tyndaridae sidus (v. 31), by which Horace probably hoped to suggest to his reader his previous use of Pollux for Castor and Pollux. (Cf. geminos Scipiadas, Aen. 6. 842.) So here again, when after reiectae . . . minae we imply eius qui . . . rediit, the difficulty disappears, and we have associated with the pair Zama and Karthago incensa the common designation eius qui . . . rediit. In the use of Pollux for the Twins (Od. 3. 29) the fourfold structure is not so evident, being obscured by the transference of scapha from the subject (scapha et) aura to the predicate feret.

In these three passages, to designate a pair connected by nature, name, qualities, or exploits, we have the name or designation of one of the pair, and that the second in each case. The first passage is of the Dioscuri, a pair usually expressed by the dual even in Doric, where the dual is rare. We have here, then, a case of the dual not passing into the plural, as has been assumed to be always the case, but into the singular. The dual has left few traces in Latin, where there was no such endeavour to restore it as is evident in post-Homeric Greek. Its earlier and speedier disappearance in Latin may well account for its passing to the singular there oftener than in Greek; for the dual that the Greeks tried to restore was a dual that was encroaching more and more on the domain of the plural. If the dual primarily indicates two objects paired in nature, such as the two eyes or the two hands, as is usually believed, it may be asked whether it is nearer the plural than the singular. I read in Gauthiot, Du nombre duel: En Indo-Européen Skr. akṣī, Gr. ὄσσε, lit. akì, ne signifient pas proprement ‘les deux yeux’, ni ‘la paire d’yeux’, ni même ‘l’œil et l’autre œil’, mais 8 ‘l’œil autant que double’ (Festschrift für Vilhelm Thomsen, p. 131). We know how in Homer ὄσσε is repeatedly joined with a singular verb, though not so often as with one in the dual or plural. It is a neuter dual; and this use has been held to be parallel with that of the neuter plural with a singular verb; but it is no true parallel. If it were, the use of ὄσσε with the singular verb would be far more usual than its use with the dual or plural.

The dual certainly passes into the plural far oftener than into the singular; and the syntax of ὄσσε supports this. Our English use of ‘to bow the knee’ points to the same syntax, as does Marlowe’s verse: Than has the white breasts of the Queen of Love (Faust, 1. 1. 132); and when Horace writes:

Gaude quod spectant oculi te mille loquentem (Ep. 1. 6. 19),

does he mean that only five hundred constitute the audience? If he means a thousand, then oculus must mean to him a pair of eyes. So too in: in laxa nixa pedem solea (Prop. 2. 29. 40) we have both pedem and solea singular, but with dual force. We have a picture of this syntax of the pair in: clarum Tyndaridae sidus (Od. 4. 8. 31), representing it as a plural enclosed in a singular; and a similar one in: geminique sub ubere nati (Aen. 5. 285), where the dual has passed to the plural in nati, but to the singular in ubere; and when to: includunt caeco lateri (Aen. 2. 19) Servius notes: caeco lateri pro caecis lateribus, he is noting this use of a singular for a plural which is for an old dual. We have the opposite use of a plural, which is an old dual, for a singular in: Atridas Priamumque et saevum ambobus Achillem (Aen. i. 458), where ambobus can hardly be for three, and the union of Atridas with Priamum points to its use for Agamemnonem. When did Achilles condescend to reproach Menelaus?



Virgil in his third Georgic, after a fine description of the horse, begins to illustrate it with examples thus:

Talis Amyclaei domitus Pollucis habenis

Cyllarus (vv. 89-90).

The reader will notice that here we have Pollux in a rôle not usually assigned to him. Horace tells us:

Castor gaudet equis, ovo prognatus eodem

Pugnis (Sat. 2. 1. 26),

following Homer’s:

Κάστορά θ’ ἱππόδαμον καὶ πὺξ ἀγαθὸν Πολυδεύκεα (Il. 3. 237).

Servius’s note is: Atqui Castor equorum domitor fuit. Sed fratrem pro fratre posuit poetica licentia, ut quas illi Philomela dapes pro Progne, item revocato a sanguine Teucri pro Dardani; aut certe ideo Pollucem pro Castore posuit, quia ambo licenter et Polluces et Castores vocantur; nam et ludi et templum et stellae Castorum nominantur. The last argument might account for the use of Castor for Pollux, it hardly accounts for that of Pollux for Castor; while his fratrem pro fratre reminds me of the question of our boyhood, ‘Who killed Cain?’

But we have a way in English of expressing briefly a series of four or five by the first and last. Murray tells me: ‘First and last: all, “one and all”.’ Is this also Latin idiom? Horace writes:

Primosque et extremos metendo

Stravit humum sine clade victor. (Od. 4. 14. 31-2.)

In the verses:

album mutor in alitem

Superne, nascunturque leves

Per digitos humerosque plumae (Od. 2. 20. 10-12)

in: digitos humerosque Horace gives a series of four by the first and last; and again in:

Caementis licet occupes

Tyrrhenum omne tuis et mare Ponticum (Od. 3. 24. 3-4)

he indicates the whole system of the Mediterranean seas by its eastern and western extremes. So Catullus in:

Unam Septimius misellus Acmen

Mavult quam Syrias Britanniasque (45. 21-2)

signifies in like fashion all the provinces of the Empire. We saw in 10 the last chapter how Horace indicated a pair by one of its members. Here we have two pairs indicated by the first and last of their members.

Let us set down the pairs in order, beginning with the horses, as the passage is about horses. Of the two horses, Cyllarus and Xanthus, Cyllarus is the horse of Castor. Valerius Flaccus tells us:

Vectorem pavidae Castor dum quaereret Helles

Passus Amyclaea pinguescere Cyllaron herba (1. 425-6).

So we must order the pairs: Cyllarus—Xanthus—Castor—Pollux. Virgil here uses Cyllarus for the pair Cyllarus—Xanthus, and Pollux for the pair Castor—Pollux, but the secret of this figure was lost by Servius’s day, and has not been recovered till now. Acron’s note to: hunc equis illum superare pugnis nobilem (Od. 1. 12. 26) is: amatorem equorum Castorem dicit ut est Virg.: talis Amyclaei domitus Pollucis habenis Cyllarus. The only way we can account for his use of this verse to prove that Castor, not Pollux, was the amator equorum is to assume that Acron felt that all his readers, like himself, would at once see that Pollucis was here used by metonymy for Castoris.

Let us turn to Servius’s Procne—Philomela example:

Aut ut mutatos Terei narraverit artus,

Quas illi Philomela dapes, quae dona pararit,

Quo cursu deserta petiverit, et quibus ante

Infelix sua tecta super volitaverit alis. (Buc. 6. 78-81.)

Servius’s note is: Philomela dapes: atqui hoc Procne fecit; sed aut abutitur nomine, aut illi imputat propter quam factum est. But in his note to v. 78 we read: Omnes in aves mutati sunt: Tereus in upupam, Itys in fassam, Procne in hirundinem, Philomela in lusciniam. Take his order of the dramatis personae; Virgil names them here by the first and last, but lest we might not understand that Philomela here is for Procne—Philomela, he appends two descriptions, one of the nightingale’s dwelling: quo cursu deserta petiverit (poetic for: quo volatu silvas petiverit) and: quibus ante . . . alis for that of the swallow. With this specimen of Virgil’s art we may compare his description of Romulus in the sixth Aeneid:

Viden ut geminae stant vertice cristae? (v. 779)

to which Servius’s note is: omnino in omnibus hoc egit Romulus, ut cum fratre regnare videretur, ne se reum parricidii indicaret: unde omnia duplicia habuit, quasi cum fratre communia.

Unde omnia duplicia habuit. Nowhere does this seem truer of Virgil, or more striking than in the example I follow Servius in citing:


Quid loquar aut Scyllam Nisi, quam fama secuta est

Candida succinctam latrantibus inguina monstris

Dulichias vexasse rates et gurgite in alto

Ah! timidos nautas canibus lacerasse marinis? (Buc. 6. 74-77.)

Here we have the exploits of Scylla, the daughter of Phorcus, apparently attributed to Scylla, the daughter of Nisus. We can again construct the figure: Scylla—Scylla—Phorci—Nisi, but it differs from the last example in that none of the description given fits Scylla Nisi. It looks to me like a very strong case of ἀπροσδόκητον; we can hardly assume a blunder on the part of Virgil, who had already set forth at great length in the Ciris the crime of the daughter of Nisus, and in Geo. 1. 406-9, repeats the last four verses of that poem. What he seems to mean is that Silenus told the tale of both Scyllas, but he specifies only the doings of the Scylla whose story he had not already told. But Propertius (4. 4. 39-40) confuses the two Scyllas, as Virgil seems to do here.

Easier to deal with is Servius’s question (ad Aen. 1. 235), why Virgil always calls Dido Sidonia Dido, though she was from Tyre, not Sidon, and has no direct connexion with Sidon. Of course there was the indirect connexion, in that Tyre came from Sidon; it was settled by the Sidonians. This we could state in our fourfold form: Sidonia—Tyros—Tyria—Dido; by using the more remote epithet got by this figure Virgil calls forth in his reader’s mind the origin both of Dido and of Tyre; and this is evidently one of the constant aims of poetic diction. The aim of prose is to state with all possible clearness and elegance the fact you wish to narrate, that of poetry to call forth in your narration by suggestion all the facts associated with the fact in question; and hence the use of a figure like hypallage.

To turn now to the remaining example cited by Servius, where Virgil seems to have gone wrong in his mythology:

Certe hinc Romanos olim volventibus annis,

Hinc fore ductores revocato a sanguine Teucri,

Qui mare, qui terras omni dicione tenerent,

Pollicitus. (Aen. 1. 234-7.)

Servius’s note is: Teucrum pro Dardano posuit: Dardanus enim de Italia profectus est, Teucer de Creta: quia solent poetae nomina de vicinis provinciis vel personis usurpare. But Italia and Creta are hardly neighbouring states like Tyre and Sidon; and one is rather led to think of the confusion between Teucer and Dardanus in the mind of Anchises, that led to the settlement of the Aeneadae in Crete. But our fourfold arrangement gives Romani—Cretenses—Dardanus—Teucer; 12 and Virgil’s use of Romani and Teucer here may be another use of this figure. But Servius’s note to revocato here leads further: dicendo revocato, ostendit Italiam, unde Dardanus fuerat. But by naming Teucer instead of Dardanus he straightway contradicts this; for how can the descendants of Teucer be called back to a country from which Teucer did not come? May not Virgil intend us to take re- in another sense?

Brugmann relates red-, the older form, to vret, which contains vr, a reduced form of the root ver that is found in vermis ‘the wriggler’, and in verto. Re- has primarily to do with turning back or repeating an action. But it is often used to emphasize a notion of iteration already expressed in the verb as in renovo, constantly in use for novo, ‘I renew’. So in:

Sic pater Aeneas intentis omnibus unus

Fata renarrabat divum cursusque docebat (Aen. 3. 716-17)

renarrabat is simply: dictis iterata narrabat. In:

Vergilium finibus Atticis reddas incolumem precor (Od. 1. 3. 6-7)

Virgil is not returning to Attica, but Attica is his proper and purposed destination. Most interesting here is recipere, the correlative of reddere, which is in use throughout Latin letters from Plautus down, not only for ‘to take back’, but also for ‘to take as one’s own’. It is well distinguished from accipio in: (Peneus) accipit amnem Horcum, nec recipit, sed olei modo super natantem brevi spatio portatum abdicat, poenales aquas dirisque genitas argenteis suis misceri recusans (Plin. 4. 8 (15). 31). So in: Dissolve frigus ligna super foco large reponens (Hor. Od. 1. 9. 6-7), reponens seems to convey the idea that the hearth in winter is the proper place for the firewood; just as the right place for a book which I take down from my friend’s shelves is the place from which I took it.

In Aen. 1. 235, then, Virgil by substituting Teucri for Dardani indicates to his thoughtful reader that revocato is not to be taken here in its usual sense of ‘called back’, for Teucer had never been in Italy. The word revocato suggests to the reader that Dardanus the founder of the city of Troy was from Italy, and so it was natural that on the fall of Troy the Trojans should return to Italy, their old home. He reads on, and is surprised to find Teucer, and not Dardanus, connected with revocato; and his surprise at this serves to impress more strongly on his mind the idea which Virgil tries to convey throughout the Aeneid, that this return to Italy is a right and due recall, a recall by the gods. We have here one of the many cases, which call for 13 a fuller treatment than they have yet received, where Virgil joins to an obvious meaning a second and less obvious, but one often marked by greater majesty and sublimity. He often uses the figure I have illustrated in this chapter for this purpose, and when I show that this expression of two related pairs by one member of each, as well as the more general expression of four objects by two, the first and last named, is not confined to proper names, but is often used with common nouns and adjectives, as well as verbs, I hope to convince my reader that we have here a figure that was readily comprehended by Virgil’s readers in his day, but the secret of which was lost by the fifth century, the period of Servius, Acron, and Porphyrio. We find examples of it in English; in Bishop Bickersteth’s well-known hymn, at the beginning of the last stanza:

See the feast of love is spread;

Drink the wine and break the bread.

while bread and wine give us a symmetrical pair, this is not true of break and drink, which are short for: break and eat, fill and drink; a use of two for four. For this alternate ellipsis in two connected pairs, a figure not named till now, I would suggest the name of Antallage, a term closely related to the names of the allied figures of Enallage and Hypallage, and not hitherto in use for grammar to my knowledge.

There seems another example of this figure in the Aeneid not mentioned by Servius in this connexion, where Virgil seems to contradict not merely current tradition, but the account he has himself given a little before. Hyginus notes (Gell. 10. 16. 12-13) that, though Virgil had named Theseus among those who had descended to the shades and returned to earth (qui ad inferos adissent et redissent), and had said:

Quid Thesea magnum,

Quid memorem Alciden? Et mi genus ab love summo est (Aen. 6. 122-3),

yet afterwards he inserts in his poem:

Sedet aeternumque sedebit

Infelix Theseus. (Aen. 6. 617-18.)

‘How is it possible’, asks Hyginus, ‘that he should sit forever in the shades, whom Virgil has already named with those who have descended thither and again escaped thence? especially when such is the tradition about Theseus, and if Hercules himself plucked him from the rock and led him forth to the upper air?’ His question 14 calls for a franker and better answer than it has yet received from commentators whether ancient or modern. Servius thinks that, though Hercules did deliver Theseus, still: eum ita abstraxit, ut illic corporis eius relinqueret partem (ad v. 617). Sidgwick thinks that in v. 122 Virgil represents Theseus as freed by Hercules, but that in v. 618 he has adopted a variant account. Most of our editors, like Conington, follow Heyne, who apparently fails to catch Hyginus’s meaning, and adopts Servius’s treatment of v. 122, where he speaks of Theseus as a durum exemplum, passed over quickly by Aeneas as not having ascended like Castor, Pollux, and Hercules; to render this more probable, he is for dividing magnum from Thesea and joining it with the following Alciden. But Conington rightly refuses to make this division, and compares Thesea magnum with Cissea durum (Aen. 10. 317). It seems to me that in: quid Thesea magnum, quid memorem Alciden? we have an easy poetic distribution for: quid memorem Thesea magnum? quid memorem Alciden magnum?, and that Servius’s attempted evasion of the difficulty involves a departure from the plain sense of the passage.

No doubt Virgil had in mind two versions of the legend: that implied in Homer (Od. 11. 630 ff.), where we are told that, if Ulysses had remained longer in Hades, he would have further seen men of the foretime, Theseus and Pirithous, renowned children of the gods; and that given us by Apollodorus (2. 5. 12), and followed by Horace (Od. 4. 7. 27), where Hercules succeeds in rescuing Theseus, but fails to reach Pirithous. The latter is the account commonly received in Virgil’s day, and plainly accepted by him in v. 122, though in v. 618 he seems at first sight to follow Homer. A century or so after we find that Plutarch does not take Theseus to Hades with Pirithous; in return for Pirithous’s help in his abduction of Helen from Sparta he joins his friend in attempting to carry off Kore, the daughter of Aïdoneus and Persephone, sovereigns of the Molossi (Th. 31). In this attempt Pirithous was gobbled up (ἠφάνισε) by the king’s dog, Cerberus, but Theseus escaped to be pushed off the rocks of Scyros by Lycomedes a little later (Th. 35). But though dead, he had not forgotten his beloved Athens; and in the charge and rout of Marathon, he returned to sustain his Athenians—an older Angel of Mons.

But did Virgil really show himself guilty of so flagrant a contradiction of himself in this second account, where to the cursory reader he seems to turn back to Homer’s story? Let us examine more closely the passage from which Hyginus cited the apparent contradiction:


Saxum ingens volvunt alii, radiisque rotarum

Districti pendent; sedet aeternumque sedebit

Infelix Theseus, Phlegyasque miserrimus omnes

Admonet et magna testatur voce per umbras:

‘Discite iustitiam moniti et non temnere divos.’ (Aen. 6. 616-20).

We have seen how Virgil seems to use Pollux for Castor (Geo. 3. 89), Philomela for Procne (Buc. 6. 79), Teucer for Dardanus (Aen. 1. 235); how he really uses Scylla Nisi for Scylla Phorci (Buc. 6. 74), and Sidonia for Tyria as the standing epithet of Dido. May he not be using Theseus here for Pirithous? But in all the cases enumerated two pairs were involved, of which the first and last terms alone were expressed; here Theseus and Pirithous are plainly involved, but where is our second pair? Very close to the rock of Theseus here Virgil has set the wheel of Ixion; and according to the usual tradition Phlegyas was Ixion’s father, and Pirithous Ixion’s son. Both were whelmed in Tartarus, Ixion for his attempted rape of Juno, Phlegyas for burning the shrine of Apollo at Delphi in revenge for Apollo’s seduction of his daughter Coronis. Most appropriate to both seems the second clause of Phlegyas’s discourse: et non temnere divos. But where does Virgil get the first clause: discite iustitiam? Heyne cites Pindar (Pyth. 2. 39 ff.): ‘They say that Ixion under injunctions from the gods proclaims this to mortals, as he is rolled on his winged wheel: Him that doeth good service draw near and recompense with fair reward.’ So that while the second part of Phlegyas’s speech here seems as appropriate to Ixion as to his father, the first clause is given by Pindar to Ixion, and transferred by Virgil to Phlegyas. The magna voce has occasioned surprise, and our editors compare with it the vox exigua proper to the shades (v. 493); but torments, like actions, speak louder than words, and magna voce is clearly a metaphor. Virgil, in his anxiety to give his reader the clue to his meaning, is not content with bringing the wheel of Ixion as close as he can to the rock of Theseus; looking back you will see that the last pair of proper names before the verses we have cited are those of the missing pair, Ixiona Pirithoumque (v. 601). We seem, then, to be following Virgil’s express indications in taking Theseus Phlegyasque as short for Theseus et Pirithous, Ixion Phlegyasque. While Phlegyas is for Ixion Phlegyasque, an example of synecdoche, Theseus is used for Pirithous, an example of metonymy, both of which figures arise out of the use of one for a pair. So in Cyllarus (Geo. 3. 90) we have synedoche; the steeds are of like merit.



In Dr. Rendel Harris’s Cult of the Heavenly Twins, he tells us of a pair of Italic deities, Picumnus and Pilumnus, the Castor and Pollux of Italy. I recalled Pilumnus in the Aeneid, but had no recollection of his association there with Picumnus. But if Virgil used the name Pilumnus for the pair, as Horace did Pollux, I felt sure that I should at once find the pair named in Servius’s scholium to the verse. So I turned to the verse where he first names Pilumnus:

Luco tum forte parentis

Pilumni Turnus sacrata valle sedebat. (Aen. 9. 3-4.)

If Picumnus and Pilumnus are the Italian Castor and Pollux, then a grove sacred to Pilumnus will be sacred also to Picumnus, and Servius’s note runs: Parentis Pilumni. Pilumnus et Pitumnus fratres fuerunt dii. Horum Pitumnus usum stercorandorum invenit agrorum—unde et Sterculinius dictus est—, Pilumnus vero pinsendi frumenti: unde et a pistoribus deus colitur. Quidam Pilumnum et Pitumnum Castorem et Pollucem accipiunt: nonnulli laudum deos: Varro coniugales deos suspicatur.

But Servius calls the brother of Pilumnus Pitumnus, and not Picumnus. Virgil knows Picus, but he knows nothing of Picumnus, and the only source of information about Picumnus of which I know is Nonius Marcellus’s Compendiosa Doctrina (p. 518, M.): Picumnus et avis est Marti dicata, quam picum vel picam vocant, et deus qui sacris Romanis adhibetur. Virgil’s name for the bird and god in question was certainly Picus, of which Picumnus may have been a later and corrupt form. And of Nonius’s citations four identify Picumnus with Picus, and only one: idem Iuris Pontificii Lib. III; Pilumno et Picumno, associates him with Pilumnus; and even here Lucian Müller tells me that the codices give Picum and that Picumno is Bentley’s emendation. Bentley is following a second note of Nonius (p. 528, M.): Piluminus et Picuminus di praesides auspiciis coniugalibus deputantur; and on this note seem to rest the corrections by Roscher, Wissowa, and Aust of Pitumnus in Servius to Picumnus. Ettore Pais retains Servius’s spelling, rightly as it seems to me, for it is 17 repeated four times in his commentary, a work far freer from blunders than Nonius’s Doctrina. Of all the authorities Leonhard Schmitz cites in his article on Picumnus in Sir Wm. Smith’s Dictionary of Mythology not one gives us the name Picumnus; and the article hardly seems an honour to English scholarship.

Can we derive the name Pitumnus? Has it any connexion with the function assigned him by Servius, the usus agros stercorandi? Servius tells us that the pilum or pestle, used in early times to pound the grain to prepare it for cooking, took its name from Pilumnus; but it was the other way, as we know. Pilumnus got his name from pilum. And Pitumnus gets a second name, Sterculinius from stercus ‘dung’. Nor is it hard to see how the name Pitumnus is derived. Take the Greek ψίθυρος ‘false’; it is a dissimilation of ψύθυρος, from a reduced form of the root we have in ψεύδω. So Pitumnus is plainly a dissimilation for Putumnus ‘the stinking god’, from the root we see in pŭter, of which we have a strong form or guna in pūtidus. We need not wonder that Virgil had no room in his Epic for such a name; we might as well expect to find the verb ‘to puke’ in the Paradise Lost.

But we see from Nonius that the name Picumnus was used for Picus, the bird and the god. It is probable that this name was connected with Pilumnus, as appears from Nonius’s second note. Some related that Danae, after landing in Italy, married Pilumnus, and bore him a son called after her Daunus, the father of Turnus. So that we have the following genealogies:


Picus       —Faunus—Latinus.

The association of Latinus and Turnus in Italian legend leads to an association of Daunus and Faunus, whence a rhyme association arises, which extends to Pilumnus and Picus, forming from Picus the second name Picumnus. That Picumnus, the corruption of Picus thus produced, was later confused at times with Pitumnus, seems not unlikely.

We meet Pilumnus again in:

Cui Pilumnus avus, cui diva Venilia mater (Aen. 10. 76);

and again in:

Pilumnusque illi quartus pater (10. 619),

where the quartus puzzles Servius, who thinks of this Pilumnus as the avus of the Pilumnus mentioned in v. 76. But it seems more likely that both Virgil’s avus and his quartus pater refer to the same 18 Pilumnus, who is, strictly speaking, the tertius pater, according to the usual form of the legend. In:

Poscit equos gaudetque tuens ante ora frementes,

Pilumno quos ipsa decus dedit Orithyia,

Qui candore nives anteirent, cursibus auras (Aen. 12. 82-4)

the analogy of the white steeds of Castor and Pollux leads us to believe that Pilumno stands for Pilumno et Pitumno. Of course, as horses live only about thirty years, these could not be the same horses that Orithyia gave Pilumnus, but were descendants of them. Servius thinks the story a figment. How could Orithyia, born in Attica, and carried off to Thrace by Boreas, give horses to Pilumnus in Italy? In his note to Aen. 7. 410 he tells us that Danae came to Italy alone; in the note to Aen. 8. 345 we read that she came to Italy accompanied by two sons, whom she had of Phineus, Argos and Argeos, and held the place which is now Rome. In some families it is not so much the sexus as the genus that tends to variation. In Minos’s line the taurus plays an important part; think of the nuptials of Europa, and the incest of Pasiphae. In the household of Boreas and his son-in-law Phineus this seems true of the horse. In the Iliad we read of the three thousand mares of Erichthonius, with whom Boreas falls in love as they graze and:

αἱ δ’ ὑποκυσάμεναι ἔτεκον δυοκαίδεκα πώλους (20. 225).

Possibly Argos and Argeos, ‘white’ and ‘whitey’, whom Danae brought with her from Phineus, were the white horses sent by Orithyia. Servius identifies this Argos with the Argos whose death gave its name to the Argiletum according to some.



That geminus Pollux could be used for gemini Castores or Polluces will seem easier to believe, when we see how liberi is used in Latin for a single son or daughter, and how far this use has been extended by analogy. We have here the opposite of the use of one for two, an opposite which we find as a rule in case of any abnormal construction. We read in the third Aeneid:

Idem (nos) venturos tollemus in astra nepotes

Imperiumque urbi dabimus. (vv. 158-9.)

Servius’s note to this is: In astra nepotes significat Gaium Iulium Caesarem, qui primus inter deos relatus est. He evidently refers nepotes to Caesar as founder of the Empire; otherwise he would have taken note of Romulus.

The construction plainly starts with liberi, as we find it in: ex Miseno autem eius ipsius liberos . . . a praedonibus esse sublatos (Cic. Leg. Man. 33. 12). Plutarch tells us that it was a daughter of Antonius that was carried off (Pomp. 24). So in: occisus est cum liberis M. Fulvius consularis (Cic. Cat. 1. 4. 2), and: cum Ameriae Sex. Rosci domus, uxor, liberique essent (Pro Rosc. Am. 96. 34), liberi is for a single son; and in: coniugem et liberos meos (Tac. Ann. 1. 42. 2) Caligula is the only child of Germanicus in the camp. In: et Plancinam et filios variis modis criminari (Ann. 2. 57. 3) M. Piso is the only son with his father in Syria. Here it is extended from liberi to filii, as in: vagamur egentes cum coniugibus et liberis (Cic. Att. 8. 2. 3), where coniugibus (i.e. Terentia) is evidently attracted to the plural by the following liberis; as is parentibus in: quod me parentibus (i.e. Antonia), liberis, patriae intra iuventam praematuro exitio raperent (Tac. Ann. 2. 71. 2).

In: coniugiumque domumque patres natosque videbit (Aen. 2. 579) we remember that Helen had only one child, her daughter Hermione. About patres there seems more uncertainty among scholars; it may be for Tyndareus alone or for Tyndareus and a second consort; and so a plural for a dual, like soceros (v. 457). But it seems most natural to take patres natosque here as a concrete expression for ‘the delights 20 of home’, to put it more familiarly, ‘papa and the children’; in which case patres is for Menelaus and is attracted by natos like parentibus in the last example. Coniugium and patres, the first words of each colon, should have a common reference. To this class of plurals will belong amores in: Acmen Septimius suos amores (Catull. 45. 1), Pompeius nostri amores (Cic. Att. 2. 19. 2), non ille oblitus amorum (Aen. 5. 334). Draeger has referred this syntax to the general or typical use of the plural, as we find it in: et maribus Curiis et decantata Camillis (Hor. Ep. 1. 1. 64) ‘chanted by manly Romans of old like Curius and Camillus’. But all cases we have cited thus far are plainly of individuals, and not typical. It is easy to see how from such cases we pass to typical uses; from:

Qui illum decrerunt dignum suos quoi liberos committerent (Ter. Hec. 212),

where one child is vaguely referred to, it is easy to pass to:

Domus, uxor, liberi inventi invito patre. (And. 891.)

So in:

mussat rex ipse Latinus

Quos generos vocet aut quae sese ad foedera flectat (Aen. 12. 657-8)

Virgil transfers the plural generi from the leaders to their peoples, a transference easy in case of the Aeneadae, who are viewed as the family of Aeneas. We have a further step in: aut ensem . . . ereptum reddi Turno, et vim crescere victis (Aen. 12. 799), where the plural is not now a collective, but an abstract designating the class to which Turnus belongs. This is the natural course in language; from the concrete we get the collective, and from this the abstract or general use, but by far the greater number of uses like liberi for a single son or daughter are concrete and of individuals.

That liberi is found only in the plural is no sufficient answer to our problem; the singular of liberi is in use in Liber Bacchus and in Libera Proserpina (cf. Κόρη Περσεφόνη). The real solution may be that given by Gaius: non est sine liberis, cui vel unus filius unave filia est (de Verb. Signif. 1, cited by Gesner). So by a facile popular logic we have liberi used for either unus filius or for una filia, as we have seen.





Numbers are three, Dionysius Thrax tells us, Singular (ἑνικός), Dual (δυϊκός), Plural (πληθυντικός). But there are singular expressions used also for many, e.g. δῆμος, χορός, ὄχλος; and plurals used both for singulars and duals; for singulars, e.g. Ἀθῆναι, Θῆβαι; and for duals, e.g. ἀμφότεροι (p. 30, Uhlig). Such is the statement about numbers in our oldest grammar; but it is far from being full or exhaustive. Let us state the matter a little more fully and in detail.

(1) The singular is used for the plural. Dionysius’s examples are all nouns of multitude, where no one would dream of counting the units, or treating them as individuals, e.g. ψάμμος or ψάμαθος ‘sand’. Here the Greek usually has either the singular or the plural with the same meaning; e.g. λαός or λαοί; while Latin uses the singular populus. In English we use now the singular, now the plural, saying ‘wheat’ and ‘barley’, but ‘oats’ and ‘pease’. When Virgil ventured in imitation of ἄλφιτα to write hordea ‘barleys’, he met with Bavius’s reproach:

Hordea qui dixit, restat ut tritica dicat,

a reproach fully justified by Latin idiom.

(2) The plural is used for the singular. Dionysius’s examples for Greek are names of cities, like Athens or Thebes, and Latin gives us similar examples in Veii, Gabii, Corioli. The same is true of festivals held on a single day like Θεσμοφόρια, Διονύσια, Ἐλευσίνια, Saturnalia, Kalendae, Idus.

(3) The plural is used for the dual; indeed Homer uses the plural δοιοί instead of δύω (Il. 5. 206), and in the first Iliad Achilles greets the two heralds: χαίρετε, κήρυκες (v. 334). After Aristotle the dual passes out of use in Greek, and for a pair the plural is regularly used. In Latin duo and ambo are the only two duals recognized by the Romans, and for the remaining cases of these words plural forms are used, like ambos or duobus. Speaking generally, when the dual goes out of use the plural takes its place.


(4) The dual is used for the plural. In Greek for several pairs we often find the dual used, and not the plural, as in: οἱ δ’ ἅμα πάντες ἐφ’ ἵπποιϊν μάστιγας ἄειραν (Il. 23. 362) ‘and they all with one accord raised their whips upon their teams’. Dionysius does not proceed to give us this use, probably from a natural piety. Zenodotus, the founder of the Alexandrian school of grammarians, maintained that Homer used the dual for the plural; but Aristarchus of Samothrace, the teacher of Dionysius, wrote in refutation of this a treatise now lost. No doubt he tried to show that in all examples of this use cited by Zenodotus there was an idea of pairing involved that justified the use of the dual. But the syntax of ἁλόντε with γένησθε in:

μήπως, ὡς ἀψῖσι λίνου ἁλόντε πανάγρου

ἀνδράσι δυσμενέεσσιν ἕλωρ καὶ κύρμα γένησθε (Il. 5. 487-8)

seems to involve no such idea of pairing. In Latin while octo and viginti, old duals, are used as plurals, this idea of pairs is primarily involved, though no longer felt. But equae ‘the mares’, primarily a dual (= Skt. açve), is used for any number of mares.

(5) We have examined in the previous chapters a number of cases where in Latin the singular is evidently used for the dual. This use seems rarer in Greek; but where in Skt. we have nāsā ‘the nose’, a duale tantum, in Greek we find ῥίς or ῥῖνες, a singular or a plural.

(6) The dual is used for the singular, but not in Greek or Latin. But in Latin we find plurals used for older duals in use for the singular. In Sanskrit we find two duals used for two related singulars in: Mitrā Varunā. So in: Lugete o Veneres Cupidinesque (Catull. 3. 1) we have two plurals used similarly for two related singulars.

I might proceed to ask how far we have reason to believe that the dual number is coeval with the singular and the plural. The dual seems to have existed in all branches of the Indo-Germanic family of languages; and so we have a right to consider it an inflexion of the primitive Indo-Germanic. But probably it was much later and more imperfect in its development than the singular and the plural. The inflexions for the dual of nouns, pronouns, and verbs seem to show from their lack of development that the dual was later, and was never in regular use even for pairs. Its use is most fully developed and most regular in Sanskrit and Gothic, both artificial or book languages as we know them.



That the dual existed in Latin, is recognized by Donatus and his school. In Donatus’s Ars we read: Est et dualis numerus, qui singulariter enuntiari non potest, ut hi ambo, hi duo (IV. 376. 23, K.). Servius, in his commentary to Donatus’s Ars, adds that this is why they are irregular in declension; they have dual forms for the nominative, and plural for the oblique cases (IV. 408. 17, K.). To duo and ambo Sergius adds uter and neuter (IV. 540. 7, K.). Later anonymous grammarians under the influence of philosophy seem disinclined to accept the dual: quia non est in natura rerum hic tertius numerus (Ars Anon. Bern., Suppl. 84. 18, K.); sed hunc non recipimus, quia, qui singularitatem excedit, in pluralitate deprenditur (Comm. Einsidl. in Don. Artem, Suppl. 240. 14, K.).

The earlier Roman grammarians like Donatus were fortunately free from this influence, and were determined in their view by the form of duo and ambo; for duo corresponds exactly to the Greek δύο, and ambo, mutatis mutandis, to ἄμφω. Duo is shortened from the older duō by the law of brevis brevians, the same that gives us benĕ and malĕ. Porphyrio’s note to Hor. Sat. 2. 3. 248 is: ludere par impar, uni dui (Cod. Med.); so in the fifth century of our era duo seems to have developed a plural dui, just as δύο developed δοιοί. The φ in ἄμφω is for an older bh (cf. Skt. ubhau); the Roman, who could not aspirate as did the Greek, and down to 150 B.C. wrote Corintus as he pronounced it, for bh wrote b in ambo.

But the Latin grammarians had a further motive for emphasizing the fact that they found a dual in Latin. Latin grammatical studies, Varro tells us, begin in 157 B.C., when Crates of Mallos, sent to Rome as ambassador from King Attalus, while taking a walk on the Palatine, fell and broke his leg. During the inactivity consequent on his accident, he found time to give some attention to the Latin language, which he decided was a depraved derivative of Greek. The circle of the Scipios and the Aemilii eagerly adopted the notion that associated their language with Greek; and grammatical studies 26 of this tendency came into fashion. But a century later Romans were no longer so disposed to accept this view; and it was to refute one Hypsicrates, who wrote from this standpoint, that Varro composed his De Lingua Latina. That Latin is an independent language, he maintains, and points to the number of Latin cases, one more than in Greek, to prove his point. All the older Latin grammarians follow him in this endeavour to show that Latin is at least as rich as Greek in grammatical inflexions. The two forms ambo and duo seem to some of them a narrow basis for their claim of a Latin dual; so we read in Cledonius: et communis est numerus, qui et dualis dicitur apud Graecos, ut species, facies, res (V. 10. 19, K.). In assigning genders, when a word had the same form for the male and the female, as ἵππος or homo, the Greeks and Romans agreed that its gender was ‘common’. Cledonius finds that species, facies, res have the same form for the singular and the plural, and so sets up a ‘common number’, equivalent to the Greek dual, he tells us. In this he seems to have found no following.

But the Greek verb also has forms for the dual distinct from the plural forms; and so some Romans claimed that forms like legere, fecere, conticuere were duals (vide Cledonius, V. 60. 6, K.). All Latin usage is against this; Donatus denies that they are duals, and Macrobius, to prove they are not, cites Virgil’s conticuere omnes (Aen. 2. 1), and una omnes fecere pedem (5. 830). In the Commentum Einsidlense (Suppl. 256. 3, K.) we find legēre confused with the infinitive legĕre: dicimus enim legere volo et legere volumus in singulari numero et in plurali;—another attempt to establish a common number as a dual. Legēre seems the old and genuine form of the third plural perfect, which later, on analogy of the present, became legērunt.

But our great Latin grammarians like Vossius and Ruddiman make no mention of a Latin dual. They seem to have held the grammar of Priscian to be of higher authority than that of Donatus; and Priscian knows nothing of a Latin dual; nor do his sources, Charisius and Diomedes. In A.D. 327 Constantine removed the seat of empire from Rome to Byzantium; and at once the Greeks, who have henceforth to administer Latin law, feel the need of a knowledge of the Latin language. To satisfy this the great grammars of Charisius and Diomedes are composed; and as both are intended to teach Greeks Latin, both state with emphasis that Latin, unlike Greek, has no dual; and Priscian, who like them composed his great work in Constantinople 27 about 150 years later, follows them in this. But Donatus and the older grammarians were right in their claim that there was a Latin dual, and that traces of this number and of forms arising from its presence still exist in Latin.

We may review briefly these traces. Like ambo and duo, octo is the Greek ὀκτώ, the Sanskrit aṣṭau, which Fick thought meant primarily ‘the two points’, i.e. the two hands held out with the thumbs folded into the palms. So viginti is the two tens, like the Greek εἴκοσι (old ϝίκατι) and the Sanskrit viṅçati; and in all three the ending ī is the regular ending of the neuter dual in Sanskrit. Wilamowitz thinks that in the inscription: M. C. Pomplio No. f. dedron Hercole, Pomplio is a dual, and his view is favoured by Leo (Pl. Forsch. 333); and Schulze thinks that in: Q. K. Cestio Q. f. Hercole donu dedero (C.I.L. 14. 2891) Cestio is a dual.

The ending o found in ambo, octo, Pomplio, Cestio, seems the ending in ἵππω and in the Skt. açvau, where the u seems a reduction of the vi- in viginti and means ‘two’. It is the root of the Skt. vidya ‘knowledge’ (cf. scio and descisco), and of the old verb vido found in the compound divide. When we read in Horace:

Hoc iter ignavi divisimus, altius ac nos

Praecinctis unum (Sat. 1. 5. 5-6),

the translation of divisimus induced by its opposition to unum, ‘we made two of’, seems justified by its derivation.

Sommer (Lat. Laut- u. Formenlehre, 424) explains ū in neuters of the fourth declension as got by analogy from the ū in genū and cornu, which are old duals like the Skt. sūnū ‘the two sons’. Cornū = die beiden Hörner = das Gehörn = das Horn. So in English, speaking of the two knees, we use the phrase ‘to bow the knee’. He thinks genūs is for an older genuos, where the ending os is that of the gen.-abl. dual in Skt. Brugmann is inclined to agree that genūs is the old genitive dual, but refuses to accept genu as a nominative dual till he has more evidence that sūnū is an Indo-Germanic form. It is interesting to meet this fresh case of a dual passing into the singular.

But the great majority of Latin forms once dual, if we follow Brugmann, now appear as plurals. Equae is the exact equivalent phonetically of the Skt. açve, and meant primarily ‘the pair of mares’; θύραι is primarily ‘the pair of doors’, and corresponds to a Latin forae still found in the acc. foras and the abl. foris. The later development of the dual in Greek, τὰ κόρα, which has nothing to 28 correspond with it in Sanskrit, confirms this theory. The plural of aśve is aśvās; and in Oscan totas, the plural of tota a city, is evidently formed in the same way. We find traces of this old and genuine plural in Latin; Nonius quotes from Pomponius: quot laetitias insperatas modo mi irrepsere in sinum; under the lemma: accusativus pro nominativo (500. 33, M.). But a comparison with Umbrian and Oscan makes it probable that laetitias here is really the old nominative plural, which was superseded by the dual laetitiae. Still the close connexion of this dual form with the pronominal genitive ending -som makes it likely either that with this dual was associated a form got by analogy from the pronominal plural populoi (= populi), or that this dual was regarded later as a similar pronominal plural. Its association with the pronominal genitive seems even closer than that of the genuine pronominal plural; for in Latin the use of terrarum (old terrasom) for *terrum (old terrom) seems older than that of liberorum for liberum; while in Greek we have the corresponding χωράων, but ἱππόων is not developed.

The adjectives uter and neuter are not duals, but singulars with an ending -ter that associates them with the dual. While the pf. pl. legēre is not a dual, the ending -tis in estis seems to Brugmann a form of the Skt. ending -thas, used for the second dual of primary tenses in the active voice. Still this ending may be the old plural ending -te, still in use in the imperative regite, but changed to regitis in the indicative after the analogy of the second sing., regis. These, then, are the Latin inflexions of nouns or verbs that have been thought dual in origin.



I translate from Diomedes, de Numero (I. 334-5, K.): Verbs, moreover, are inflected by both numbers, the singular and the plural. The dual is in use among the Greeks only, but by us is excluded, just as it is in nouns. Nowhere can there be found in Latin any expression which shows the dual number. For the Romans, mindful of antique usage, refused to use the dual, which had been adopted as a novelty in the usage of later ages. For though it (the dual) is claimed as belonging from the beginning to discourse as produced by nature, it was disregarded and kept in obscurity, and for a considerable period lay hidden in uncertainty between both numbers, as well the singular as the plural. Later, however, as age followed age, it was adopted through a regard for scrupulous accuracy, and crept in as an intercalar number, and for this reason it is rare in old writers, since its frequent use involves constant error. To such an extent is this the case that all expressions of the Greek tongue that are unintelligible are explained as due to usages of this nature (i.e. of the dual). Only in Attic does it flourish to any great extent, and most of all in Homer, who, whilst he used the Attic dialect, as one favouring his mother-tongue, to follow the opinion of some, was after all not ignorant of ancient usage, as that well-known verse attests. For though they (the heralds) were two, himself mindful of antiquity set forth (Achilles’) greeting of them after this fashion: χαίρετε κήρυκες ἄγγελοι (Il. 1. 334). Besides, it seemed superfluous to the ancients, especially when the inflexion of the dual was shaped after the likeness of the plural.

As the only detailed statement about the dual proceeding from an ancient Greek, I find this passage interesting. Diomedes feels that the dual is later in development than the singular or the plural, that it is not a necessary number, nor one in general, but only in occasional, use; and that from a striving after scrupulous exactness of expression that commonly led to error. Perhaps an examination of the inflexions of the dual for nouns, verbs, and pronouns may make his meaning clearer; we shall see that in languages like Sanskrit and Greek, where the dual is most in use, it developed few inflexions in comparison with 30 the singular and the plural; and these in consequence were likely, if much used, to occasion frequent confusion of meaning. If a number has only one case form, as has the Greek dual, to express the genitive, dative, ablative, locative, and instrumental relations, it is easy to see what confusion would arise from its frequent use even for natural pairs. And so it seems likely that never in any branch of Indo-Germanic was the dual in constant popular use; at most it was used only to some such extent as we find it in Homeric Greek. In Sanskrit the dual has three case forms for nouns: one for the nom.-acc. ending in -au for masculines, -ī for neuters, and -e (= ai) for feminines; a second in -os for the gen.-loc.; a third in -bhyām for the dat.-instr.-abl. For a late and artificial number like the dual it is natural to expect that the endings should be significant, and that is what we find. The ending -e (= ai) is really the neuter ending -ī added to the fem. stem -ā, originally the ending of the so-called neuter plural; and so it may be classed as a mere variety of the neuter ending -ī. The ending -au is probably for older -āvi, where the vi means ‘two’, as it does in viginti. Whether the neuter -ī is another form of vi is a question suggested by forms like νῶϊ and σφῶϊ, which scholars are disposed to answer in the affirmative. In the ending -os of the gen.-loc. the -s seems a reduced form of the genitive ending -as (= os) which has been added to the nom. ending -āu, giving us -au-s (= os). The ending -bhyām is a variation of the dat.-abl. -bhyas, whose nature will appear more clearly when we come to pronouns. It seems to have no cognate parallel except in Avestan; and as it is not found in any of the Central or Western groups, it cannot be claimed for primitive Indo-German. But the ending -os is claimed for Latin, as we saw.

For Greek nouns the dual shows two case forms: the nom.-acc. in for ο stems, in -ᾱ for α stems, and in for the rest; and the gen-dat. ending in -αιν for α stems and in -οιν for the rest. The endings for α stems, as we shall see, are late developments following the analogy of those for ο stems. We find for pronouns νῶε for νῶϊ, which suggests that is a degeneration from -ῑ, the ending still in use for pronouns and ϝίκατι. Probably this -ῑ is for ϝι, identical with vi, of which the reduced form υ is part of the ending (= au). In the gen.-dat. ἵπποιν is shortened from the Homeric ἵπποιϊν (= ἵπποιϝιν), and the relation of this -ϝιν to -ϝι will be seen when we come to pronouns.

In Latin we noticed Sommer’s view that genū and cornū are old duals. It is hard to understand the ū on any other theory; for in the 31 Greek γόνυ and the Sanskrit janu the u is short; and in γουνός (= γονϝος) it is reduced to a consonant. If genu is really an old dual we can understand forms of the gen. sing. found in inscriptions like conventuus (C.I.L. 2. 2416) as for conventuos, showing the ending of the Sanskrit gen.-loc. dual.

For verbs we find a full set of personal forms for the Sanskrit dual, but a defective one for Greek. The so-called secondary or aorist forms are simpler and more primitive than the so-called primary or present ones—a state of things which appears natural only when you think of the aorist as the old timeless form of the verb, out of which the present was evolved. Sanskrit shows for the aorist tenses active: 1st dual -va, 2nd -tam, 3rd -tām; and for the middle: -vahi, -thām, tām. In the present tenses active we have -vas, -thas, -tas, and for the middle -vahe, -the, -te; for the pf. active -va, -thur, -tur, and for the middle -vahe, -the, -te. The imperative shows only one variation from these, -vahai for the 1st middle. In Greek we find for aorists 2nd -τον, 3rd -την in the active, and 1st -μεθον, 2nd -σθον, 3rd -σθην in the middle; for presents 2nd -τον, 3rd -τον in the active, and 1st -μεθον, 2nd -σθον, 3rd -σθον in the middle. At first sight the present endings seem the simpler here; but it may be questioned whether it is really simpler to use the same inflexion for ‘you two’ as for ‘they two’.

The only point of coincidence between the Greek and Sanskrit inflexions is in the endings -τον, -την of the aorist tenses, which correspond phonetically and in meaning to the Sanskrit -tam, tām; and we may take this as our starting-point. The middle endings -σθον, -σθην seem to come from the 2nd pl. ending -σθε on the analogy of -τον, -την; and -μεθον to have come from -μεθα, the 1st pl. ending, on the analogy of -σθον, -σθην. In the aorist tenses at times -τον is used for the 3rd, and -την for the 2nd (Hirt, Gr. Laut- u. Formenlehre, 403. 3), and it seems that from this confusion has arisen the use of -τον and -σθον for both the 2nd and 3rd persons in the present tenses. How this came about we shall understand better when we consider the use of σφῶϊ for ‘you two’. Why the 1st dual should develop much later than the 2nd and 3rd, and not at all for the Greek active, we may understand when we see that -va, its oldest ending in Skt., seems originally a word meaning ‘two’ or ‘both’, but later ‘we two’ or ‘we’. If the English ‘we’ is really the same word as the Skt. va, probably the 1st plural was originally very closely connected with the idea of ‘two’, and it was long before the need of a separate form for the 1st dual was felt.


In Sanskrit let us start from -va, the ending of the aorist 1st dual, related to -vi, the noun-ending of the dual, as dva is to dvi. The impf. has -va; the pres. has -vas, on analogy of the 1st pl. -mas; and the middle has aor. -vahi, pres. -vahe, on analogy of the 1st plurals -mahi and -mahe. Following -vas comes the 2nd dual -thas from the 2nd pl. -tha; and following -thas from the 3rd dual -tām we get -tas. For the middle the aorist middle has -tām for the 3rd, which it changes to -thām for the 2nd and to -the and -te for the 2nd and 3rd present after the plural endings -dhvam, -dhve, -nte; and the pf. endings -thur and -tur evidently follow the 3rd pl. pf. ending -ur. So -vahi, -vahe, -vahai follow the first plurals -mahi, -mahe, -mahai. So, like the Greek endings, excepting -τον and -την, all Sanskrit endings excepting -tam and -tām are evidently late formations, and indicate the truth of Diomedes’ view that the dual was a later formation, possibly from a striving after meticulous accuracy.

While in the Latin verb we find no endings in use as dual, Brugmann and Thumb agree that -tis in estis or ducitis is the Skt. -thas (= -thes), and that the primary Italic -tes has changed to -tis after the analogy of -is in ducis. If cornūs is really the old genitive dual, and estis really the old 2nd dual of the pres. active, the only dual inflexion for cases in Latin is in use as a singular inflexion, and the only dual inflexion for persons as a plural inflexion. This balance, too, might seem to indicate that, when the Latin dual disappeared, while it usually passed to the plural, at times it became a singular.

To understand the dual endings of the verb in Greek and Sanskrit we must give some attention to the Latin, Greek, and Sanskrit personal pronouns, and especially to their dual forms in Greek and Sanskrit. Let us begin with νῶϊ, ‘we two’. Homer has an older form, νώ (Od. 15. 475 and 16. 306), which has been adopted by Attic, where it develops the gen.-dat. νῶν, not found in Homer. But Homer has usually νῶϊ, and from it he forms a gen.-dat. νῶιν. We feel that νώ is an old dual like in ending to δύω and ἄμφω; what then is νῶϊ? Brugmann thinks it is for an older νῶϝι, the -ϝι being the same as the ϝι in ϝίκατι and the vi- in viginti; so that νῶϝι is equivalent to ‘we two two’. The Greek νώ is probably the Skt. nau, used as an enclitic for the acc. dat. gen. dual ‘we two’. In later Greek we have νῶε, with the same ending as δοῦρε ‘the two spears’, which may develop on its analogy. The Greek νώ appears in Latin as nos, passing into the plural just as ambo passes into ambos and duo into duos. That nos 33 is first an acc. and later a nom. is confirmed by the use of nau in Skt. as an acc. and not a nom., while in Avestan the corresponding enclitic acc. has developed a nom. nā̊. It may be that we have here the starting-point for the use of the acc. pl. as a nom. in Latin, which we see in homines, a use which spreads to the sing. in the French homme (= hominem), while on (= homo) has become a specialized term.

While nau in Skt. is the enclitic acc. for ‘us two’, nas is the corresponding enclitic pl. for ‘us’. The English ‘us’ is the German uns (= n̥s), which is a reduced grade of nas, which may have been nos or nes in Indo-Germanic. This reduced form appears in the first syllable of the accented acc. asmān (= n̥sman), and in the Aeolic ἄμμε (= n̥sme), the Skt. asma-, found in the dat. asmabhyām and the instr. asmābhis. The Doric form is ἅμε, which gets its aspirate from ὕμε ‘you’, just as ἐσμέν gets its first syllable from ἐστέ. From ἄμμε and ἅμε, old accusatives, come the nominatives ἄμμες and ἅμες. The Attic ἡμεῖς; has changed α to η and -ες to -εις after the analogy of τρεῖς, thinks Brugmann; for it is a plural, not a dual.

We have found the root of ‘us’ in a reduced form of nos; but ‘we’ seems an older form of vos. This form appears in the Skt. vām, the enclitic acc. for ‘you two’, in vayam, the Skt. nom. pl. for ‘we’, and in āvām and yuvām, the nom. duals for ‘we two’ and ‘you two’, as also in vas, the enclitic acc. for ‘you’. We may begin with vām, according to Brugmann for va-am, the va being the same word as is used to form the 1st dual aorist, to which is added -am after the analogy of aham, tvam, vayam, and yūyam. This va ‘two’ is our ‘we’ and the Latin vos, got from the same root and in the same way as the Skt. vas, and being the plural form. The root va ‘two’ has become ‘we two’ by association with verbal forms in the first pl. and ‘you two’ by association with verbal forms in the second pl.

This vām (= two) is joined to the root yu- in yuvām, the Skt. nom. dual ‘you two’. The root yu-, also found in the Skt. nom. pl. yūyam ‘you’, is the root of our ‘you’, and seems the same as the root of Latin iuvo ‘I help’. In Greek we have the Aeolic ὕμμε, and the Doric ὕμε, corresponding phonetically to the Skt. yuṣma- in the instr. yuṣmābhis ‘with you’. It passes to the Attic ὑμεῖς, formed like ἡμεῖς, except that the rough breathing is not got by analogy, but represents the Skt. y. The ν added in νῶιν, σφῶιν, ἡμῖν, and ὑμῖν is probably the same that we have in ποσίν and ἀνδράσιν, though that does not explain it, for it seems older in νῶιν.


Nos and vos, then, are to be associated in derivation with the Skt. enclitic duals nau and vām, and with the Skt. enclitic plurals nas and vas. Brugmann notes that they, too, are used as enclitics at times, as in: ob vos sacro, old for obsecro vos, and: Quo nos cumque feret melior fortuna parente (Od. 1. 7. 25). Both of them have a double genitive, one with a singular ending and a singular meaning: nostri—vestri; and the other a plural in form and meaning: nostrum—vestrum. It is worth noting that these twin genitives are formed, not from nos and vos directly, but from the dual possessive forms noster and vester; so that here again we have a balancing of singular and plural in connexion with the dual, such as we have repeatedly observed.

When we turn to the 2nd and 3rd dual forms we get a balancing, not of number but of person. We noticed how in the aorist tenses of the verb the 3rd dual is used at times for the 2nd, and the 2nd for the 3rd; while in the present tenses the 2nd is used throughout, when we turn to dual pronouns of the 2nd and 3rd persons, we have the 3rd person with a slight variation used throughout. We have seen how νῶϊ and νῶε are used for ‘we two’; σφῶϊ is used for ‘you two’, and σφῶε for ‘they two’. Brugmann’s idea, that in σφῶϊ the φ is for an older ϝ and that it must be connected with σύ, seems most improbable. Both σφῶϊ and σφῶε are dual forms got from σφι after the analogy of νῶϊ and νῶε. Brugmann shows us (Vergl. Gr. II2 p. 413) how from οὗ (= σϝοῦ) we get an instrumental σφι or σφιν, mistaken later for a dative and expanded to σφίσι or σφίσιν; how this develops a genitive σφέων, an accusative σφέας or σφέα, and a nominative σφεῖς. There can be little doubt that σφῶϊ and σφῶε are variant duals developed from the same instrumental, and that the problem here is much the same that we have in the German use of Sie for Ihr, and later for Du.

The use of ἡμεῖς and nos for ἐγώ and ego is very common in Greek and Latin, and we shall speak of it presently when dealing with the Plural of Modesty. That of ὑμεῖς and vos for σύ and tu is not developed in either classical Greek or Latin, though we find uses approaching it from which the later uses in Low Latin and French are derived. These consist in the choice of a single person out of the number addressed, so that he or she alone of that number is indicated by the noun or pronoun used in the address; as in: νῆα ἰθύνετε, φαίδιμ’ Ὀδυσσεῦ (Od. 12. 82), προσέλθετ’ ὦ παῖ (Soph. O. C. 1104), heus foras exite huc aliquis (Pl. Epid. 399), ibitis Aegaeas sine me, Messala, per undas (Tibull. 1. 3. 1), vos, O Calliope, precor, adspirate canenti (Aen. 9. 525).


The use of Sie for Ihr probably arose from a substitution of indirect for direct address, similar to that which substitutes ella for voi in Italian; and in Greek, too, it is likely that, in addressing a herald or a vassal chief, the king feeling that the absent ruler, and not the present herald, should determine the form of address, used ‘they two’, and not ‘you two’, in addressing him and his lord. Latin prefers the second person to the third in such a case, but the Greek use here seems to follow the lines of modern courtesy.



While words ending in -ter are not duals, as some Roman grammarians taught of uter and neuter, in origin they are connected with the dual notion, the ending -ter being a comparative ending. It is interesting to see how far they continue to express this notion, how far they pass on to the expression of three or more, or pass back to one. In this respect the uses of inter will prove noteworthy.

Inter has cognates in all branches of Indo-Germanic: in Sanskrit and Zend, in Armenian and Old Bulgarian, in Greek and Italic, in German and Celtic. In English we have under, in German unter meaning ‘among’ as well as ‘below’. In Latin, while it usually means ‘between’, or ‘among’, it is at times the equivalent of per. Its derivation is not obscure, the first syllable being the Latin in, Greek ἐν, English in. When en- is followed by a syllable with an initial mute, the e becomes i in Latin, and so the Greek ἐντός appears there as intus. We can see the same tendency in our pronunciation of England. The cognate of inter in Greek, ἔντερα, is a substantive, not a preposition.

The ending -ter is evidently the same as the Greek ending -τερος, and we have it in the Latin exterus, which develops a double comparative, exterior. The corresponding interus has disappeared, interior having taken its place. It is the usual ending for comparatives in Sanskrit; and there, too, it is often joined with prepositions, as in uttaras ‘higher’, from ud ‘up’. In English and German the usual comparative ending is -er, which we see in the Latin superum and inferum. While in Latin the idea of below is usually given by inferus and infra, inter in composition has this meaning at times, as in interire ‘to go down’ and interficere ‘to slay’. We shall understand this shifting better if we compare the use of imus ‘lowest’, primarily the superlative of in (= in-mus) and meaning inmost, as we see it in Catullus’s phrase: imis exarsit tota medullis (64. 93). But it is commonly felt to be a superlative of inferus, as in: imis avolsam solvit radicibus (Aen. 8. 237-8), where in relation to the earth ‘lowest’ is also ‘inmost’.


Inter, then, as a preposition means primarily ‘between two objects’, as in: qui (mons) est inter Sequanos et Helvetios (B. G. 1. 2). Virgil often places it so as to present us with a picture of this meaning, as in: terras inter caelumque (Aen. 4. 256). But just as the dual passes from a pair to two or more pairs, so we find inter passing to two pairs in: namque manus inter maestorumque ora parentum (Aen. 2. 681), and then to several in: ante oculos interque manus sunt omnia vestras (11. 311). In this way it seems to have been transferred to a plurality of objects with the meaning ‘among’, as in: micat inter omnes Iulium sidus, velut inter ignes Luna minores (Od. 1. 12. 46-8).

Scholars have been puzzled by a curious repetition of inter ‘between’, found in both prose and verse. We read: inter Hectora Priamiden animosum atque inter Achillem ira fuit capitalis (Sat. 1. 7. 11-13), and again: Nestor componere lites inter Peliden festinat et inter Atriden (Ep. 1. 2. 11-12). Wickham’s note is: ‘an illogical, but a Latin use’, and he compares: quid intersit inter popularem . . . et inter constantem (Cic. Lael. 95. 25). But though we feel this use illogical for inter ‘between’, it seems quite logical for inter ‘among’, as in: Lycus inter et hostes inter et arma fuga muros tenet (Aen. 9. 556-7), with which compare: fortunate senex, hic inter flumina nota et fontes sacros frigus captabis opacum (Buc. 1. 51-2). Either is right with inter ‘among’; but for inter ‘between’ we feel that the logical use is that in: inter Padum atque Alpes (Liv. 5. 35. 2). We have then inter ‘between’ at times in Latin following a syntax not unusual for inter ‘among’; a syntax which gives a higher degree of weight and balance to the pair in competition.

But we read in Lucretius: inter saepta meant voces et clausa domorum transvolitant (1. 354-5) ‘voices pass through walls, and fly through houses shut’, Munro. It seems clear that here inter is used for per; and in Virgil:

Ecce autem flammis inter tabulata volutus

Ad caelum undabat vertex turrimque tenebat (Aen. 12. 672-3),

‘and lo! a spire of flame wreathing through the floors wavered up skyward and held a turret fast’, Mackail. In this sense of per it is more usual for time than for place. Cicero writes: qui inter annos tot unus inventus sit (Leg. Man. 68. 23), and: quae inter decem annos . . . nefarie flagitioseque facta sunt (Verr. 1. 37. 13); and Livy: inter ipsum pugnae tempus decem naves regiae . . . ad Thronium in sinu Maliaco stabant (36. 20. 5). In all these examples it seems 38 more natural to use per than inter. We find Gellius writing: qui plus cernunt oculis per noctem quam inter diem (9. 4. 6). But it is easier to see how the use arose for place than for time. We expect to find: inter ripas fluit Tiberis; but Virgil writes:

ubi Lydius arva

Inter opima virum leni fluit agmine Thybris (Aen. 2. 781-2),

where arva is substituted for ripas, the whole for the part. So we have inter arva, where in prose we should write per arva, or per agros. And so we read:

hunc inter (lucum) fluvio Tiberinus amoeno

Verticibus rapidis et multa flavus arena

In mare prorumpit. (Aen. 7. 30-2.)

Of course in return we may expect per for inter, and in:

Principio delubra adeunt pacemque per aras exquirunt (Aen. 4. 56-7),

Sidgwick translates per aras, ‘amid the altars’. In: concussam bacchatur fama per urbem (4. 666) we have a poetic expression for: turbatos bacchatur fama inter cives. Perhaps we might connect with these examples the verbs interire and perire with closely approximate meanings. But it is quite plain that inter has a singular, as well as a plural and dual force.



Speaking generally, alius seems to play the part in older Latin that alter plays in Silver and later Latinity. In archaic Latin alius is often used to denote the other of two as well as of three or more. It is also found at times with the meaning of ‘any one’, like aliquis. In Golden Latinity as a rule alter is ‘one of two’, and alius is ‘the other’ of three or more. But in Virgil alter is already used for ‘the other’ of three or more; and in later Latinity and the Romance languages altro or autre is the regular word for ‘other’, while of alius few traces subsist. The Romance languages follow poetic diction, which tends of two words to choose the stronger and more emphatic; and that in this case was plainly alter.

Both start from the root ali-, which we have in the English ‘else’, and in the Greek ἄλλος (= alyos), ἀλλότριος, and reduplicated in ἀλλήλων, where the duals ἄλλω-ἄλλω passed to neuter plurals ἄλλα-ἄλλα, contracted to ἄλλᾱλα (Ion. ἄλληλα) with a single λ in the 2nd syllable, parallel to the single s in misi, as following a long vowel. We have another grade of this al- in ol or ul, that we find in olli, archaic for illi, and in ultro and ultra ‘on yon side’. We read:

alii ventosis follibus auras

Accipiunt redduntque, alii stridentia tingunt

Aera lacu. Gemit impositis incudibus antrum.

Illi inter sese multa vi bracchia tollunt

In numerum versantque tenaci forcipe massam. (Aen. 8. 449-53.)

It is clear that illi is parallel to the preceding two alii, and that it denotes a third class of smiths, who wield the hammer; and Servius’s note is: ‘illi’ quidam pro ‘alii’ accipiunt. Henry wished to read alii here, but the kinship of the words was felt by Virgil.

While autre goes back to the root ali-, not so our ‘other’; we see its root in ‘and’ and in the German andere, in the Sanskrit anyas, and the Greek ἔνιοι; it was once onþer; but the n is lost in ‘other’ just as the n of Gans and hanser is lost in goose. In Greek ἔνιοι has 40 lost the sense of ‘others’, and is ‘some’. So in Latin aliquis has as a rule lost its old meaning of ‘some one else’ and means ‘some one’. But I read in Tacitus: ne eis quidem annis, quibus Rhodi specie secessus exsul egerit, aliquid quam iram et simulationem et secretas libidines meditatum (Ann. 1. 4. 4). Our editors usually correct aliquid to aliud, against all the manuscripts, but aliquid here probably still conveyed to Tacitus and his readers the meaning ‘anything else’. So the Skt. anye ‘others’ passed to the Greek ἔνιοι ‘some’; and in Latin aliquos usually means ‘some’; but we read:

Quique sui memores aliquos fecere merendo. (Aen. 6. 664.)

While most editors change aliquos to alios, Sidgwick retains aliquos, the reading of all good manuscripts, but finds a ‘Virgilian pathos’ in his rendering ‘some hearts’ suggesting the narrow range of gratitude for human merit; but it is far more likely that Virgil still felt here in aliquos, the old meaning ‘others’. And in:

Sin aliquem infandum casum, Fortuna, minaris (Aen. 8. 578),

the meaning must be ‘some dread fate other (than his return)’; and Cicero has: vellem aliquid Antonio, praeter illum . . . libellum (Brut. 163. 44), and Seneca: aliqua res extra eloquentiam (Cont. Top. 22), and Porphyrio: neminem posse aliquid agere quam quod consuerit (ad Od. 1. 1. 16).

An anonymous grammarian tells us that aliquando is compounded of alius and quando, and is in use for the past and the future; evidently he feels that it naturally marks some other time than the present. But generally we have for ‘something else’ aliquid aliud, for ‘somewhere else’ alicubi alibi, alicunde aliunde for ‘from some other quarter’, and alias aliquando for ‘at some other time’; i.e. the loss of the idea ‘other’, constant in ἔνιοι, is only usual in aliquis. This loss seems to have come from such repetitions as we have in: sin aliquis exstiterit aliquando (Cic. de Orat. 3. 80. 21), or: verum aliquando aliqua aliquo modo alicunde ab aliqui aliquast tibi spes mecum fortunam fore (Pl. Epid. 331-2), where we feel all but the first ‘other’ superfluous. But in: hic opus est aliquot ut maneas dies (Pl. Poen. 1421) the natural translation is ‘it is well for you to wait a few days longer’. And so in Nonius’s reading of: nec nobis praesente aliquis quisquam nisi servus (Pl. Amph. 400).

In return we have a number of cases where alius seems to mean the same as aliquis or ullus; e.g.: neque maius aliud neque praestabilius invenias (Sall. Jug. 1. 2), non alia ante Romana pugna atrocior fuit 41 (Liv. 1. 27. 11), neque enim aliud . . . difficilius reperient (Quintil. 4. 2. 38), quamvis non alius flectere equum sciens aeque conspicitur gramine Martio, nec quisquam citus aeque Tusco denatat alveo (Od. 3. 7. 25-8) (where alius is balanced by quisquam), Eridanus, quo non alius per pinguia culta in mare purpureum violentior effluit amnis (Geo. 4. 372-3). Probably with examples like these we are to associate such uses as: Fama, malum qua non aliud velocius ullum (Aen. 4. 174), or: mulier qua mulier alia nulla est pulcrior (Pl. Merc. 101), so that for the union aliud ullum we have above aliud used in the sense of ullum, a usage we must examine under Metonymy. So in:

Sed non ante datur telluris operta subire

Auricomos quam quis decerpserit arbore fetus (Aen. 6. 140-1),

ante quam quis is poetic for alii quam qui; and the direct prose would be: illi soli dabitur qui et.

While in archaic and classical Latin alius is usual where three are in question, it is used at times to express the other of two, i.e. for alter; e.g. in: per illam tibi copiam copiam parare aliam licet (Pl. Epid. 323-4), remedium tumultus fuit alius tumultus (Tac. Hist. 2. 68. 4). In uses like: ex loco in alium locum (Plin. Ep. 10. 69), aliudque cupido, mens aliud suadet (Ov. Met. 7. 19-20), alius est amor, alius cupido (Afran. com. 23), alius seems the opposite of idem, a force heightened to mutatus in:

Non vires alias conversaque numina sentis? (Aen. 5. 466.)

Here, too, when the clause is negative, alius is equivalent to alter, as is clear when we compare: malum qua non aliud velocius ullum (Aen. 4. 174) with: Misenum . . . quo non praestantior alter (6. 164). In: quos alios muros, quae iam ultra moenia habetis? (9. 782) we are very close to the use of alius instead of alter for ‘second’, which we reach in: alius Latio iam partus Achilles (6. 89).

This explains the idiom unus . . . alius for alter . . . alter, in: (leges) duas promulgavit, unam . . . aliam (B. C. 3. 21), and: unam . . . epistolam acceperam . . . in qua significabatur aliam te ante dedisse (Cic. Att. 7. 12. 1). Alius is not used for alter in union with another numeral, as in: altero vicesimo die (Cic. Fam. 12. 25. 1); but we read: quarum unam incolunt Belgae, aliam Aquitani, tertiam . . . Celtae (B. G. 1. 1. 1), ab alio exspectes, alteri quod feceris (Com. Inc. 82), geminae . . . portae, quarum altera . . . aliam (Val. Fl. 1. 833-5), duo agmina parant quorum altero populatores invaderentur, alii castra Romana adpugnarent (Ann. 4. 48. 4). We have in: quaeritur huic alius (Aen. 5. 378) 42 alius used for compar, a use that would be striking even for alter.

Alius is at times constructed with the ablative, as though it were a comparative, e.g.: alius Lysippo (Ep. 2. 1. 240), alium sapiente bonoque (Ep. 1. 16. 20), expertis alia experiri (Liv. 5. 54. 6), si accusator alius Seiano foret (Phaedr. 3, Prol. 41). True, the ending -ius hardly seems the same as that we have in maius -oris, or in plus (= plouios) -pluris. Hence Sommer (302. 1) thinks the comparative force inherent in the root ali-, just as minor-minus, with no proper comparative ending, gets its comparative force from the root mi- ‘to lessen’. But Brugmann believes that it is from words like alius that the ending -ios gets its comparative force. We have this ending in medius, δεξιός, and σκαιός (older σκαϝιός, Latin scaevus). No doubt δεξι- and σκαϝι- are old locatives like ali- ‘on yon side’. Tertius has this same comparative ending, and illustrates the use of alius for one of three, as well as for one of two; for if alius is originally a comparative, the latter will be its primary force.

We find, then, that alius, often a comparative in force and probably primarily in form as well, had in old Latin already passed on to a use for one of three as well as for one of two, a use it retained there to a considerable degree. Through its association with words like quis or ullus, it had lost this comparative force at times, and had with it lost all reference to a dual or plural, i.e. was absolutely singular. We may note its omission in:

Impastus ceu plena leo per ovilia turbans,

Suadet enim vesana fames, manditque trahitque

Molle pecus mutumque metu (Aen. 9. 339-41),

where the last phrase seems equivalent to: et mollis pecoris alia mandit trahitque alia.

We read in Servius: alter enim de duobus dicimus, non de tribus (ad Buc. 3. 34), and when we find it used of three and not of two, we seem to have entered on a new period in its history. When we read in Cato: (vinum) in dolium infundito, . . . transfundito in alterum dolium, post dies XX in alterum dolium transfundito (R. R. 112. 2), it is not quite certain that it is three objects, and not a succession of pairs with which we have to do. But in:

Hoc erat, alma parens, quod me per tela, per ignes

Eripis, ut mediis hostem in penetralibus, utque

Ascanium, patremque meum, iuxtaque Creusam

Alterum in alterius mactatos sanguine cernam? (Aen. 2. 664-7),


the use of alter for one of three is plain. But we have also:

Tum geminas vestes auroque ostroque rigentes

Extulit Aeneas . . .

Harum unam iuveni supremum maestus honorem

Induit (11. 72-3 and 76-7).

And in like fashion we find: unum exuta pedem (4. 518) and: unum exserta latus (11. 649), where we have, not, it is true, alter for unus, but unus for alter. And thus we find alter, too, like inter, advancing from two to three, and apparently receding from two to one; for it is replaced by unus.

Properly, in designating the members of a pair we should have alter . . . alter, as in: alter istinc, alter hinc adsistite (Pl. Rud. 808). But often, as with Castor and Pollux, only one need be expressed, as in: in altera parte fluminis (B. G. 2. 5. 6); or the other is designated by a different word, as in: summus ibi capitur meddix, occiditur alter (Enn. Ann. 328); and so alter takes on the meaning of ‘second’, as in: erus . . . et erus alter (Pl. Capt. 1005), becoming a preferred competitor of secundus that may imply inferiority, as we see in: haec fuit altera persona Thebis, sed tamen ita secunda ut proxima esset Epaminondae (Nep. Pelop. 4. 3). We find it used for second in a series of three; e.g.: primus . . . alter . . . tertius (Aen. 5. 310 ff.) or: una . . . alter . . . extremus (5. 563 ff.). Hence we get unus et alter ‘one or two’ and unus aut alter ‘one or perhaps two’, which in later Latin comes to mean the same as Cicero’s unus et alter. While alius is not used as an ordinal numeral in union with other ordinals, alter is often thus used, as in: litteras quas mihi altero vicesimo die reddidit (Cic. Fam. 12. 25. 1). In: alter ab undecimo tum me iam acceperat annus (Buc. 8. 39) Servius is quite positive that the thirteenth, not the twelfth year, is meant, the twelfth being too far from puberty.

Familiar is the reciprocal force of alius repeated, as in: ceteri verbo alius alii adsentiebantur (Sall. Cat. 52. 1), ut ipsi inter se alii aliis prodesse possent (Cic. Off. 1. 22. 7). When alter is thus repeated, and only two are in question, it sometimes has this force, as in: ut alter alterius iudicium non modo reprehendat, sed etiam rescindat (Cic. Cluent. 122. 43), but sometimes not, as in: consules primum religiones, deinde alterum alterius mors et comitia . . . impediunt (Liv. 41. 16. 7). But for more than two it is commonly reciprocal, as in: ut nemo memoria dignus alter ab altero videri nequiverint (Vell. 1. 16. 5), cum alter alterum indignaretur imperare (of four) (Capitol. Alb. 1. 2), omnes rediere . . . inconsideratae dementiae alter alterum arguentes (Amm. 31. 15. 15).


As early as Lucretius we find the union alius . . . alter in: ex alio terram status excipit alter (5. 835), hic odor ipse igitur, naris quicumque lacessit, est alio ut possit permitti longius alter (4. 687-8). But this confusion was not felt in some phrases. We feel how different from altero die or alio die used for it in: servolos rogitabam . . . item alio die quaerebam (Ter. And. 89) ‘on the next day’, is alio die in: mox quasi alio die studebat (Plin. Ep. 3. 5. 11) ‘as on an ordinary day’, or: confecto negotio bonus augur . . . ‘alio die’ inquit (Cic. Phil. 2. 83. 33) ‘not to-day’. By altero die we mean ‘on the morrow’, ‘on the day after’, as in: altero die pervenit Caesar (B. C. 3. 30. 6), altero die quam a Brundusio soluit (Liv. 31. 14. 2). But in the Itin. Anton. (Plac. re. R. 30) alia die de natale domini is ‘the day after his master’s birthday’, and alia die (Pallad. 9. 8. 6) ‘on the following day’.

We have alter for ‘the neighbour’ in: qui nihil alterius causa facit et metitur suis commodis omnia (Cic. Leg. 1. 41. 14), cave ne portus occupet alter (Hor. Ep. 1. 6. 32). Just as we have alterum tantum for ‘as much again’, we have alter ego for ‘a second self’ in: te me esse alterum (Cic. Fam. 7. 5. 1), and so alter idem in: est enim is quidem tamquam alter idem (Lael. 80. 21). But it is obvious that this may be carried too far; and just as from ad and salto we have adsulto, from ad and alter we have adulter, the ‘too neighbourly’ man. We read in Festus: et adulter et adultera dicuntur, quia et ille ad alteram et haec ad alterum se conferunt (Paul. p. 22). Probably the alter in: fruitur nunc alter amore (Tib. 1. 5. 17), quam vacet alterius blandos audire susurros (Prop. 1. 11. 13) has much the same significance; for in late Latin alterare is used for ‘to spoil’, much in the same sense as adulterare.

We noticed how alius is used as the opposite of idem, and in Virgil advances to the sense of par or compar. But in Horace we have alter used as the opposite of idem in: quotiens te speculo videris alterum (Od. 4. 10. 6). In late Latin instead of alius . . . alius we have alter . . . alter as in: altera substantia divinitatis, altera humanitatis (Vinc. Ler. 13. 19), et aena et lignea et fictilia simulacra, et alterius alteriusque materiae (Prosper, in Ps. 113. 4). So for more than two alter becomes usual, as in: alter adulescens decessit, alter senex, aliquis praeter hos infans (Sen. Ep. 66. 42),

Et nunc ex illo forsan grege gentibus alter

Iura dat Eois, alter compescit Hiberos,

Alter Achaemenium secludit Zeugmate Persen (Stat. Silv. 5. 3. 185-8),


cum alter maneret in Capitolio, alter in Palatio, alter . . . alter . . . alter (Lampr. Hel. 30 4). Hence we have the mixture in: altera detur si similis tellus, aliaeque . . . exsurgant rupes (Sil. 12. 72-4), altera nox aliisque gravat plaga caeca tenebris (Stat. Theb. 8. 16).

When we see the clearer and more definite force of alter in classical Latin, we need not be surprised that it proved the victor. And yet it was not always felt clear or forcible enough to give the sense ‘one of the two’, and we have often alteruter used for this. But in such a case, when it is repeated, it is expressed by alter only, as in: ne . . . alteruter alterum praeoccuparet (Nep. Dio 4. 1), aut etiam alterutrum, nisi terminet alterum (Lucr. 1. 1012).

Sommer feels that -ter in alter is the same as -ter in aliter. Alter is for the old aliteros where, because of the two morae following ali-, the i has fallen out, while in aliter, the old neuter singular, there followed but one mora, the short syllable -ter; and he compares validus with valdē, where the i is lost before the final -dē, the equivalent of two morae. Aliter is the same kind of adverb as we have in dulce ridentem ‘laughing a sweet laugh’, a cognate accusative neuter of the adjective. Osthoff accounted for the -ter in breviter as being contracted from itere, breviter being primarily brevi itere (Woelff. Arch. 4. 455); but Delbrück pointed out that the ending -iter in obiter and pariter must be connected with the -ter in inter, subter, and propter. Probably propter was originally propiter, and perhaps behind inter lies an older initer (cf. ἐνί); but the proclitic use of these words as prepositions would naturally lead to shortening as well as loss of accent, giving us propter and inter. Cette for *cedite Sommer thinks was shortened very much as was our ‘good-bye’; he calls it an allegro form. Aliter looks like a starting-point for such adverbs as celeriter. Very clear becomes the sense of atque in union with it in: omnia plena pacis aliter, ac mihi Calvena dixerat (aliter) (Cic. Att. 14. 9. 3), and we see at once the force of aliter et in: si aliter est, et oportet (aliter esse) (Att. 11. 23. 1). Mela’s aliter a ceteris agunt (1. 57) shows the way to Cledonius’s velocius equus ab equo ‘one horse runs faster than another’.

Iterum, unlike aliter, has assumed the ending of the acc. sing. masc.; perhaps we may connect it thus: semel et iterum pervenire ‘to reach one meal and then a second’. Its root is the same that we have in ibi (= thereby) and ita; and iterum rogo is: ‘I ask that second thing.’ The other form itero, found in inscriptions, would stand naturally for iterum in: ac primo quidem decipi incommodum est, iterum stultum, tertio turpe (Cic. de Inv. 1. 71. 39).


Ceteri we are constantly using in et cetera, not seeing, as the Romans, too, failed to see, that ceteri was a contraction for caeieteri, the Roman equivalent of καὶ ἕτεροι ‘and the others’. For ἕτεροι the Doric ἅτεροι seems the older form, being for sm̥teroi ‘the one party of the two’ (cf. ἕν for sem). In common use it is ‘the remaining majority’ as opposed to reliqui ‘the remaining few’; but, as we shall see, there are many uses of it which show this to be a later and acquired meaning. We have cetera for alia in: ceu cetera nusquam bella forent (Aen. 2. 438-9) and alia for cetera in:

Inde alias animas quae per iuga longa sedebant

Deturbat, laxatque foros (6. 411-12),

and in:

Obstipuere animis alii, sed Troius heros

Agnovit sonitum (8. 530).

The spelling in inscriptions is at times caeteri. The form ceterus is not in common use, and the meaning makes this natural. In caetera multitudo and caeterum triticum the use of the nominative is easy, but caeterus vir would be impossible. Still we have masculine collectives; and Gesner1 cites: ceterus ornatus domi Pompeiis emptus est (Cato R. R. 22. 3), which I do not find in Keil’s edition. Usually alius is used for it, as in: et alius exercitus ratibus iunctis traiectus (Liv. 21. 27. 6). Ceterum the conjunction, and cetera are got from the same construction as we have in: cetera Graius (Aen. 3. 594).

1 Gesner did not receive this into his own text of the Auctores de Re Rustica, but notices it there as a reading got from the notes of Politian, and adopted in the editions of Jenson and Jo. Gymnicus.



Of all words ending in -ter that arise from the dual, uter seems most puzzling in its formation. Clearly it should be cuter (old *quoteros, Att. πότερος, Ion. κότερος, Skt. kataras), and we do find in inscriptions necuter (C.I.L. vi. 1527. 64), as well as necubi, sicubi, necunde, alicunde. Neuter, says Sommer, is a later formation, developed after the change of the old *cuter to uter (Lat. L. & F. Lehre, p. 469). Either, says Brugmann, we have a wrong division into nec-uter, nec-ubi, &c., aided by the presence of a uter ‘one of the two’, which appears as ater in Old Slavic, or the initial cu- was altered to u by internal phonetic change (Vg. Gr. II2, 2. 346. 3). What this phonetic change may be Brugmann does not specify.

What is the relation between quis ‘who?’, and quis ‘any one’? Clearly they are originally one and the same word; but quis ‘who?’ has an acute accent, which quis ‘any one’ has lost. The change is old; for the same is true for the Greek τίς and τις. We have already noticed how ali- in aliquis lost its force. So numquis hoc dixit? plainly meant to begin with: ‘now who has said this?’. The strong phrase accent on the first word is analogous to the strong initial stress accent on aliteros, which led to the loss of i and changed the word to alter; and in like fashion the quis in the second place is weakened in meaning and accent, and we get: ‘Now has any one said this?’. Quis thus reduced becomes an enclitic, as we find it in aliquis and ecquis.

The -uter in neuter does not mean ‘which of the two’, but ‘either’. So with the uter in uterque ‘either soever’; but uter standing alone is usually ‘which of the two’. Still, when we have uter repeated, the second does not mean ‘which of the two’, but seems to be the uter which Brugmann identifies with the old Slavic ater, meaning the same as alter. For in: quaerere . . . uter utri insidias fecerit (Cic. Mil. 23. 9) or: ambigitur . . . uter utro sit prior (Hor. Ep. 2. 1. 55) the second uter has the same force as it has in: neuter utri invidet (Pl. Stich. 733); and we have a very close parallel to: si quis quid contra rempublicam 48 fecerit. So to the Roman the second uter (= alter) might seem to bear the same relation to the first uter (= cuter), as quid here bears to quis. And so from the proportion quis? : quis :: cuter : uter we should get the use of uter for ‘which of the two?’, as well as for ‘the other of the two’.

Cuter was for quo-ter ‘which of the two?’; but the u in uter is a reduction of the ve- that we find as vi- in viginti, as the dual vau in Sanskrit, and as the plural vos and ‘we’ in Latin and English. It appears in the Skt. ubhau ‘both’; and uter is ‘one of the two’, while alter is rather ‘one of yon two’. We have, then, in alteruter very much such a reduplication as we have in quisquis; and in: omnium controversiarum, quae essent inter aratorem et decumanum, si uter velit, edicit se recuperatores daturum (Cic. Verr. 2. 3. 35. 14) uter seems short for alteruter. Quintilian tells us: de praemiis quaeruntur duo: . . . ex duobus uter dignior; ex pluribus quis dignissimus (7. 4. 21). So far as I know uter is used only of two persons or things, or of two parties or sets, as in: aliquando utrimque sunt testes, et quaestio sequitur, ex ipsis: utri meliores viri (Quint. 5. 7. 34). Unlike alter, uter is a dying word in later Latinity, and quis tends to take its place; as is plain from: quos igitur anteferret (Ann. 1. 47. 2).

In uterque the -que is worth a note. I have no doubt that in origin it is the same -que that we have in geminusque Pollux. From qui (= how?), the accented locative of quis?, we get an indefinite qui ‘somehow’, just as from quis? we get quis ‘some one’. Ribbeck saw that we had this qui in a weakened form in neque (= nequi ‘no how’); but he distinguished that que from the conjunction que, and thought of: neque opes neque arma habebant as involving an asyndeton. But in: hostium currus arma castra cepit we have an older syntax than in: hostium currus et arma et castra cepit; for et as a conjunction is late, and never developed in Greek, where ἔτι remains the adverb. But que, Greek τε, Skt. ca, is very old. When we remember, however, that ut is primarily ‘how?’ and que, ‘somehow’, and compare Cicero’s use of ut in: cum machinatione quadam aliquid moveri videmus, ut sphaeram, ut horas, ut alia permulta (N. D. 2. 97. 38), with the use of que in:

Captivi pendent currus curvaeque secures . . .

Spiculaque clipeique ereptaque rostra carinis (Aen. 7. 184 and 186),

we feel that in the examples cited both ut and que are conjunctions with meanings closely allied.


Que, then, in uterque will mean ‘somehow’ or ‘soever’ and uterque ‘one of the two soever’ or ‘either of the two’. We have in Horace’s: mihi cumque salve rite vocanti (Od. 1. 32. 15) cumque for ‘whensoever’, and in: indignor quandoque bonus dormitat Homerus (A. P. 359) quandoque with the same sense. And we have cumque used as the corresponding indefinite ‘at any time soever’ in:

Contemplator enim cum solis lumina cumque

Inserti fundunt radii per opaca domorum (Lucr. 2. 114-5),

‘observe, pray, when at any time the sun’s rays are admitted, and pour their light through the shaded chambers’. Munro feels that cum . . . cumque is for quandocumque and means ‘whenever’. No doubt it was in this meaning that cumque was subjoined to quando, and it was in this union that it got the force of ‘soever’ instead of the older ‘whenever’, in which meaning it usually takes the place of the simpler que in later Latin. In this sense it is subjoined to uter ‘which of the two’, while to uter (= alter) it is que that is subjoined.

Uterque is, then, originally ‘either’ rather than ‘both’, and it seems that the meaning ‘both’ was evolved from double uses such as we find in: quia uterque utrique est cordi (Ter. Phorm. 800) or: cum uterque utrique esset exercitus in conspectu (B. G. 7. 35). It was easy and usual to omit the second uterque, as in: eodem die uterque eorum ex castris stativis . . . (utrumque) exercitum educunt (B. C. 3. 30. 3), where the eorum pluralizes the verb, and with the omission of utrumque uterque assumes the force of ‘both’. With this meaning it is used in the singular for two individuals, and in the plural for two classes, but in poetry it is often used in the plural for individual objects, as in: palmas utrasque (Aen. 6. 685). Vossius (ad Vell. 2. 34. 3) notices the use in archaic Latin of uter for uterque, as in: utris summo studio pugnantibus (Quadrig. apud Gell. 9. 13. 8), probably a use of uter for alteruter.

Neuter, a trisyllable according to Priscian, is ne+uter (= alter). It is joined with a following utri (= alteri) in: neuter utri invidet (Pl. Stich. 731); but oftener with a following alter, as in: neutra alteri official (Quint, 1. 1. 14), and Quintilian’s usage shows that his Latin is not neuter neutrum diligit, but neuter alterum diligit; following which most editors have changed utri to alteri in the verse of Plautus cited above. Like alter it soon passes on from two to three, designating usually an excluded third, as in: quid bonum sit, quid malum, quid neutrum (Cic. Div. 2. 10. 4). In this use it comes to designate the neuter gender and the neuter verb; and by an easy and usual abbreviation 50 we have neutra verba for verba neutrius generis, and neuter anguis (Cic. Div. 2. 62. 29), i.e. anguis nec mas, nec femina. In this new use its genitive is no longer neutrius, but neutri. This passage to a meaning that obscures its relation to alter is easy; and in this meaning it often ceases to have any connexion with two or more, passing to a singular sense.

Noster and vester by their ending -ter designate an opposition of ‘ours’ to ‘yours’ and vice versa, analogous to that felt in meum and tuum. Their use for a single person is apparently to be connected with the plurals of Modesty and Majesty. Noster standing with a proper name seems short for vir nostri ordinis, or for nostras, as in: quisquis es . . . noster eris (Aen. 2. 149-50); or for nobis favens (cf. suus), as in: sin nostrum adnuerit nobis victoria Martem (12. 187). Horace’s use of noster for ego in: per totum hoc tempus subiectior in diem et horam invidiae noster (Sat. 2. 6. 47-8), Acron explains: verba invidorum refert; but in: minime istuc faciet noster Daemones (Pl. Rud. 1245), we should rather say: ‘your friend Daemones’. We have already spoken of the use of vos for tu; in Ov. Her. 19. 62 Burmann reads: pectora nunc iuncto vestra fovere sinu, where vestra would be for tua; but A. Palmer reads nostra. In: crimen amor vestrum (Aen. 10. 188) vestrum seems of Cupid and Venus, though Servius says that some took vestrum for tuum.

Superum is related to super as is alterum to aliter. It seems to have lost much of its comparative force in such uses as: omnes supera alta tenentes (Aen. 6. 787), or: supera ardua linquens (7. 562); and it is easy to understand the formation from such uses of the double comparative superior, just as the use of inferi for the underworld would lead to the use of inferior for Tartara. Super is a comparative from sub (= ex upo), meaning either ‘from beneath’ (i.e. up), or ‘beneath’. With the accusative super is the opposite of subter; but with the ablative it often loses its comparative force, becoming equivalent to de, and losing all idea of ‘two’. Subter is both the preposition opposed to super, and the adverb opposed to supra.

While extra, intra, citra, ultra offer nothing of interest here, contra (= quom-tra) ‘to what extent on the other side’, and so ‘facing’ or ‘over against’, at times loses its force of opposition, and passes to ‘before’ or ‘to’. We read in the Vulgate: peccatum meum contra me est semper (Psal. 51. 3), and: flens orare contra Caesarem coepit (Bell. Alex. 24. 3).

It is to the doubtful point of division in a word like posterus that 51 Sommer attributes the rise of the ending -teros out of the older -eros. Probably this -eros was subjoined to poste, the opposite of ante (older postid and antid); the root pos- is plain in *posne, later pone, and so -teros here came to be regarded as the comparative ending. Words like ἀριστερός from ἄριστος, which gave rise by analogy to δεξιτερός instead of the older δεξιός, would aid in this development. Of the pair dexter and sinister in Latin, the origin of the former seems clear: it is the welcoming hand (cf. δέχομαι) as well as the hand with which the orator gesticulates (cf. δείκνυμι); my old teacher, Studemund, thought that sinistra marked the hand the speaker held in sinu; and I know of no other probable etymology.

‘Accompanying you’ and ‘at your side’ (sequos) gives the sequester (= sequent-ter). On the analogy of this term of local significance we get words like equester and pedester, paluster and campester; terraster, rurester in Apuleius, and tellustris in Capella. Sommer thinks that positives like agrestis and caelestis helped to extend this formation, The ending -iester seems reduced to -ister in magister, minister, and sinister; that the last is of late development is indicated by its double comparative sinisterior, which probably follows deterior in formation.

What of words like poetaster? Beside the avunculus or ‘little grandfather’, the elder brother who at the father’s death took his place as the protector of the sister and her children, we have his wife the matertera (= matritera, Walde) or ‘second mother’, who takes the place of the sister, should she die; and should this happen, the filia is the filiatera or filiatra of the matertera. On the analogy of words in -aster and -ister the son would now be the filiaster of the uncle, who in turn would be his patraster—neither the son nor the father in the full sense of the term. So we get calvaster ‘tending to baldness’ and surdaster ‘tending to deafness’, pilaster, oleaster, lotaster, poetaster. Our mulatto and the French mulâtre testify to a Low Latin mulaster. We read in Plautus’s Epidicus: sed quis haec est muliercula et ille gravastellus qui venit (v. 620). Festus gives this reading, but in another part of his compendium he cites the word as ravistellus—perhaps the oldest variant on record for Latin literature. Gravastellus is evidently a diminutive from gravaster ‘the man turning grey’; while ravistellus is a like diminutive from ravister, a pejorative of ravus ‘gray’ following the analogy of minister. We see that beside ravus, the classical form, there existed an older gravus, German grau, our ‘grey’. And so it is possible to think of Roma as a later development 52 of groma, as lactis is probably for an older *glactos. This is confirmed when we compare Nova Carthago (Aen. 1. 366) with Roma quadrata. Nova Carthago in the verse cited is the old, not the new, Carthage; and the epithet novus is joined with it here because the word Carthage itself meant New Town. So perhaps quadrata is joined with Roma, because Roma was itself the ‘square’ of the gromaticus.



For the syntax of the dual we are most indebted to Berthold Delbrück, who, starting from the view that it was primarily intended to designate natural pairs like ὄσσε, ὀφθαλμώ, χεῖρε, πόδε, μηρώ in Greek, showed how it passed on to two objects usually found paired, like ἵππω or βόε, the horses or oxen of the span, or δοῦρε, the pair of spears carried by the Homeric warrior. But when it passed on from pairs to the expression of any two objects not paired in such fashion, it was then used for the plural, and this called forth in return a use of the plural for the dual. Delbrück notes an example of this in:

καὶ γὰρ νῦν δύο παῖδε, Λυκάονα καὶ Πολύδωραν,

οὐ δύναμαι ἰδέειν Τρώων εἰς ἄστυ ἀλέντων (Il. 22. 46),

where the δύο παῖδε are not a pair, but two of the fifty sons of Priam. So in Il. 3. 246 the ἄρνε δύω are any two lambs from the flock. The plural is properly opposed to the singular, its function being the expression of more than one; and so here we have an open invasion of its realm by the dual.

The dual is often used to express several pairs, as in:

ὄσσε δ’ ἄρα σφέων

δακρυόφιν πίμπλαντο, γόον δ’ ὠΐετο θυμός (Od. 20. 348-9),

or in:

Ξάνθε τε καὶ σύ, Πόδαργε, καὶ, Αἴθων Λάμπε τε δῖε

νῦν μοι τὴν κομιδὴν ἀποτίνετον (Il. 8. 185-6),

or in:

κούρω δὲ δύω καὶ πεντήκοντα

κρινάσθων κατὰ δῆμον (Od. 8. 35-6).

In: δύω δέ τέ οἱ Θύραι εἰσίν (Od. 13. 109) we have the plural thus used, and δύω does not emphasize the duality, but gives the number of pairs.

In: λεύσσετον πάντα (Aesch. Eum. 255) we have the dual used of the chorus of twenty-four, which falls into two ἡμιχόρια; and in: καὶ περὶ τούτου πάντες ἑξῆς οἱ σοφοὶ πλὴν Παρμενίδου συμφερέσθων, Πρωταγόρας τε καὶ Ἡράκλειτος καὶ Ἐμπεδοκλῆς, καὶ τῶν ποιητῶν οἱ ἄκροι τῆς ποιήσεως ἑκατέρας κτλ. (Plato Theaet. 152 E) we have the dual used 54 for a plurality of persons falling into two divisions, not paired nor necessarily equal in number. In Russian I find that the dual is used after the numerals two, three, four, and both. In:

αἴ κ’ ἀποκηδήσαντε φερώμεθα χεῖρον ἄεθλον (Il. 23. 413).

Lang and Leaf translate: ‘if through heedlessness we win but the worse prize’; and at first sight one might suppose that ἀποκηδήσαντε was here used of three, Antilochus and his team. We should, however, rather, translate: ‘but if through the heedlessness of you two we win’, &c., recognizing here a fine example of the Σχῆμα καθ’ ὅλον καὶ μέρη. In Latin we find:

Atridas Priamumque et saevum ambobus Achillem (Aen. 1. 458),

where it seems at first sight that ambobus is used for three. Servius’s note runs: atqui tres dixit; sed Atridas pro uno accipe, quos unius partis constat fuisse. But the union here with Priamum leads me rather to accept Atridas for one, it is true, but to take that one as Agamemnon.

We have already noticed:

μή πως, ὡς ἀψῖσι λίνου ἁλόντε πανάγρον,

ἀνδράσι δυσμενέεσσιν ἕλωρ καὶ κύρμα γένησθε (Il. 5. 487-8).

Here the Scholiast, possibly repeating Aristarchus’s explanation, explains ἁλόντε as in agreement with σὺ καὶ οἱ ἄλλοι, which he supplies as the subject of γένησθε. But plainly there is no idea of pairing necessarily implied here, and we have a use of the dual for the plural with nothing to palliate it. In:

ὅτι νῶϊν ἀνήκεστος χόλος ἔσται (Il. 15. 217),

νῶϊν is evidently not of two, but of five: Poseidon, Athene, Hera, Hephaestus, and Hermes. In:

ὡς δ’ ὅτε χείμαρροι ποταμοὶ κατ’ ὄρεσφι ῥέοντες

ἐς μισγάγκειαν συμβάλλετον ὄβριμον ὕδωρ (Il. 4. 452-3),

where we have the two armies of the Greeks and Trojans compared to two wintry streams that fling together their stormy waters, we have the two streams in the plural, the verb in the dual, and the waters in the singular. In Latin the plural octo is an old dual denoting two groups of four each, and viginti is ‘two tens’.


εἴπερ γάρ κ’ ἐθέλοιμεν Ἀχαιοί τε Τρῶες τε

ὅρκια πιοτὰ ταμόντες ἀριθμηθήμεναι ἄμφω (Il. 2. 123-4),

ἄμφω is used not of two individuals, but of two nouns of multitude.

So in:

Se satis ambobus Teucrisque venire Latinisque (Aen. 7. 470).


In prose we often find ambo for bini as opposed to singuli, as in: oculi vel ambo vel singuli (Cels. 6. 6. 14). In an enumeration of warriors Virgil has Assaracique duo (Aen. 10. 124) but in the next verse is: germani Sarpedonis ambo, where ambo is merely a poetic variety for duo. In return we find him using duo for bina in: duo quisque Alpina coruscant gaesa manu (Aen. 8. 661-2), where the dual would be used for a number of pairs. From such uses it is easy to explain Virgil’s use for twelve of now bis seni in: pueri bis seni quemque secuti (Aen. 5. 561), now bis sex in: bis sex thoraca petitum perfossumque locis (11. 9-10). When we compare: tum pendere poenas Cecropidae iussi, miserum, septena quotannis corpora natorum (6. 20-2) with Theseus’s words in Euripides:

σώσας κόρους

δὶς ἑπτά, ταῦρον Κνώσσιον κατακτανών (H. F. 1326-7),

we see a use of septena corpora for bis septem corpora, which we shall consider further in a following chapter.

The plural is often used for the dual. By this I do not mean to assert that the plural is often used in Greek or Latin to express two objects. For two objects not forming a pair the plural, not the dual, is the proper idiom. Indeed, even when the plural is used for a natural pair, like oculi or humeri, we are not sure that there ever was a time when it was not properly so used. An examination of the inflexions of the dual makes it probable that the dual appeared later than the plural, and in natural language was never constant in use, but always sporadic. True, in Sanskrit it is used with much regularity and constancy, and I am told the same is true of Gothic; but these are both book languages like mediaeval Latin. When I read in Homer: δύω κανόνεσσ’ ἀραρυῖαν (Il. 13. 407), I have no right to say that κανόνεσσι is a dative plural used for an older dative dual κανόνοιιν: Homer has no such dual. Of natural pairs ὄσσε is the only one he uses always in the dual; he never uses πόδε, though he has ποδοῖιν; χεῖρες is far commoner with him than χεῖρε, and in:

καί ῥ’ ἀπομόρξατο χερσὶ παρειὰς φώνησέν (Od. 18. 200),

‘and with her two hands she wiped off her two cheeks and spake’, Ohler explains that when two duals meet thus in a phrase, for both of them a plural is used. Ohler seems to think there was a time when all natural pairs found expression by the dual only; but there seems little ground for this belief.

Gottfried Hermann thought δύω was a mark of the later use of the dual, as opposed, to the natural use; that ἵππω meant the team, but 56 δύω ἵππω any two horses from the herd. But uses like δύω Αἴαντε (Il. 18. 163), Ἀτρείδα δύω (Il. 1. 16) make me feel that the effect of δύω, as of ἄμφω, is merely to emphasize the duality, just as una emphasizes ambo in: una ambo abierunt foras (Ter. Eun. 702).

But words like θύραι, or foras, or equae, which were once duals, but are now plurals, seem certain examples of the use of the plural for the dual. Words like Μοῦσαι, or νῖκαι, or mensae, or animae, may never have been duals, but may be merely later plurals formed on the analogy of θύραι or equae. But elliptical plurals like Αἴαντες (Il. 7. 164) for Ajax and Teucer, Castores for Castor and Pollux, δεσπόται for ‘the master and mistress’, fratres for ‘brother and sister’ may be asserted as uses of the plural for an older dual.

Some editors have suspected such a dual in: patres natosque videbit (Aen. 2. 579). But Leda is certainly dead by this date, as is probably Tyndareus; even if he is still thought of as living, the idea that patres here is for Helen’s father and stepmother hardly seems to suggest the height of domestic felicity here implied. We have seen how liberis turned the preceding coniuge to the plural in Cicero (Att. 8. 2. 3); probably in like fashion natos here has changed patrem to patres, which is here for Menelaus.

But no editor of Virgil, to my knowledge, has suspected an old dual in: vidi Hecubam centumque nurus (Aen. 2. 501). Heyne’s note is: centum nurus latius dictum; quinquaginta enim erant filiae, totidem filiorum uxores seu nurus. But κουράων . . . δώδεκ’ ἔσαν τέγεοι θάλαμοι (Il. 6. 247-8) clearly implies that Priam and Hecuba had only twelve daughters. Servius has five explanations to offer, each more improbable than the preceding; he begins: aut finitus est numerus pro infinito ὑπερβολικῶς, and ends by referring the centum to the following per aras, though he finds that in v. 523 aras has become haec ara, which would be a transition from a hundred to one. (Aras in v. 501 is a plural for a singular, following the analogy of altaria.) It is small wonder that he adds: est autem haec plena adfectu et dolore repetitio. Only his first explanation has any probability in itself, and it seems barred by the quinquaginta thalami in v. 503. But we read:

infelix qua se, dum regna manebant,

Saepius Andromache ferre incomitata solebat

Ad soceros (Aen. 2. 455-7),

which Virgil evidently hoped would explain to his readers his parallel use of nurus less than fifty verses after. For soceros and nurus are correlatives, and soceros is certainly for socerum et socrum, as Heyne 57 saw. In nurus we have a plural used for an old dual, which stands, not for one pair, but for fifty pairs, each of a filius and a nurus. In such a case the masculine form is usually preferred, but here we have the close association with Hecuba to determine the gender. Then from the story told by Deiphobus in the sixth Aeneid it is clear that it was in effect a ballroom into which the Greeks were so rudely intruding.

Another plural used like the Greek elliptical dual is: geminosque Triones (Aen. 1. 744 and 3. 516), where there is general agreement among scholars that it is the Great and Little Bear that are meant. Septentriones is shortened to Triones, as is Acroceraunia to Ceraunia (Aen. 3. 506). In: septem subiecta trioni (Geo. 3. 381) we have a tmesis that shows how easy this shortening was; and in Cicero we have for the Little Bear: Minor Septentrio (N. D. 2. 111. 43). In: egressi superant fossas (Aen. 9. 314) fossas seems a similar elliptical plural for: vallum et fossam; we have already noticed geminos sub rupe Quirinos (Juv. 11. 105).

Bentley mentions as a crux for grammarians: fortissima Tyndaridarum (Hor. Sat. 1. 1. 100). Is Tyndaridae, of which we have here the genitive, a nom. pl. masc. of Tyndarides, and so a dual in form merely, but really a plural in meaning, being ‘the four children of Tyndareus’? or is it the plural of Tyndarida, the Latin form of Tyndaris, and so dual in meaning as well as in form, being the two daughters of Tyndareus? Bentley favours the former view, notwithstanding the fortissima, which must have made the phrase seem feminine in meaning to the Roman reader. Lucian Müller, favouring the same view, quotes dulcissime rerum (Sat. 1. 9. 4) as a parallel for the lack of agreement in gender. Wickham goes so far as to warn the student against connecting fortissima in meaning with liberta (v. 99), with which it seems in agreement; for it is surely her exploit Horace is lauding here. When we recall the old Roman use of cratera for κρατήρ, and that of Ancona for Ἀγκών, we shall see that the Roman would naturally use Tyndarida, the Greek accusative form, as the nominative for Τυνδαρίς. This seems a case where from the Latin Scholia we may learn how the Romans themselves understood the phrase. Porphyrio’s note is: At hunc, etc., Eleganter, quia Clytaemnestra Tyndarei filia fuit, quae Agamemnonem maritum securi percussit; Acron’s: Tyndaridarum, Graecarum mulierum, pro Tyndaridum; significat autem Clytaemnestram aut Helenam. Nam Clytaemnestra Agamemnonem, Helena Deiphobum interfecit, 58 et ambae Tyndarei filiae. Acron’s scholium plainly points to Tyndaridarum as the gen. pl. of Tyndarida, used for Τυνδαρίς. But Tyndaridae here seems used to designate the class of women who kill their husbands, a class that finds its prototype, according to Acron, in either Clytaemnestra or Helen. Fortissima of course belongs to liberta, to whom Horace awards the palm for vigour in her class; divisit medium, making a clean and clear cut. Tyndaridarum, then, is neither a masculine plural nor a feminine dual, but a feminine plural.

With the Homeric use of Αἴαντες for ‘Ajax and Teucer’ we may associate the use of Ἕλληνες for ‘Hellen and his people’ and Teutones for ‘Teuto and his people’ noticed by Hirt (Gr. L.- und F.- Lehre, p. 213). He further cites as a parallel the Skt. śvaśurās, ‘the father-in-law and his connexions’; and Reichelt notices a like use in Avestan (Av. El. p. 221). Along the same line of development seem the Skt. compounds ending in -adi, as in deva indradayas, ‘the gods beginning with Indra’ (Whitney, Skt. Gr. 1302 D). In οἱ περὶ Φαβρίκιον (Plut. Pyrrh. 20) for Fabricius, we have the opposite. So Herodotus has οἱ ἀμφὶ Μεγαρέας καὶ φλειασίους, meaning the same as οἱ Μεγαρέες καὶ φλειάσιοι, used immediately after in 9. 69. Out of the Attic use of οἱ ἀμφὶ Πρωταγόραν for ‘Protagoras and his school’ develops the use of this phrase for Protagoras himself, and from οἱ περὶ Ἡράκλειτον ‘the school of Heraclitus’, or οἱ περὶ Ἀρχίαν πολέμαρχοι (Xen. Hell. 5. 4. 2) ‘Archias and his colleagues’, seems to develop the phrase for Fabricius cited above.



In introducing the cases of the dual Brugmann writes: As regards formation the dual seems to have been a singular of which the essential formative elements originally expressed the quality of being coupled or paired (Vergl. Gr. Il2. 2. 194). In discussing the Irish da fer (= two men) he notices the use of the nom. sing. masc. fer as a dual, as likewise of the masc. sing. tene, and the neut. sings. dliged, tech, ainm, ascribing the use to a likeness in form that developed there. We noticed Sommer’s belief that genu and cornu are old duals that have become singulars, and that he bases the change, not on form, but on the notion of pair in ‘the two knees’, or in ‘the two horns’, making meaning, not form, the cause of the transfer. The use of Castor, or of Pollux, to name both the Twins, seems based not on the form of the word, but on the idea of pair in the Gemini, because of which the name of the one at once suggests to the reader or hearer the name of the other. Meaning seems more important than form in producing this change: if in a dual you emphasize the idea of ‘two’ or division, it will pass to the plural; if you dwell on the notion of pair, it may pass to the singular.

This change from the dual to the singular, though it escaped the notice of the Roman grammarians, seems to me certain for Latin and probable for Greek, though to a less extent. Just as in Latin we find nasus or nasum in the singular and nares in the plural, while in Skt. we have only nāsā in the dual, in Greek we have ῥίς, ἡ, and ῥίν, τό, as well as ῥῖνες for the nose. We read in Horace: gaude quod spectant oculi te mille loquentem (Ep. 1. 6. 19), not, however, of five hundred spectators, for oculus here evidently means a spectator, i.e. a pair of eyes. Usually we have the plural, as in: qui siccis oculis monstra natantia (vidit) (Od. 1. 3. 18); but what of: quisquis ingentes oculo irretorto special acervos (Od. 2. 2. 23-4) or: solus mullisne coheres, veloci percurre oculo (Sal. 2. 5. 54-5)? We read in Horace: non islic obliquo oculo mea commoda quisquam limat (Ep. 1. 14. 37-8), but 60 in Ovid: altera, si memini, limis subrisit ocellis (Am. 3. 1. 33). In: cum tibi sol tepidus plures admoverit aures (Hor. Ep. 1. 20. 19) I feel that the singular of plures aures is una auris (= auditor unus); and I read in Martial: aurem non ego tertiam timerem (12. 24. 10) ‘I should not fear a third pair of ears’. So in Virgil: simul hoc dicens attollit in aegrum se femur (Aen. 10. 856-7) femur must be for utrumque femur; for his wound was imo inguine; (hasta) ima sedit inguine (imo) v. 785-6. More interesting is:

tum pendere poenas

Cecropidae iussi, miserum, septena quotannis

Corpora natorum. (Aen. 6. 20-2.)

Heyne asks: Cum pueri septem septemque puellae mitterentur, quidni alterutrum tantum poni potuit? But he has no answer. Conington’s answer is very close to the mark: ‘The story mentioned seven youths and seven maidens, but Virgil has chosen only to name the former’. But we have here, too, a dual passing to a singular, each corpus natorum consisting of a youth and a maid.

Virgil uses the plural of geminus as an adjective to designate well-matched pairs, as in: gemini . . . inmensis orbibus angues (Aen. 2. 203-4), geminae columbae (6. 190), geminae belli portae (7. 607), geminae somni portae (6. 893), geminae quercus (9. 681), huc geminas nunc flecte acies (6. 788), geminae slant vertice cristae (6. 779). The last two examples are noteworthy; in the last the crests are on the same head, and when pairs are thus connected we commonly have the singular, not the plural, as in: gemina teguntur lumina nocte (Catull. 51. 11), gemino demittunt bracchia muro turriti scopuli (Aen. 3. 535), geminum pugnae proponit honorem (5. 365), gemina super arbore sidunt (6. 203). So we may regard: geminique sub ubere nati (5. 285) as short for: gemini (gemino) sub ubere nati, an abbreviation of four to three, from which hypallage takes its origin; cf. geminos huic ubera circum ludere pendentes pueros (8. 631-2). So, too, with geminum (geminae) pugnae proponit honorem. In: solem geminum et duplices . . . Thebas (Aen. 4. 470) we have, probably, the exchange of number for the sake of variety. That foris in the sing. and fores in the pl. are both used for the pair of doors is plain from: ad geminae limina prima foris (Ov. Her. 12. 150) and: frustra clavis inest foribus (Tib. 1. 6. 34). Along with θύραι ‘the pair of doors’, which Brugmann thinks an old dual, Homer uses θύρη for ‘the door’. Probably foras is the acc. pl. and foris the abl. pl. of an old *fora, corresponding with this; and foris is a new nom. sing. formed on the 61 analogy of fores, the old plural of fora. (Cf. manus ‘good’, and manes ‘the kindly ones’.)

By Homer’s time duals like θύραι had become plurals; in their stead had begun the development of duals in for α-stems following the analogy of duals like ἵππω. We see this beginning in Homer, not for feminines, but for masculines like Ἀτρείδᾶ; but the development is soon extended to feminines. It was an irregular and partial development; for the use of the dual in Greek was passing away. We find side by side ταῖν χεροῖν and τοῖν χεροῖν, τὰ κόρα and τὼ στήλα. Attraction played its part, as we see from ἀμφοῖν τοῖν χεροῖν, but ταύταιν ταῖν ἀδελφαῖν. The form ταῖν seems to have developed for the article before τά; Plato and the Orators have ταῖν ὁδοῖν, but τὼ ὁδώ, Sophocles and Aristophanes write τὰ κόρα, but τὼ χεῖρε. One of the results of the struggle thus prolonged of the dual against the prevailing plural seems to have been that, when the dual disappeared, it was almost always the plural that took its place. There is no trace of such a revival of the dual in Latin; it probably disappeared there more readily and speedily; and so more of its uses passed to the singular.

Our most interesting example of this use of the singular for the dual occurs in Horace’s first Canidia epode, where the boy, whom the witches are about to kill to obtain an unguent from his marrow, after a vain appeal to their feelings as women, in a second speech threatens them with the vengeance of gods and men. Its opening words have proved a crux for all editors:

Venena magnum fas nefasque non valent

Convertere humanam vicem (Epod. 5. 87-8).

To this text, given by all manuscripts of any value, Bentley adds the note: Durissimus locus, neque, quocumque modo vertas et excutias, sententiam commodam praebens. Scio equidem ut conati sint explicare veteres novique interpretes; sed si quid video laterem lavarunt. . . . Multa quidem ipse nequidquam tentavi, quae piget hic referre; tamen, ne omnino asymbolus veniam, dicam aliquid, quod etsi ne mihi quidem placeat, Rutgersiano haud deterius fore credo. Ergo ecce tibi correctionem, qualiscumque est:

Venena magica fas nefasque non valent,

Non vertere humanas vices.

Of our novi interpretes Kiessling reads:

Venena maga non fas nefasque, non valent

Convertere humanam vicem.

In maga non he follows an improvement on Bentley’s magica 62 suggested by Moritz Haupt, and he calls the magnum of the manuscript meaningless (sinnlos); but he does not venture to change the number of humanam vicem.

Turning to the veteres we read in Acron: Venena magnum f. n. n. v. c. h. v. Lex enim humana habet malis poenam, bonis praemia pollicenda, et ideo furens puer dicit, haec eas carminibus mutare non posse. Porphyrio gives us: Ven. magnum fas nefasque. Magnum fas venena sunt, si hostibus dantur, magnum nefas, si amicis. Non valent conv. hum. vicem. Sic sensus est: Quamvis venena multum possint, non tamen valent merita in contrarium vertere, ut liberentur poena, aut mala mereantur. Vices autem appellantur poenae, quae in scelerosis admissis regerentur.

It is evident that both accept the text cited by Bentley as that of all good manuscripts, and it might seem significant that both give in full magnum, the word rejected by Bentley and found meaningless by Kiessling. There are two things of the greatest importance to note in Porphyrio’s second scholium: (1) in his explanation he substitutes vices for the vicem of his text, (2) in arranging Horace’s words he joins non valent with convertere. We have in the distich non valent in the first verse, and convertere in the second, giving us a not very difficult example of distribution; we must fill out each of these, understanding for each convertere non valent. So we have: venena magnum fas nefasque convertere non valent; convertere non valent humanam vicem (= vices, Porph.). Acron’s note gives us the true meaning of humanam vicem: it is a singular where prose would use a plural; a singular for a dual, the praemia and poenae of the lex humana, which here balance the fas nefasque of the gods. It is put in the singular to vary the diction of the couplet, just as in: et solem geminum et duplices se ostendere Thebas (Aen. 4. 470) solem geminum is put in the singular to give variety to the verse.

Why do Bentley and Kiessling feel magnum to be inadmissible, in face of such clear testimony from Scholia and manuscripts? Venena ait magica, says Bentley, id enim epitheton necessarium hic videtur; quippe venena per se et absolute posita non possunt rem magicam denotare. If venena were here used in its proper and absolute sense of ‘drugs’, one might readily supply magica from the theme of the whole poem, which is of the maga Canidia; and Horace might well deem its expression here unnecessary. But Bentley is hopelessly prosaic; usually in poetry words of importance are not used ‘per se et absolute’; and Acron tells us implicitly that venena is here used 63 for carmina ‘spells’. Magnum is ‘sinnlos’, thinks Kiessling. True, it is not here applied to a material object, the size of which can be determined by the modius or the decempeda. But magnus is at times applied to animus; we have them connected in magnanimus ‘high-souled’. From the humanam of the second verse we might have expected divinum in the first, but divinum is already implied in fas as opposed to lex. Convertere means ‘to change wholly’, or ‘to reverse entirely’. In balancing magnum fas nefasque by humanam vicem, the poet has chosen for variety to use the singular, not the plural, to express the dual idea of the rewards and penalties of human law. We shall translate then: Your spells have not the power wholly to reverse the right and wrong of high heaven; they have not the power to reverse even the rewards and punishments of men.

It seems strange at first sight that Brugmann is willing to accept cornus (old cornuus) as for an older gen. dual cornuos corresponding to the Skt. gen.-loc. dual in -os, but is unwilling to accept cornu as the old nom. dual. A similar irregularity is noted, however, for Polish by Delbrück; while reçe, the old nom. dual, ‘the two hands’, has become a plural, reku, its gen.-loc., has passed to the singular (Vergl. Synt. I, p. 145). From examples like: iam cornu petat et pedibus qui spargat harenam (Buc. 3. 87 et Aen. 9. 629), cornu ferit ille (Buc. 9. 25), qui vexat nascenti robora cornu (Juv. 12. 9) it is plain that the singular cornu in Latin stands at times for a pair of horns. The plural is far more common; as is genua in: tarda trementi genua labant (Aen. 5. 431-2) but we read: nuda genu nodoque sinus collecta fluentes (1. 320), and: impressoque genu nitens terrae adplicat ipsum (12. 303). In:

quamquam tardata sagitta

Interdum genua impediunt cursumque recusant (12. 746-7),

Wagner relying on inferior manuscripts changed tardata to tardante, and he is followed by Forbiger and Ladewig. Only one of Aeneas’s knees is affected by the arrow, and interdum seems to show that it is this one knee which at times impedes him in the race. Ribbeck has retained tardata, as all the manuscripts on which he relies give this reading, which is clearly the more difficult; and this is the reading of Conington and Sidgwick, who do not explain Wagner’s difficulty in translation, however. Probably we have here a use of genua for genu, the opposite of genu for genua.

We read: et gemina auratus taurino cornua voltu (Geo. 4. 371). We have also geminae nares (Geo. 4. 300), geminas aures (Culex 64 150), geminos lacertos (Moretum 21), geminas acies (Aen. 6. 788), showing that gemini is felt to be the numeral appropriate to pairs. While neither duo nor ambo is joined with the singular, we noticed solem geminum (Aen. 4. 470) ‘twin suns’, gemino muro (3. 535) ‘twin walls’, geminum honorem (5. 365) ‘a pair of prizes’, gemino ab ovo (Hor. A. P. 147) ‘from the twin eggs’, and geminae foris (Ov. Her. 12. 150) ‘a pair of doors’. It is reasonable to place here: geminus Pollux (Hor. Od. 3. 29. 64) and geminus Castor (Ov. A. A. 1. 746). If the Latin dual passes into the singular at times, it is only to be expected that the numeral for two appropriate to the dual shall also appear as a singular.

The dual is used as the singular at times, in return for the uses of the singular for the dual. I have already noticed the use of Atridas (Aen. 1. 458), and of tardata sagitta genua (Aen. 12. 746-7). Of course in Latin it is the plural used for the dual that is in question.

In poetry a part of the body is often expressed by the plural, where we should expect the singular. We read: coniugis ille suae complexus colla lacertis (Ov. Met. 1. 734); why colla? Many parts of the body occur in pairs; e.g. oculi, aures, genae, manus, pedes, genua; and after analogy of such parts it became a poetic fashion to use the plural even for such parts as were not paired, like cervix, collum, mentum, dorsum. Often such uses are plainly to vary the diction, as in: nunc animis opus, Aenea, nunc pectore firmo (Aen. 6. 261), with which compare: violenta pectora Turni edocet (10. 151-2) and: his animum arrecti dictis (1. 579). Here animis is the mind of Aeneas, but animum the minds of Aeneas and Achates; while both pectore and pectora are of a single person. In: manus iuvenem . . . post terga revinctum (Aen. 2. 57) terga of a single person seems to follow the analogy of pectora, a paired part. So in Homer we read:

τὸν . . . ὁ μὲν ἰῷ

βεβλήκει τελαμῶνα περὶ στήθεσσι φαεινόν (Il. 12. 400-1),


ἥ τ’ ἀνὰ νῶτα θέούσα διαμπερὲς αὐχέν’ ἱκάνει (Il. 13. 547),

where νῶτα seems to follow the analogy of στήθεα in like manner.

Still more noteworthy is the use of the dual in Sanskrit in the double elliptical form Mitrā Varunā, the two Mitras, the two Varunas, for Mitra and Varuna; or in pitarau-mātarau, the two fathers, the two mothers, for the father and the mother. We find the same idiom in Mithra Ahura in the Avesta (Reichelt, Av. Synt. p. 222). In: lugete, O Veneres Cupidinesque (Catull. 3. 1) Schwyzer (Ig. F. 14, p. 28) 65 thinks that the plurals Veneres Cupidinesque are for old duals, and mean merely Venus and Cupid.

The development of this construction seems interesting, and can be traced with the help of Anglo-Saxon and old Irish. In Skt. the dual Mitrā is used for Mitra and Varuna, just as in Greek Ἀτρείδα for Agamemnon and Menelaus (Il. 1. 16), or in Latin Castores for Castor and Pollux. But at times uses like Mitrā ‘the two Mitras’ did not convey clearly enough the name of the second member of the pair, and so the second name would be added, at first in the singular. So we read: ā yad ruhāva Varuṇaś ca navam (R. V. 7. 88. 3) ‘when we two and Varuna get on board the ship’, with the meaning ‘when I and Varuna get on board’, &c. We have the same syntax in the Anglo-Saxon phrase: wit Scilling ‘we two Scilling’ for ‘I and Scilling’. Zimmer gives us from Irish (K. Z. 32, p. 152 ff.): doronsat sid ocus Fergal ‘they made peace and Fergal’ for ‘he and Fergal made peace’; a syntax reflected in the Latin life of Fintan: in illo autem die ante vesperum venit Fintanus ad consilium, et salutaverunt se in vicem et Lasserianus. But instead of advancing to two duals as in Sanskrit, or to two plurals as in Latin, Irish changes the verb back to the singular, influenced, Zimmer thinks, by an impersonal use of the verb; and we have a form which Zimmer translates thus: She came from the East in the shape of two swans and her maid.

In Greek Wackernagel cites from Homer:

ἐς δ’ ἐνόησ’ Αἴαντε δύω, πολέμον ἀκορήτω

ἑσταότας, Τεῦκρόν τε νέον κλισίηθεν ἰόντα (Il. 12. 335-6),

where he thought Αἴαντε δύω . . . Τεῦκρόν τε parallel to the āvām Varuṇaś ca of the example cited from the Rig-Veda, where he amplifies āvām from the -va in ruhāva. And this view seems confirmed by:

τὶν δ’ ἐν Ἰσθμῷ διπλόα θάλλοισ’ ἀρετά,

Φυλακίδα, κεῖται, Νεμέα δὲ καὶ ἀμφοῖν

Πυθέᾳ τε, παγκρατίου (Pind. Isth. (5). 17 ff),

‘further, Phylacides, a double crown of glory is at Isthmus stored and at Nemea both for thee and for Pytheas, a pancratiast’s crown’ (Myer’s transl.). The Scholiast explains ἀμφοῖν Πυθέᾳ τε by ἀμφοτέροις ὑμῖν, σοί τε καὶ τῷ Πυθέᾳ, making it parallel with Αἴαντε . . . Τεῦκρόν τε. But the Greek never advanced to Αἴαντε Τεύκρω τε as did the Sanskrit to Mitrā Varunā.

It seems clear that this advance in Sanskrit was due to an assimilation of the second to the first in number. So in Latin from an older 66 Veneres Cupidoque, where Veneres primarily meant Venus and Cupid, we have Veneres Cupidinesque (also in Catull. 13. 12 and Martial, 9. 11. 9 and 11. 13. 6). But the assimilation of the first to the second was also possible, as we find it developing in Irish from: they and Fergal, to: she . . . and her maid. This is what takes place in Greek, giving us the Schema Alemanicum, as we see it in:

ἧχι ῥοὰς Σιμόεις συμβάλλετον ἠδὲ Σκάμανδρος (Il. 5. 774),

where συμβάλλετον becomes intelligible only when we think of its primary subject as Σιμόεντε, ‘the Simois and Skamander’, which, after Σκάμανδρος was added, passed back to Σιμόεις under the influence of its form and meaning.

Of course it would be absurd to think of the Latin construction here as directly influenced by the Skt. syntax. Gauthiot, in the paper Du Nombre Duel already cited, shows that this curious doubling of the dual is not uncommon in Finnish; and from its development in languages so independent of each other we may conclude that it is a natural and easy development of the dual wherever found. Other peculiarities of the dual, such as its passage to the singular, may be even more easily assumed as probable developments of the dual number wherever it occurs. (See Appendix A.)

Delbrück holds that the primary use of the dual is for natural pairs, like ὄσσε or χεῖρε. But that this use, though very old, is older than, or even as old as, that for twins, as in the Skt. Aśvinau, is opposed to much that we see in the development of number and gender. In the next chapter we shall see reasons for believing that the plural was used primarily of persons, and was extended to things later. The use of gemini, as the numeral appropriate to the dual in Latin, and accompanying it in its passage to the singular as well as to the plural, tends to confirm this. The sensation, at times of joy, but oftener of horror, called forth by a birth of twins among savages of to-day, may give us some idea of its importance among primitive men. In Russian the use of the dual is extended from dva, ‘two’, and oba, ‘both’, to trī, ‘three’, and chetyre, ‘four’ (Figgis, Russ. Gram., p. 91). The extension to ‘four’ seems the same as that we cited from Il. 8. 185-6; but its use with three seems easiest to understand when we think of the evident association of gemini with tergemini. Delbrück notes that Slovenian, and Upper and Lower Serbian have kept the same construction and almost to the same extent to which it existed in Old Bulgarian; he thinks it an effect of analogy (Vergl. Synt. I, p. 144-5).



Before the threefold system of gender that marks the Indo-Germanic languages there seems to me to have existed a twofold system still found in many languages of more primitive structure and sometimes styled higher and lower. This division has been commonly identified with that between living and lifeless objects, but as a rule women and children belong to the lower class, while the sword and shield of the warrior are found at times in the higher. When we turn to Greek we find that adjectives in rarer use, like ἄλογος, have often only two forms, and that some, like ἴφθιμος, are of either two or three forms in Homer; cf. ἰφθίμους ψυχάς (Il. 1. 3) with θυγατέρ’ ἰφθίμῃ; (Od. 15. 364). So βέβαιος in Thucydides has only two forms for gender, but three in Euripides; χρήσιμος has usually two forms in Attic, three appear in the Orators. In Latin hilaris with two forms changes to hilarus with three. In Roman law women and children are res, as opposed to the free man, the vir sui iuris, who is a persona; and this distinction, which lies at the basis of Roman law, may reasonably be held to represent a distinction of great importance and antiquity in primitive Indo-Germanic society. As we have already noted, there is strong reason for assuming that before the threefold distinction for number which seems to have marked Indo-Germanic, there existed a twofold distinction of the singular and the plural. Older grammarians held that ‘while there may be multeity in things, there can be plurality only in persons’ (Farrar, Gr. Synt. p. 65, note). Irregularities in the use of the singular for the plural, that do not arise out of the disappearance of the dual, seem mainly due to this primary restriction of plurality to persons.

In many of the languages that have only two genders, a higher and a lower, nouns of the lower gender form no plurals. In both Greek and Latin words denoting materials, grain, flocks, and herds of cattle show much confusion in their use of the singular and the plural. This confusion is very familiar to us in English, where we say oats, pease, and beans, but wheat and barley. Greek uses for ‘blood’ αἷμα and αἵματα, and Latin cruor and cruores; in Greek we have ζειά and ζειαί, in Latin far and farra; Greek has for ‘fat’ only στέαρ, but Latin adeps and adipes; in Greek ‘dust’ is κόνις, in Latin pulvis and pulveres. 68 But while in Thucydides we have: ἄμπελον κόπτοντες (4. 90. 2), but: λίθοις τε καὶ κεράμω βαλλόντων (2. 4. 2), still there seem traces of a definite tendency here, differing in Greek and in Latin.

For materials Greek tends to use the plural for the general name and the collective sense, while the singular rather means a bit of the material in question. Thus κρέα is usually ‘dressed meat’, as in: κρέα ὠπτημένα (Ar. Plut. 894), while: ἄρνειον κρέας (Pherecr. Δουλ. 1) is ‘a bit of lamb’. But in Latin caro is the general term for flesh, as in: carne pecudum famem propulsare adacti (Ann. 14. 24. 1), while carnes is used for ‘bits of flesh’, as in: pedes pones super concisis carnes (Vulg. Exod. 29. 17). We find many deviations from these tendencies, however, both in Latin and Greek. While σάρκες is the usual word for ‘flesh’, σάρξ is also in use. Ὕδωρ is Homer’s usual word for ‘water’, except for the sea, where he has ὕδατα ἀενάοντα (Od. 13. 109). But later ὕδατα is much more common and varied in use, being for showers in οὐρανίων ὑδάτων ὀμβρίων (Pind. Ol. 10. 2), and for mineral springs in Ὕδατα Σέξτια for the Latin Aquae Sextiae. Homer uses δάκρυ in: δάκρυ χέων (Il. 1. 357) as the collective; and the old grammarians regarded δάκρυα as from δάκρυον, and not from δάκρυ, which seems the older form. While Homer uses either ξύλα or ξύλον for firewood, in later Greek ξύλα is firewood, and ξύλον a plank or bench. It seems clear that in Greek, too, the singular is older than the plural here. In Latin lignum is wood and ligna firewood, and it is natural to regard the specialized term as later than the general.

Turning to grain, we have already cited Bavius’s censure on Virgil for writing hordea for hordeum in imitation of the Greek ἄλφιτα or κριθαί. Later Greek writes ἄλφιτα, but Homer has the singular ἄλφιτον as well; he uses κριθαί thrice, but κρῖ occurs eight times in the Homeric poems. No Greek word ends in θ, and so he cannot write κρῖθ, of which κριθή would be the old form of the plural, κριθαί being a second plural following the analogy of πυροί and ἀλείατα ‘wheat’. He uses πυροί four times, but the singular πυρός five times. He has not ἄλευρα which later Greek uses for wheat. Latin has no plural for triticum, but it uses frumenta for ‘kinds of grain’, as in: tritico vel aliis frumentis (Columella 8. 9. 2). Avena is ‘oats’, but avenae ‘wild oats’ in: steriles dominantur avenae (Geo. 1. 154). Here too the plural seems late, and in Latin rare.

For cattle, where in Greek we have πρόβατα, or κτήνεα, pluralia tantum, in Latin pecu is the generic term. This has the plural pecua or pecuda, later pecudes, for members of the herd, and pecora for 69 herds. Latin pluralizes equus and bos freely as does Greek ἵππος and βοῦς; but, unlike Greek, for the smaller creatures of the farmyard it tends to use the singular as a collective, as in: Villa abundat porco, haedo, agna, gallina (Cic. Sen. 56. 16).

We see the same respective tendencies in Greek and Latin, when we turn to bodies of men. In Homer λαός for a body of men, as in λαὸν ἀγείρω (Il. 16. 129), is far more usual than the plural in λαοὶ . . . ἀγροιῶται (Il. 11. 676); but in Attic Greek the reverse is the case. Plato has still ὁ πολὺς λεώς, but the herald’s cry: ἀκούετε λεῴ, ‘hear, O people!’, our Oyez, oyez, shows the prevalence of the plural. In Aristophanes λεῴ seems the equivalent of plebs; in N.T. Greek λαοί becomes specialized as the opposite of κλῆροι. In Latin populus is pluralized only to designate the peoples of different states, till we reach the African Latin of Augustine and Apuleius. Plebs is the collective, having plebeii for individuals, and is never pluralized till Columella uses it for swarms of bees in: duas vel tres alveorum plebes (9. 11. 1). But στρατός is singular like exercitus; στρατιώτης like miles is individual or collective; and we find ὁ πεζός and ἡ ἵππος like pedes and eques for infantry and cavalry. Manus is the hand, and the band of hands. Uses like: ὀκτακισχιλίην ἀσπίδα (Herod. 5. 30), or: παρμένοντας αἰχμᾷ (Pind. Pyth. 8. 58) ‘standing by their spearmen’, have their nearest and perhaps only parallel in Cicero’s: hasta infinita (Phil. 4. 9. 4) ‘endless auction sales’. The use of pondo for librae arises from a use parallel to that of Pollux for Castor et Pollux, or of uter for alteruter, that of pondo for pondo librae ‘pounds in weight’, where pondo is the ablative of an old pondum, which later became pondus.

The neuter plural in Greek is regularly constructed with a singular verb. The exceptions to this rule are in the proportion of one to three in Homer, but become fewer in later Attic. In many of the Homeric exceptions the neuter plural is accompanied by a numeral as in:

οὐδ’ εἴ μοι δέκα μὲν γλῶσσαι, δέκα δὲ στόματ’ εἶεν (Il. 2. 489)

Many can be explained by the σχῆμα πρὸς τὸ συνώνυμον as in:

ὣς τῶν ἔθνεα πολλὰ νεῶν ἄπο . . . ἐστιχόωντο (Il. 2. 91-2),

where ἔθνεα is constructed as though it were λαοί. The rule is pretty closely followed in Plato and the Drama; but Thucydides and Xenophon show many exceptions, mainly of the classes noted above, as in: τὰ τέλη (= οἱ ἄρχοντες) τῶν Λακεδαιμονίων . . . αὐτὸν ἐξέπεμψαν (Thuc. 4. 88. 1) or: φανερὰ ἦσαν καὶ ἵππων καὶ ἀνθρώπων ἴχνη πολλά (Anab. 1. 7. 17).

While in the accusative absolute we should have the singular here, 70 as in: δόξαν ἡμῖν ταῦτα (Plat. Prot. 314 C), we find also the plural, as in: δόξαντα δὲ ταῦτα καὶ περανθέντα τὰ μὲν στρατεύματα ἀπῆλθε (Xen. Hell. 3. 2. 19), where the δόξαντα seems attracted by the following ταῦτα, just as is ἀδύνατα in: ἀδύνατά ἐστι ταῦτα ποιείν. Just as the use of ἀδύνατα for ἀδύνατον is extended by analogy to phrases where there is no following plural to attract, as: τὴν πεπρωμένην μοῖραν ἀδύνατά ἐστιν ἀποφυγεῖν (Herod. 1. 91), so here, too, we have the plural in the acc. abs. with no following neuter plural, as in: δεδογμέν’, ὡς ἔοικε, τήνδε κατθανεῖν (Soph. Ant. 576). This construction of the neuter plural with a singular verb is not found in classical Latin, but occurs in Plautus: quae imperasti . . . factumst ilico (Bacch. 726).

‘There can be plurality only in persons’; is the plural in -a really a plural to begin with? or merely an augmentative? In Homer we find from μηρός the plural μηροί for the separate thighs in: μηρούς τ’ ἐξέταμον (Il. 1. 460), but in: αὐτὰρ ἐπεὶ κατὰ μῆρ’ ἐκάη (v. 464) μῆρα designates the pile of thighs heaped upon the altar. This view is strengthened by the Homeric use of ὕδατα for the ocean, and of κέλευθα in: ἐπέπλεον ὑγρὰ κέλευθα (Il. 1. 312) ‘they sailed the watery ways’. The use of οἰκία, a neuter plural in Homer, for a single dwelling, and its transition to a feminine singular in Attic, both point strongly in the same direction. There seems good reason to think that abstracts in and in Greek and in -ia in Latin are really this old neuter plural in a collective sense, used later as singulars to denote the abstract idea deduced from the collection in question.

Just as τὰ οἰκία, a plural of parts, is used to designate a single house in Homer, like μέγαρα ‘the hall’, and then passes to ἡ οἰκία in Attic with a new plural αἱ οἰκίαι, so ἀναλκείη ‘a lot of weak acts’ has already in Homer attained the force of ‘weakness’ and has formed a new plural ἀναλκεῖαι ‘acts of weakness’. A like collective in Latin is Italia, which means (1) the collection of Itali, (2) the land where they dwell. Prose usually admits only the second sense of such collectives, but poetry often shows them in their primary meaning, as in: Graecia Barbariae lento collisa duello (Hor. Ep. 1. 2. 7) ‘the Greeks ground down in slow warfare with the Trojans’. It is not often that in Latin literature we can see the old plural Graecia passing into the feminine singular; or the collective sapientia ‘a gathering of sages’ passing into the abstract sapientia ‘wisdom’. But in Sallust’s: interea servitia repudiabat, cuius initio ad eum magnae copiae concurrebant (Cat. 56 fin.), does not the syntax of cuius with servitia make it appear a singular rather than a plural?


When in Tacitus we read: Firmius Catus senator, ex intima Libonis amicitia (Ann. 2. 27. 2) ‘of the intimate friends of Libo’, or: plures seditioni duces (Ann. 1. 22. 1) ‘the mutineers get more leaders’; or in Propertius: turpior et saecli vivere luxuria (1. 16. 12) ‘to live in greater shame than the wantons of her time’; or in Virgil: adsit laetitiae Bacchus dator (Aen. 1. 734) ‘let the winegod, bestower of joy, cheer the joyous throng’, we understand the force of virtus in: exigui numero, sed bello vivida virtus (Aen. 5. 754) ‘few in number, but a band of heroes strong for war’; or in: non aliter socium virtus coit omnis in unum (Aen. 10. 410) ‘so all the hero-band of allies gathers together’. It is plain that the collective sense in point is not confined to abstracts in -ia, but may extend to others. We read: ianua Tarpeiae nota pudicitiae (Prop. 1. 16. 2) ‘the well-known gate of the virgin Tarpeia’, where the abstract is transferred to a single person; and now we can understand: spes quoque suas ambitioni donant ‘they consecrate their young hopefuls, too, to politics’; . . . cruda adhuc studia in forum pellunt ‘they hurry the raw schoolboys into the lawcourts’ (Petron. Sat. 4).

Draeger thought it worth his while to collect all plurals of abstract nouns that he found in Latin, and he tells us that of 3,814 found, only 2,500 are used in the plural. Of such plurals rarest seem those which are used in the abstract sense, as in: virtutes voluntariae, quae quidem proprie virtutes appellantur (Cic. Fin. 5. 38. 13). Far more usual is the plural in a concrete sense as in: heroum mira virtutes indicat arte (Catull. 64. 51) ‘exhibits with wonderful skill the deeds of heroes’. So color, the abstract quality in: qui color, nitor, vestitus, quae habitudost corporis (Ter. Eun. 242), passes to ‘blossoms’ in: aurave distinctos educit verna colores (Catull. 64. 90), and to ‘gauds’ in: furtivis nudala coloribus (Hor. Ep. 1. 3. 20). Vitia is usually concrete, in the sense of ‘twists’ or ‘perversities’. Rarest is the transference to the person exercising the quality: artes is commonly used of the various arts, or the works of art in which they are shown; rare and late is the use of ars for the book teaching them, as in: volvitque Palaemonis artem (Juv. 6. 452), and still rarer its use for those skilled in them, as in: Mnemosyne Iovi fecunda novies Artium (i.e. Musarum) peperit chorum (Phaedr. 3, prol. 18-19).



Homer uses οἰκία and μέγαρα, neuters plural, for a single house and a single hall. As in Greek we have οἶκος ὁ and οἰκία τά for a single house, so in Sanskrit we have gṛham, a neuter singular, and gṛhās, masculine plural, for a house; and Boehtlingk-Roth explain this use of gṛhās as of a house consisting of several structures or rooms. So in Greek we have δῶμα and δώματα for a house; in Latin aedes in the singular is the temple, a single room with its hearth or altar (cf. αἴθω ‘I burn’), but in the plural a house, which consisted of several rooms. In Horace’s: te . . . Glycerae decoram transfer in aedem (Od. 1. 30. 3-4), with aedem we should supply futuram, as when Venus arrives, the house will have become a temple. In poetry sedes is often used like aedes for a single dwelling, as in: ipsius at sedes . . . fulgenti splendent auro (Catull. 64. 43-4); such a use of domus is rare, but there seems an example of it in: ubi . . . perventum ad limina sedis antiquasque domos (Aen. 2. 634-5).

In Greek names of places are commonly singular, e.g. Ὀρχομενός, Δυφνοῦς (= δαφνόεις ‘abounding in laurel’), Σικυών ‘the cucumber bed’, Μαραθών ‘abounding in fennel’; many of them being clearly epithets of τόπος. But some, like Ἀθῆναι and Θῆβαι, are plurals. For both of these we have the names of the older town or citadel, the Cecropia and the Cadmeia, to which later additions were made, constituting the new cities of Athens and Thebes; so that we have reason for regarding them, too, as plurals of parts. In Italy most towns were primitive hill forts; and the names Arpinum, Nomentum, Privernum, Tusculum, Ferentinum, seem epithets of oppidum or castellum. Some like Velitrae, Fidenae, Faesulae, may be plurals of parts like Mycenae, or Thebae. But just as Mycenae becomes Mycena in Virgil, so we find Fidena for Fidenae (Aen. 6. 773), following the tendency to make names of cities epithets of urbs or civitas, like Nola, Aricia, Bola, Cora, Norba. The Greek use of Μυκήνη for Μυκήναι probably represents the same tendency. Masculine plurals like Falerii, Corioli, Gabii, Arpi seem primarily names of peoples, which is clearly the case with Veii; for we read in Livy: 73 Romani Veiique in armis erant (5. 1. 1). Veientes like Falisci is evidently a later formation. And in Greek we read of Delphi in Herodotus: Κροῖσος . . . τέμφας αὖτις ἐς Πυθώ, Δελφοὺς δωρέεται (1. 54).

Turning to common names of places, we have λιμήν or λιμένες for the harbour in Greek, ὄχθη or ὄχθαι for the river bank, δυσμή the sunset but δυσμαί the west, ἀκτή the promontory but ἀκταί the coast-line. In Latin castra is the camp, angustiae and fauces the mountain pass, hortus the garden but horti the park, finis the boundary but fines the territory. Rostra is plainly the tribunal adorned by the rostra brought from Antium. Πύλαι, θύραι, and fores are old duals, but perhaps they are rather to be regarded as instruments, with which we proceed to deal in our next paragraph.

The use of τόξα rather than τόξον for the bow (Il. 1. 45) is likely to puzzle the student. It used to be explained as a plural of excellence; but then why should it be τόξα, a plural, while still hanging on the shoulders, but βιός, a singular (v. 49), when in Apollo’s hands, wreaking destruction on the Greeks? In the Odyssey the bow of Ulysses is τόξα (21. 349), but τόξον (v. 352) without any apparent change of meaning. Arcus in Latin is singular not plural for a single bow. And we have for pincers forceps and forcipes, forfex and forfices for scissors, volsella and volsellae for tweezers. In Greek λαβίς and πυράγρα are singulars. So tabellae the memorandum book is at times tabella, and codicilli is a short note, but codicillus is at times used for the note appended to the will. Bracca as well as braccae is used for the breeches, as is brax as well as braces. Clitellae is usual for the saddle, but the singular forms cletella, cretella, cratella occur in the glosses. Molae is the mill, and mola the upper millstone, but there are examples of mola the mill, like the Greek μύλη, which is either mill or nether millstone, the upper being ὄνος and the union at times αἱ μύλαι. You will see that in all these cases the instrument consists of two parts, giving naturally a dual for the union, which may pass to the singular as well as to the plural. Fides in the singular is ‘the string’, in the plural ‘the lyre’; but the singular is often used with the latter meaning. Virgil has currus and currūs for the chariot, as Homer has ἅρμα and ἅρματα. Scala is a step, and scalae a collection of steps, a ladder; but I read of Jacob in the Vulgate: vidit in somnis scalam stantem super terram (Gen. 28. 12). And it may be that the habit of instruments of twofold structure has passed by analogy to those of more complex nature.


Both Greek and Latin use plurals to designate a single day. Kalendae is used for the first day of the month, Idus for the middle, and Nonae for the fifth or seventh intervening. Delbrück seems right in his belief that this plural, too, is a plural of parts. For ‘night and day’ Homer uses: νύκτας τε καὶ ἦμαρ (Il. 5. 490), and the Greek for midnight is μέσαι νύκτες. The division of the night into three watches must have been very old; it gives rise to the phrase: trir aktun for ‘night’ in the Vedic Hymns, and in Homer we read:

παρῴχωκεν δὲ πλέων νὺξ

τῶν δύο μοιράων, τριτάτη δ’ ἔτι μοῖρα λέλειπται (Il. 10. 252-3).

From this use of the plural for the night arises by analogy a habit of indicating feasts or special holidays by the plural, though they last but a day. Weihnachten ‘holy nights’ is the German for Christmas, and feriae the Latin for a holiday, of which word nundinae the market day, nuptiae the wedding, and exsequiae the funeral seem to be epithets. By analogy we get epulae the banquet, funera the funeral, tenebrae darkness, and somnia sleep.

We have already spoken of the use of the plural for single parts of the body, e.g. colla for the neck, and fauces for the throat. Many of the human organs occur in pairs, e.g. βλέφαρα, ὦτα, ῥῖνες, oculi, genae, tempora; and hence arises a tendency to use the plural for parts, too, that are not paired, e.g. γενειάδες the chin, σπλάγχνα, viscera, exta. In return we note a tendency to use the singular for organs that occur in pairs, as in: dexter adi pede sacra secundo (Aen. 8. 302). Fauces the jaws is usually plural, but the abl. sing. fauce is common, and the nom. faux is attested. We have here a further example of a dual finding expression by the singular as well as by the plural. In: linguis micat ore trisulcis (Aen. 2. 475) the plural seems used to give variety to the phrase, and at the same time to indicate the triple division of the lingua, following the analogy of the plural of parts.

Very common both in Greek and Latin is the Plural of Modesty, e.g. the use by an orator of nos for ego in speaking of himself and his acts. So often Cicero, as in: Moloni . . . dedimus operam (Brut. 307. 89), and he often joins nos with a following singular, as in: dissuasimus nos; sed nihil de me (Lael. 96. 25). In the primary use of this figure the speaker sinks his individual personality, and identifies himself with the people, class, or craft to which he belongs; whence the name. Of this we have a fine example in the speech of the physician Eryximachus: ἄρξομαι δὲ ἀπὸ τῆς ἰατρικῆς λέγων, 75 ἵνα καὶ πρεσβεύωμεν τὴν τέχνην (Plat. Symp. 186 B). Very usual and appropriate to orators is the speaker’s identification of himself with his audience, as seen in the familiar phrases: ὡς ἀκούομεν, or: ὥσπερ πυνθανόμεθα. The poets, too, pass without an effort from the first plural to the first singular, as in:

ἥλιον μαρτυρόμεσθα δρῶσ’ ἃ δρᾶν οὐ βούλομαι (Eurip. H. F. 858),

or in:

deus nobis haec otia fecit.

Namque erit ille mihi semper deus. (Buc. 1. 6-7.)

The Plural of Majesty seems to arise from the principle of collegiality in office so usual in Greece and Rome. We read: τά τε Συρακοσίων ἔφη ὅμως ἔτι ἥσσω τῶν σφετέρων εἶναι (Thuc. 7. 48. 5), where σφετέρων is oblique for ἡμετέρων, the speaker who uses the plural being one of the associated generals. In the days of the Augusti and Caesares there developed from this a form of court speech in which phrases like nostra maiestas or nostra serenitas appeared and were with the Roman Empire perpetuated to our own times.

Plurals of Modesty and Majesty are usually of the first person, or of the third in oblique narration. But when the speaker uses nos for himself, it is only natural that he should use vos in addressing a representative of the opposite party. In the beginning of his tenth epistle, Horace uses the singular for his friend Fuscus, but the plural for himself in:

Urbis amatorem Fuscum salvere iubemus

Ruris amatores;

but when in v. 9 he addresses his friend as vos in:

Quae vos ad caelum effertis rumore secundo,

we feel that he is performing what is due to courtesy in putting Fuscus, the leader of the opposing party, on a level with himself; while in:

Quo iste voster expolitior dens est (Catull. 39. 20),

voster seems to mean ‘a man of your nation’; but in:

Furi, villula vostra non ad Austri

Flatus opposita est, neque ad Favoni (Catull. 26. 1-2),

or in:

tenet ille immania saxa,

Vestras, Eure, domos (Aen. 1. 140-1),

the plural pronoun is used virtually as it is with us. The use is not common before the fourth century, but is evidently fully developed then.

Just as we have vos for tu, and nos for ego, primarily for the class 76 of men to which tu or ego belongs, so we find proper names pluralized, especially the names of persons of some marked excellence or defect, to denote men of like excellence or defect. This is very clear in: τῇδε γὰρ τῇ ἡμέρᾳ μυρίους ὄψονται ἀνθ’ ἑνὸς Κλεάρχους (Anab. 3. 2. 31), and in: οὐκ ἔφη νοῦν ἔχειν αὐτοὺς εἰ μὴ πολλοὺς ἐν τῷ παιδὶ τούτῳ Μαρίους ἐνορῶσι (Plut. Caes. 1). So, too, in:

Extulit haec Decios Marios magnosque Camillos (Geo. 2. 169),

and in:

Nenia . . . et maribus Curiis et decantata Camillis (Hor. Ep. 1. 1. 63-4),


caedunt Lepidos, caeduntque Metellos

Corvinosque simul (Lucan 7. 584-5),

and: interfectos Romae Varrones, Egnatios, Iulos (Ann. 1. 10. 3). And we have the transference clearly set forth in:

Lys. Unus tibi hic dum propitius sit Iupiter,

Tu istos minulos cave deos flocci feceris.

  Ol. Nugae sunt istae magnae; quasi tu nescias

Repente ut emoriantur humani Ioves (Pl. Cas. 331-3),

and inverted in:

Non mihi isti placent Parmenones, Syri,

Qui duas aut tres minas auferunt eris.

Nequius nil est quam egens consili servos. (Bacch. 649-51)

We have a like transition from an individual singular to a general plural for common nouns in:

Cyl. An opsono amplius

Tibi et parasito et mulieri? Men. Quas mulieres,

Quos tu parasites loquere? (Pl. Men. 320-2),

and in:

Lyco. Lusco liberto tuo:

Is Summanum se vocari dixit: ei reddidi,

Qui has tabellas obsignatas attulit. Ther. Quas tu mihi tabellas,

Quos tu mihi luscos libertos, quos Summanos somnias? (Curc. 543-5),

and in:

Aesch. Verum hoc mihi moraest,

Tibicina et hymenaeum qui cantent. . . . Dem. Missa haec face,

Hymenaeum turbas lampades tibicinas. (Ter. Ad. 904-7.)

So in Virgil: clipeum efferri iussit, Didymaonis artes (Aen. 5. 359), ‘one of the masterpieces of Didymaon’. With this compare:

Mucius, imposuit qui sua membra focis (Mart. 10. 25. 2),


and: exsulibusne datur ducenda Lavinia Teucris (Aen. 7. 359). So in Lucan’s verses:

Hoc animi nox illa dedit quae prima cubili

Miscuit incestam ducibus Ptolemaida nostris (10. 68-9),

where it is plain from nox illa that it is Caesar, and not Caesar and Antony, to whom the poet refers in ducibus. So Horace in:

quod male barbaras

Regum est ulta libidines (Od. 4. 12. 8),

uses regum for Terei. And in Claudian’s verses:

Pauper erat Curius, reges cum vinceret armis (In IV Cons. Hon. 413),


contentus honesto

Fabricius parvo spernebat munera regum (In Ruf. 1. 200-1),

reges is plainly for Pyrrhus. We read in Virgil:

superos Arruns sic voce precatur:

Summe deum, sancti custos Soractis Apollo (Aen. 11. 784-5),

where superos is explained as Apollo, and in:

Perque deos oro, quos hosti nuper ademi (Ov. Met. 13. 376),

deos is clearly for Minervam (cf. v. 380).

Servius tells us (ad Aen. 1. 139): bonum antiqui dicebant manum; and Festus that cerus manus in the Saliaric hymn is for creator bonus. Manes seems the old plural of manus, though it is treated as a plurale tantum as a rule; being used as a general plural for the spirit of the deceased, as in:

Nec patris Anchisae cinerem manesve revelli (Aen. 4. 427),

or for the spirits of the dead generally as in:

Nocturnosque movet manes. (Aen. 4. 490.)

It is used for inferi ‘the lower world’ in:

Haec manes veniet mihi fama sub imos (Aen. 4. 387),

and for: supplicia quae sunt apud manes (= inferos) according to Servius in: quisque suos patimur manes (Aen. 6. 743).

We find in Greek ὠδῖνες, and in Latin minae, nugae, inimicitiae, pluralia tantum as involving the idea of indefinite repetition. With these Delbrück joins preces, found in the sing. dat. preci, acc. precem, abl. prece, and we may add vices, found in the sing. gen. vicis, acc. vicem, abl. vice. It seems to me that here, too, is the natural place for such plurals as Alpes and Pyrenaei.

Delbrück notices that while ὁ ἥλιος is the only object of its kind 78 known to the ancients, it formed the plural οἱ ἥλιοι for ‘sunny days’ or ‘rays of sunshine’; and the Latin sol was parallel in its use of soles. When we turn to:

Soles occidere et redire possunt (Catull. 5. 4),

we see the reason at once in the naïve conceptions of primitive astronomy; even Lucretius favoured the view that a new sun was formed every morning. In Greek ὁ μήν, primarily the moon, has come to mean the month, and there is formed from it a secondary ἡ μήνη for the moon, not often used; for the usual term for moon is ἡ σελήνη, old *σελασνα, ‘the shining one’. In Latin *men has been lengthened to mensis, just as *aus (= οὖς), still plain in ausculto, has been lengthened to auris, old *ausis; and from it is formed mena the moon goddess, though luna, old *loucsna, ‘the shining one’, is the usual term. With this plural of repetition we might associate nomina in: tua sectus orbis nomina ducet (Hor. Od. 3. 27. 75-6) and: vitreo daturus nomina ponto (4. 2. 3-4), where we shall have no longer one, but two names through the repetition consequent on the transfer.


THE DUAL PRONOUNS σφώ, σφῶϊ, and σφωέ

Of the dual inflexions of the Sanskrit verb the oldest are evidently -tam and -tām, the endings of the second and third persons of the aorist dual. We cannot derive or explain them, and, what is more significant, they seem cognate with the corresponding -τον and -την, the endings of the Greek aorist dual. The Greek present active has the endings -τον and -τον for the second and third, and the correspondence of -τον and -την, the endings of the old timeless aorist, with the Sanskrit, makes it probable that from them proceed the later present pair -τον and -τον. But if these are merely the aorist endings transferred to the present, how have we -τον in the third person for the present? The cause that led to the conversion of -την to -τον in the third present, seems also to have led to the occasional use of -τον for -την in the third aorist, and of -την for -τον in the second. I think the cause of this confusion was the use of the dual pronoun of the third person for the second person as well as the third. Such a use, as we have seen, is common in German, Italian, and Spanish; and so need hardly seem a monstrous hypothesis for old Greek.

We find for the pronoun of the first dual in Greek nom. and acc. νώ, νῶϊ, and νῶε in Antimachus and Corinna, and νῶϊν for the gen. dat.; for the second person σφώ and σφῶι for the nom. acc. and σφῶιν for the gen. dat.; for the third person σφωέ for the acc. (for the nom. also in later Greek) and σφωίν for the gen. dat. Here we seem to have in reality two pronouns, and not three: for σφώ, σφῶι, σφωέ, and σφῶιν, if we disregard accent, are exactly parallel to νώ, νῶϊ, νῶε, νῶιν; and we see from τίς that change of accent accompanies a later development in meaning, such as would result from the use of the third personal pronoun σφώ for the second person as well. Most philologists of to-day derive σφώ ‘you two’ from the same root as we have in σύ, while they connect σφωέ with σφίν and σφέ. Hirt divides σ-φω, and would form it on the analogy of ἄμφω. But both pronouns have the same possessive σφωίτερος, and both have the same gen. dat. σφῶιν, if we disregard accent, and the ancients believed both to be cognates of σφεῖς and σφίσι; some of them were clearly at a loss to distinguish σφῶϊ from σφωέ.


Apollonius Dyscolus cites thus the verse:

τίς τ’ ἄρ σφῶι θεῶν ἔριδι ξυνέηκε μάχεσθαι; (Il. 1. 8),

as the reading of Seleucus and many others, and supported by manuscript authority. They justified this reading on the lines we have been following: if σφῶϊν means ‘of you two’ in Il. 8. 416, and σφωίν ‘of them two’ in Il. 8. 402, and σφῶϊ is the acc. ‘you two’ in Il. 1. 336, then by analogy σφωί must be ‘them two’ in Il. 1. 8. In all this they were analogists, not anomalists; and were beating Aristarchus with his own weapons. Apollonius thus amplifies their argument: All singular pronouns have the same ending for the accusative, as ἐμέ, σέ, , and so with all plurals, ἡμᾶς, ὑμας, σφᾶς, and the first and second dual too have the same ending, e.g. νῶϊ, σφῶϊ; therefore we expect σφωί as the third dual. In favour of σφωέ, which he thinks the right form, he argues that the difference in person is never expressed merely by a variation in accent, forgetting for the moment σφῶϊν and σφωίν. Recalling them to memory, he adds that the forms of the second dual are really wanting, and that for them accented forms of the third are used; for that all pronominal forms beginning with σφ are originally of the third person is clear from σφέας, σφίσι, σφέτερος, σφός. Such forms, when transferred to the second person, naturally take an accent; for the second persons have deictic power, while the forms of the third lack deixis; hence the use of the enclitics σφέ and μίν for the third singular as well. But the use of the form σφωέ for the accusative of the third is justified by the analogy of ἐμέ, σφέ, and (Gram. Gr. II. 22. 67-9). Apollonius, then, believes that σφῶϊ, σφῶϊν are really pronouns of the third person that have taken an accent when they were transferred to the second.

His son Herodian, in a note to Il. 1. 574, tells us that σφώ is not a contraction from σφῶι: that is clear from its accent. Its ending is the that we have in ἵππω; νώ and σφώ are old duals meaning ‘we two’ and ‘they two’, to which the suffix (= ϝι), also meaning ‘two’, is appended; so that in νῶϊ we have ‘we two two’. Σφωέ is found only in the accusative in Homer, and is formed from σφώ, after the analogy of μέ, σέ, ἄμμε, ὕμμε, as Apollonius conjectured. We find νῶε, also an accusative, used too as a nominative in later writers, since in all other duals the acc. and the nom. are alike. In examining the meanings of σφώ, σφῶϊ, and σφωέ, we must begin with σφώ.

We find σφώ used thrice in the Iliad as a nom., and once as an acc. In: εἰ δὴ σφὼ . . . ἐριδαίνετον (Il. 1. 574) Hephaestus is addressing Hera, but speaking of Zeus and her. In: σφὼ δὲ μαλ’ ἠθέλετον (Il. 11. 81 782) Nestor is addressing Patroclus, but is speaking of Achilles and Patroclus. In addressing one person and speaking to him of himself and another we use the honorific ‘you’ even though it is evident that he is inferior to the other. But this use of the honorific ‘you’ is a comparatively late development in Latin and Greek. Neither in Attic nor in classical Latin has it become the usual form of address for one person, as it has in French and English. It is not unreasonable to think that, in the older form of expression for two persons thus spoken of, the Greeks were guided by the importance of the person addressed, and not by his presence; that Hephaestus, bearing in mind the supremacy of Zeus, followed the older fashion, addressing Hera and Zeus, when Zeus was away, as ‘they two’, and not as ‘you two’. True, in Il. 11. 781 we read ὕμμ’ for Achilles and Patroclus, expressed by σφώ in v. 782. Here the later and honorific ‘you’ is used side by side with what I think the older use, which keeps in view the relative importance of Achilles and Patroclus. In: Αἴαντε, σφὼ μέν τε σαώσετε (Il. 13. 47) we have the dual pronoun joined with a plural verb, the later use for the dual. That it was the lesser Ajax that Poseidon addressed here, we may conclude from: τοῖϊν δ’ ἔγνω πρόσθεν Ὀϊλῆος ταχὺς Αἴας (v. 66). Finally in: Ζεὺς σφὼ εις Ἴδην κέλετ’ ἐλθέμεν (Il. 15. 146) we have both Apollo and Iris addressed by Hera, and the two plurals ἔλθητε and ἴδησθε (v. 147) referring back to σφώ as their subject; so that σφώ here is treated as identical with ὑμᾶς. But in three of the four uses of σφώ, of the two persons, the one inferior in rank or importance is the one addressed.

Homer uses σφῶι five times in the nominative. In: ἔμβητον καὶ σφῶϊ (Il. 23. 403) it is addressed to the horses of Antilochus, and in: φράζεσθον δὴ σφῶϊ, Ποσείδαον καὶ Ἀθήνη (Il. 20. 115) both persons addressed are present. The remaining three are addressed to two persons, of whom only one is present: in σφῶϊ . . . ἕπετον κρέα (Il. 11. 776), of Achilles and Patroclus, Patroclus is addressed; and in: σφῶϊ δ’ ἀποστρέψαντε πόδας (Od. 22. 173), of Eumaeus and Melanthius, Melanthius is addressed; but in:

Αἶαν, σφῶϊ μὲν αὖθι, σὺ καὶ κρατερὸς Λυκομήδης,

ἑσταότες Δαναοὺς ὀτρύνετον (Il. 12. 366-7),

it is clearly the more important of the two that is addressed.

Homer uses σφῶϊ six times in the accusative. Of these, one is of the horses of Achilles (Il. 17. 443). Two are of a pair both present, of each of which one is the inferior, in Il. 4. 286 of the Ajaces, in Il. 1. 336 of Talthybius and Eurybates. One (Il. 7. 280) is used by the two 82 chief heralds of the Greeks and Trojans, Talthybius and Idaeus, of the champions they represent, Ajax and Hector. Of the remaining two, one (Il. 5. 287) is of Aeneas and Pandarus, of whom Pandarus is present, the other (Il. 10. 552) is of Diomede and Ulysses, and Ulysses is present. The accusative is of less importance here than the nominative, as the idea of address is less emphasized. Many of the ancients in Il. 7. 280 and 10. 552 read, not σφῶϊ, but σφῶε, being perhaps determined by the analogy of νῶϊ and νῶε.

Σφῶϊν is twice used as a genitive, once (Il. 1. 257) of Agamemnon and Achilles, and once (Od. 16. 171) of Ulysses and Telemachus, of whom Ulysses is present. Of the eleven uses of σφῶϊν as a dative, three are of horses (Il. 17. 451; 23. 408 and 411). Of the remaining eight, in four both persons addressed are present (Il. 13. 55; 16. 556; Od. 21. 209 and 212); in four it is the inferior that is present (Il. 4. 341; 8. 413 and 416; Od. 23. 52). In Od. 4. 62 we have the later form σφῷν used of Peisistratus and Telemachus, of whom both are addressed. We see, then, that while of the oblique cases a large proportion are used of a pair of whom the inferior is addressed, in the nominatives this is the case of the great majority.

When we turn to uses of -τον and -την as verb inflexions, according to Monro we have three examples in Homer of the use of -τον in the third person (ἐτεύχετον, Il. 13. 346; διώκετον, Il. 10. 364; λαφύσσετον, Il. 18. 583), where the metre forbids -την. Zenodotus read forms in -την for the second dual in καμέτην (Il. 8. 448) λαβέτην (Il. 10. 545), ἠθελέτην (Il. 11. 782), where Aristarchus and our editors have preferred forms in -τον. With ἠθελέτην, σφώ was expressed as its subject. The dual forms of the middle in -σθον and -σθην are formed on the analogy of the active forms; so it is not astonishing to find three in -σθον as variants for the third dual: ἀφίκεσθον (Oxford Text ἐφίκοντο) in Il. 13. 613; θωρήσσεσθον (O. T. θωρήσσοντο) in Il. 16. 218; πέτεσθον (O. T. πετέσθην) in Il. 23. 506. The tendency in Attic to read -την and -σθην for both persons of the aorist dual seems an effort to distinguish them from the present, which has -τον and -σθον for both persons. Sanskrit offers no confusion parallel to that we have noticed in Greek between -τον and -την; tam is always the inflexion of the second, and tām of the third. Nor are there pronouns in Skt. liable to a like confusion with those in Greek; there a dual pronoun yuvām ‘you two’ has been evolved out of the plural yuyam ‘you’, on analogy of āvām ‘we two’.

Σφώ is probably derived from σφίν (= σϝ-φιν) the instrumental of 83 οὗ (= σϝ-ου) on the analogy of νώ. The pronoun οὗ was primarily used for both singular and plural like the Latin sui, its cognate. From σφίν after the analogy of μέ and σέ was formed σφέ also used for both singular and plural. We find this σφέ for the dual σφωέ in Il. 11. 111 and 115; Od. 8. 271; 21. 192 and 206; Hes. H. S. 62. For both σφώ and σφωέ the possessive is σφωίτερος, which is ‘of you two’ in Il. 1. 216. Just as σφε is used for σφώ or σφωέ, so σφίτερος is used for σφέτερος by Apollonius Rhodius in 1. 643 and with reflexive force in 3. 600. It is used for σός (Ap. Rh. 3. 395), just as σφέτερος is used for the singular as well as the plural.

We have what looks at first glance like an interesting parallel to this use of σφῶϊ for ‘you two’ in the Aeneid. We read:

Vivite felices, quibus est fortuna peracta

Iam sua: nos alia ex aliis in fata vocamur. (Aen. 3. 493-4.)

Servius’s explanation is satisfactory; sua: id est dura, propria Troianorum, ut (Aen. 6. 62) hac Troiana tenus fuerit fortuna secuta. But it is very probable that Virgil, in thus shaping his verse, had in mind the Homeric use of σφώ for the second person.



We turn to the Schema Pindaricum or Boeoticum. Following the analogy of: τὰ ζῷα τρέχει, thinks Riemann, certain poets, and especially Pindar, use the singular of the verb with the names of things in the masculine or feminine plural. Jelf and Riemann seem right in citing as an example of this construction the well-known couplet of Hipponax:

δυ’ ἡμέραι γυναικός ἐστιν ἥδισται

ὅταν γαμῇ τις κἀκφέρῃ τεθνηκυῖαν (Fr. 29, B.).

Gaisford conjectured εἰσιν for ἐστιν, and Bergk adopted this in his edition, because he found it in one of his manuscripts, as he was almost sure to do. His business was to account for the usual reading ἐστιν, rather than substitute for it a reading which some scribe would inevitably introduce in his copy.

Eustathius tells us that Homer invented this schema, and cites to prove his point:

καμάτῳ δὲ καὶ ἵδρῷ νωλεμὲς αἰεὶ

γούνατά τε κνῆμαί τε πόδες θ’ ὑπένερθεν ἑκάστου

χεῖρές τ’ ὀφθαλμοί τε παλάσσετο μαρναμένοιϊν

ἀμφ’ ἀγαθὸν θεράποντα ποδώκεος Αἴακίδαο (Il. 17. 385-8).

We notice that, while in our first example the subject is a single pair, in this it consists of a large number of pairs; both of these might take the dual in Greek. But we noticed that, when a woman used her two hands to wipe her two cheeks, the plural, not the dual, was the number in use, not merely for the verb, but for the pairs connected with it. And we also noticed that in Latin the dual found expression, not always by the plural, but often by the singular; and that there were traces of this usage in Greek as well. To follow the old rule, the number of παλάσσετο may be determined by the nearest subject ὀφθαλμοί, and its synonym ὄσσε takes the singular thrice in Homer. This view of the construction is strengthened when we turn to:

πναιῇ δ’ Εὐμήλοιο μετάφρενον εὐρέε τ’ ὤμω

θέρμετ’ (Il. 23. 380-1),


and we feel it natural to compare this with:

οὔτε τοι ὀξύτατον κεφαλῆς ἐκδέρκεται ὄσσε (Il. 23. 477),

and still more when we turn to: εἰ ἔστιν τούτω διττὼ τὼ βίω (Plat. Gorg. 500 D), or to:

ἡμῖν γὰρ οὐκ ἔστ’ οὔτε κάρυ’ ἐκ φορμίδος

δούλω διαῤῥιπτοῦντε τοῖς θεωμένοις (Ar. Vesp. 58-9),

or to:

καὶ Μᾶγος Ἄραβος, Ἀρτάβης τε Βάκτριος,

σκληρᾶς μέτοικος γῆς ἐκεῖ κατέφθιτο (Aesch. Pers. 318-19),

or even to: Ὅμηρος μέν νυν καὶ τὰ Κύπρια ἔπεα χαιρέτω (Hdt. 2. 118).

Wilpert (Das Schema Pindaricum, Oppeln, 1900) quotes as an example of this schema:

ἦ τοι ἐμοὶ χλαῖναι καὶ ῥήγεα σιγαλόεντα

ἤχθεθ’ (Od. 19. 337-8),

where we have plainly a duality of collectives; but it is easier to regard the ἤχθετο as determined by ῥήγεα, the nearest subject. Still this is not so plausible in: ἵνα δοκοῦντι δικαίῳ εἶναι γίγνηται ἀπὸ τῆς δόξης ἀρχαί τε καὶ γάμοι (Plat. Rep. 363 A), or in: ἔστι μέν που καὶ ἐν ταῖς ἄλλαις πόλεσιν ἄρχοντές τε καὶ δῆμος (ib. 463 A), or in: σκέλη μὲν οὖν χεῖρές τε ταύτῃ καὶ διὰ ταῦτα προσέφν πᾶσιν (Plat. Tim. 45 A), or in ἀφ’ ὧν ἐμοὶ ξενίαι καὶ φιλότητες πρὸς πολλοὺς καὶ βυσιλέας καὶ πόλεις καὶ ἄλλους ἰδίᾳ ξένους γεγένηται (Andoc. 1. 145). In the last example the two subjects are so closely allied in meaning, that we might assume that the singular verb is determined by what Gildersleeve calls a Unity of Subject. But of this more presently. It seems probable that in the examples cited we have a duality of idea in the subject; and that this follows the same rule as do the duals cited above in taking the verb in the singular. So from: σάρκες καὶ νεῦρα ἐξ αἵματος γίγνεται (Plat. Tim. 82 C) we may assume that in: οὐδ’ ἔπι χεῖρες οὐδὲ πόδες (Hdt. 6. 86 γ’ 2) ἔπι is for ἔπεστι.

This tendency to take the verb in the singular was probably strengthened by the construction with the singular of another class of plurals expressing a unity of idea, a class that seems to have fused with the duals cited above in creating this schema. We have such a plural in:

ξανθαὶ δὲ κόμαι κατενήνοθεν ὤμους (Hym. in Cer. 279),

and in:

ἐκ δέ οἵ ὤμων

ἦν ἑκατὸν κεφαλαὶ ὄφιος (Hes. Theo. 824-5),

where the snakes’ heads are felt as constituting a mane. It is felt, we 86 are told, in: τῆς δ’ ἦν τρεῖς κεφαλαί (Hes. Theo. 321) that after ἦν a subject is expected designating the head of the monster, and from this we have a singular verb, but it finds expression in τρεῖς κεφαλαί. With a feeling that this was the true explanation of the schema grammarians everywhere quoted:

ἐν δ’ ἀγαθοῖσι κεῖται

πατρώϊαι κεδναὶ πολίων κυβερνάσιες (Pind. Pyth. 10. 71-2),

and taught: The speaker begins with a single subject in mind; and this determines the number of the verb, but afterwards finds expression in a plural form. But we read in Pindar:

μελιγάρυες ὕμνοι

ὑστέρων ἀρχαὶ λόγων

τέλλεται     (Ol. 11. 4-6).

Wilpert quotes: λάχναι νιν μέλαν γένειον ἔρεφεν (Ol. 1. 68) (where all my editions read ἔρεφον). And, to turn to prose, we have in Plato: πάχναι καὶ χάλαζαι καὶ ἐρυσῖβαι ἐκ πλεονεξίας καὶ ἀκοσμίας περὶ ἄλληλα τῶν τοιούτων γίγνεται ἐρωτικῶν (Symp. 188 B). Even when the verb precedes the subject, as in:

ἦν δ’ ἀμφίπλεκτοι κλίμακες

ἦν δὲ μετώπων ὀλόεντα

πλήγματα καὶ στόνος ἀμφοῖν (Soph. Trach. 520-2),


ἐνῆν δ’ ὑφανταὶ γράμμασιν τοιαίδ’ ὑφαί (Eur. Ion 1146),

or: διήγεται σάρκες (Pind. Fr. 183), where σάρκες is the Latin caro, or in:

μέγα τοι δύναται νεβρῶν παμποίκιλοι στολίδες (Eur. Hel. 1358),

the interval between the verb and the following subject is so short as to make this view improbable. The plural subject σάρκες gives us a singular idea, and Wilpert thinks that οτολίδες designates a single garment.

But the theory deduced from the position of κυβερνάσιες does fit the latest form of this schema, and the form usually found in prose. We read in Herodotus: ἔστι δὲ μεταξὺ τῆς τε παλαιῆς πόλιος, ἣ τότε ἐπολιορκέετο, καὶ τοῦ νηοῦ ἑπτὰ στάδιοι (1. 26). So in Plato: ἔστι γὰρ ἔμοιγε καὶ βωμοί (Euthyd. 302 C) and in: ἐγένετο δὲ μετὰ τοὺς λόγους τούτους ὡσεὶ ἡμέραι ὀκτώ (Luke 9. 28). But in: ἔστι δὲ ἑπτὰ στάδιοι ἐξ Ἀβύδου ἐς τὴν ἀπαντίον (Hdt. 7. 34) there is no interval between the verb and its subject, and the unity of idea in ἑπτὰ στάδιοι may well determine the number of ἔστι. The feeling of interval is very strong in: προσξυνεβάλετο οὐκ ἐλάχιστον τῆς ὁρμῆς αἵ Πελοποννησίων νῆες ἐς Ἰωνίαν ἐκείνοις βοηθοὶ τολμήσασαι παρακινδυνεῦσαι (Thuc. 3. 36. 2), 87 where Arnold thinks that the subject the speaker has in mind, when he begins, is τὸ τὰς ναῦς τολμῆσαι. In this prose form Kühner seems justified in comparing the schema with the French: il est des hommes, or: il est cent choses, where the German seems more logical with its: es sind viele Dinge. But to account for the origin of the schema from this last stage in its history, as is so commonly done, seems absurd. Even here the original duality of the subject is often evident, as in: ἦν δὲ τοῦ δανείσματος τετταράκοντα μὲν καὶ πέντε μναῖ ἐμαί, τάλαντον δ’ Εὐέργου (Dem. 37. 4)

Our reference to il est leads us naturally to the use of ἔστιν οἵ for ἔνιοι, as we see it in: ἐνταῦθα δὴ οἱ βάλλοντες ταῖς βολαῖς ἔστιν οἵ καὶ ἐτύγχαναν καὶ θωράκων καὶ γέρρων, οἱ δὲ καὶ μηροῦ καὶ κνημῖδος (Xen. Cyr. 2. 3. 18). We find ἔστιν ἅ for ἔνια in: καὶ (Κλεόπομπος) ἀποβάσεις ποιησάμενος τῆς τε παραθαλασσίου ἔστιν ἃ (χωρία) ἐδῄωσε καὶ Θρόνιον εἷλεν (Thuc. 2. 26. 2). Ἔστιν ἅ for ἔνια is quite in harmony with Greek syntax, suitable in meaning, and of a like number of syllables. The natural result of this correspondence would be that in ἔστιν ἅ the primary meaning of the first two syllables would be obscured to some extent, just as in ἔνια the primary force of ἔνι- was partially lost, and they passed from ‘some other’ (cf. Skt. anye) to ‘some’. In ἔστιν apparently it was the idea of number that was obscured, and this seems to have taken place gradually; and by a series of stages of which ἔστιν οἵ was the last.

The idiom belongs to prose, and appears first in Herodotus. But neither Herodotus nor Thucydides gives us ἔστιν οἵ, which we find first in Xenophon. Herodotus has: εἰσὶ δὲ οἳ λέγουσι τοὺς ἀπ’ Αἰγύπτου νικῆσαι Πολυκράτεα (3. 45); and Thucydides: εἰσὶ δὲ αἳ καὶ οἰκήτορας μετέβαλον ἁλισκόμεναι (1. 23. 2) et saepius. But in cases where the masculine and feminine of the pronoun do not differ from the neuter in form, we find the transition to them already in Herodotus and Thucydides; e.g. in: προδοῦναι τὰ ῥέεθρα τῶν ποταμῶν ἔστι ὧν (Herod. 7. 187); and: αὐχμοί τε ἔστι παρ’ οἷς μεγάλοι (Thuc. 1. 23. 3); and so in Plato: ἔστιν ὅτε και οἷς βέλτιον τεθνάναι ἢ ζῆν (Phaedo 62 A); and in Xenophon: ἢ ἔστιν οἷς καὶ πάνυ ἀρέσκει; (Mem. 2. 3. 6). But in Plato we find masculine forms of the pronoun differing from the neuter thus constructed, as in: ἄκων δ’ ἔστιν οὕς ἐγὼ ἐπαινῶ καὶ φιλῶ (Prot. 346 E) and in Xenophon: ἔστι δὲ ἃς ἂν καὶ πόλεις τῆς ἀναγραφῆς ὀρεγομένας (Vect. 3. 11). Last of all we come to ἔστιν οἵ in the example first cited; but even in later Greek εἰσὶν οἵ seems more usual. Propertius has imitated ἔστιν οἷς in:


Est quibus Eleae concurrit palma quadrigae,

Est quibus in celeres gloria nata pedes (3. 9. 17),

which Paley calls ‘a bold and perhaps unique Graecism’.

We see how starting from ἔστιν ἅ, parallel to ἔνια, in about fifty years we come to ἔστιν οἵ, parallel to ἔνιοι, in which probably the ἔστιν is no longer felt as the third singular of a verb, but as the first two syllables of an indefinite pronoun. This result is undoubtedly aided by the use of ἔστιν ὅτε for ἔνιοτε (Phaedo 62 A).

Parallel to this seems the development in the force of ἐν τοῖς in the phrase: ἐν τοῖς πρῶτοι. This is a prose construction also, found in Herodotus, Thucydides, and Plato, and of the later Atticists in Lucian. The oldest example I know of is in: τοῦτό μοι ἐν τοῖσι θειότατον φαίνεται γενέσθαι (Herod. 7. 137), where Rawlinson favours the translation: ‘in my opinion this was one of the cases in which the hand of heaven was most plainly manifest’. It occurs six times in Thucydides: (1) ἐν τοῖς πρῶτοι δὲ’ Αθηναῖοι τόν τε σίδηρον κατέθεντο (1. 6. 3) ‘the Athenians were among the first who, laying aside iron,’ &c.; (2) καὶ κατὰ τὸν χρόνον τοῦτον ὅν αἱ νῆες ἔπλεον, ἐν τοῖς πλεῖσται δὴ νῆες ἅμ’ αὐτοῖς ἐνεργοὶ κάλλει ἐγένοντο, παραπλήσιαι δὲ καὶ ἔτι πλείους ἀρχομένου τοῦ πολέμου (3. 17. 1) ‘at the time when the ships were at sea, the Athenians had one of the largest fleets they ever had all assembled together, effective and in good trim, though the number of their ships was as large or even larger at the beginning of the war’, where I have changed Jowett’s translation slightly to get rid of the contradiction between his ‘the largest’ and ‘even larger’; (3) οὕτως ὠμὴ ἡ στάσις προυχώρησε, καὶ ἔδοξε μᾶλλον, διότι ἐν τοῖς πρώτη ἔγενετο (3. 82. 1) ‘to such bitter extremes did the civil strife proceed; and it seemed the more so, because it was among the earliest’; (4) μέγιστόν τε καὶ ἐν τοῖς πρῶτον ἐκάκωσε τὸ στράτευμα τῶν Ἀθηναίων ἡ τοῦ Πλημμυρίου λῆψις (7. 24. 3) ‘a very severe blow and one of the greatest that befel the Athenians was the taking of Plemmyrium’; (5) ἄλλοι δὲ . . . ἐν τοῖς χαλεπώτατα διῆγον (7. 71. 3) ‘others again . . . were involved in one of the hardest of cases’; (6) καὶ Ἀρίσταρχος, ἀνὴρ ἐν τοῖς μάλιστα καὶ ἐκ πλείστον ἐναντίος τῷ δήμῳ (8. 90. 1) ‘and Aristarchus, who had been one of the foremost and most uncompromising in his opposition to the commons’.

In Plato, Matthiae, whom I follow, notes these six examples: (1) ἣν ἐγώ, ὡς ἐμοὶ δοκῶ, ἐν τοῖς βαρύτατ’ ἂν ἐνέγκαιμι (Crito 43 C) ‘one of the saddest which in my opinion I could be bringing you’; (2) καὶ οὐχ ἥκιστα Ἀθηναίων σὲ (ἐνέξεσθαι), ἀλλ’ ἐν τοῖς μάλιστα (ib. 52 A) 89 ‘that of the Athenians you were not the least, but one of the most involved’; (3) λέγοντες ὅτι ἐν τοῖς μάλιστα Ἀθηναίων ἐγὼ αὐτοῖς ὡμολογηκὼς τυγχάνω ταύτην τὴν ὁμολογίαν (ib. 52 A) ‘saying that I happened to be one of the Athenians that had been most forward in making this acknowledgement’; (4) καὶ τούτων (ἡ ψυχή) μοι δοκεῖ ἐν τοῖς μάλιστα πρὸς ἄλληλα σκοπεῖσθαι τὴν οὐσίαν (Theaet. 186 A) ‘and methinks these are among the things of which the soul most carefully considers the essence with regard to each other’; (5) παρεγεγόνει δ’ ἐν τῇ συνουσίᾳ, Σωκράτους ἐραστὴς ὢν ἐν τοῖς μάλιστα τῶν τότε (Symp. 173 B) ‘and he had been at the gathering, being one of those who at that time were most attracted to Socrates’; (6) ἀκούω Δίωνος ἐν τοῖς μάλιστα ἑταῖρον εἶναί τέ σε νῦν καὶ γεγονέυαι διὰ παντός (Ep. 358 C) ‘I hear that you both now are and have ever been one of the closest friends of Dion’.

I have translated all these examples with the view that ἐν τοῖς is restrictive, and not intensive, in its force; but I am not certain that its force, primarily restrictive as it is, is clearly felt in each one of the last six examples. You will notice that five of these give: ἐν τοῖς μάλιστα, showing that the idiom is crystallizing in a single phrase, which more and more tends to the same meaning as μάλιστα. When in later Greek we read ἐν τοῖς μάλα, it is plain that this is an equivalent for μάλα, and that the ἐν τοῖς has become otiose.

But two other views call for a brief notice. Morris (ad Thuc. 1. 6. 3) quotes: καὶ Θηραμένης ὁ τοῦ Ἅγνωνος ἐν τοῖς ξυγκαταλύουσι τὸν δῆμον πρῶτος ἦν (Thuc. 8. 68. 4) ‘Theramenes, too, Hagnon’s son, was foremost among those that overthrew the democracy’. This is not an example of our idiom; but with its help Morris interprets Thuc. 1. 6. 3, supplying thus: ἐν τοῖς τόν τε σίδηρον καταθεμένοις πρῶτοι Ἀθηναῖοι αὐτὸν κατέθεντο, giving an intensive force. But how shall I fit this explanation to my second example? True, Jelf speaks of ἐν ταῖς with this construction; but he cites no example. While Jowett thinks ἐν τοῖς intensive in the first three examples, he makes it restrictive in the fourth, and it seems the same in the fifth and sixth. Have we then here an idiom used in two opposite meanings, and defeating one of the aims of prose, that of clear expression?

The older interpreters of this idiom dealt with it in simpler fashion, and compared it with: ἐν μαλακωτάτοις τῶν μαλακωτάτων (Plat. Symp. 195 E), and with ἐν τοῖς μεγίστοις μέγιστον (Crat. 427 E), where Stallbaum would omit μεγίστοις, making it an example of our idiom. So in Thucydides ἐν τοῖς πρῶτοι would be short for ἐν τοῖς πρώτοις 90 πρῶτοι, about the intensive force of which there can be no doubt. But ἐν τοῖς πλεῖστοι seems so adverse to this solution that its supporters wish to omit the whole passage in Thuc. 3. 17 where the former occurs. Non tali auxilio.

Morris quotes ἐν τοῖς πρῶτοι as found in Thucydides 7. 19. 4 and 8. 89. 2, and ἐν τοῖς πρῶτον in 7. 27. 3. But these are the readings of Bekker, not of the best manuscripts, which give ἐν τοῖς πρώτοις. Some editors now recognize that here Thucydides gives better expression to his idea by ἐν τοῖς πρώτοις than by ἐν τοῖς πρῶτοι or πρῶτον. But while Bekker is wrong in changing the manuscript reading, no one has questioned the correctness of his Greek; he has substituted for the adverbial phrase ἐν τοῖς πρώτοις in the predicate the adjective πρῶτοι with its modifying phrase ἐν τοῖς, and this is exactly parallel to the use in τριταῖοι ἦλθον for τῆ τρίτῃ ἡμέρᾳ ἦλθον, or in Latin of vespertinus erro for vesperi erro. This is the way in which our idiom comes to pass, passing from ἐν τοῖς πρώτοις to ἐν τοῖς πρῶτοι, where after a time the ἐν τοῖς come to be regarded as the first two syllables of the adverb, like παρα- in παραχρῆμα.

We may be asked why, when Bekker found ἐν τοῖς πρώτοις twice in Thucydides, but ἐν τοῖς πρῶτοι only once, he did not change ἐν τοῖς πρῶτοι to ἐν τοῖς πρώτοις. Plainly because of the Scholium: ἐν τοῖς — ἐν τούτοις ποιητικῶς· ὑπερβιβάζεται γὰρ ὁ δέ (‘ἐν τοῖς is used poetically for ἐν τούτοις; for the δέ is transposed’), i.e. we should have ἐν τοῖς δὲ πρῶτοι, where τοῖς δε would be for τούτοις. While this Scholium assures the text, it seems to suggest Morris’s explanation; for ἐν τούτοις must be for: ἐν τοῖς . . . καταθεμένοις. Indeed this Scholium seems very unfortunate: all three things it tells us are wrong; the expression belongs to prose, not to poetry; the δέ is placed after ἐν τοῖς πρῶτοι, because that is felt as one word; and the view that ἐν τοῖς is for ἐν τούτοις comes to grief in the next two examples we meet. Plainly the force of ἐν τοῖς had been quite obscured by the time this Scholium was written.

Just as in Greek we pass from ἐν τοῖς πρώτοις to ἐν τοῖς πρῶτοι, so in Latin we pass from absente me to absente nobis. We read:

Restituis cupido atque insperanti ipsa refers te

Nobis (Catull. 107. 5-6),


Perfida nec merito nobis inimica merenti (Tib. 3. 6. 55),

in both of which nobis is the plural of modesty for mihi; so that if 91 these examples stood alone, they might be thought due to sense construction. In:

Nescio quid profecto absente nobis turbatumst domi (Ter. Eun. 649),

where absente nobis is for absente me, Donatus wishes to supply me, and some modern editors, accepting this, connect nobis with turbatumst, ‘we have had some disturbance’. But Donatus quotes two further examples that make this improbable; one from Pomponius: praesente amicis inter cenam (Fr. 47, Rib.), and one from Varro: id praesente legatis omnibus exercitu pronuntiat (De Serm. Lat. V, Fr. 82, Wilm.). Under absente Nonius gives us (p. 76, M.): nec nobis praesentest aliquis quisquam nisi servus (Pl. Amph. 400), where all manuscripts have: nec nobis praeter me aliquis quisquamst; and: adest si hic absente nobis venierit puer (Afran. Fr. 6, Rib.); under praesente from Pomponius: quidam apud forum praesente testibus mihi vendidit (Fr. 168, Rib.), and from Accius: Est res aliqua, quam praesente his prius maturare institit (Fr. 428, Rib.); also from Fenestella, Annalium Lib. II: et quaedam exta praesente suis, quaedam absente porrecissent, and from Novius: te volumus dono donare pulcro praesente omnibus (Fr. 57, Rib.). Of these eleven examples only five can be solved as sense constructions, and the remaining six seem rather examples of a transition, as from praesente amico to praesente amicis. And this is Donatus’s second solution; for he adds: aut ἀρχαϊσμός est figura absente nobis pro absentibus nobis . . . absente nobis cum dicit, pro praepositione ponit ‘absente’, ut si diceret ‘coram amicis’. Absente and praesente, then, he feels to be prepositions like coram.

So in Cato and Gellius we find fini used as the predicate of an ablative absolute, and afterwards as a post-position and preposition. We read in Gellius: hac, inquit, fini ames, tamquam forte fortuna et osurus, hac itidem tenus oderis, tamquam fortasse post amaturus (1. 3. 30), where hac fini balances hac tenus. So: qua fini (id. 1. 3. 16), and ea fini (6. 3. 29). These phrases are ablatives absolute, in which ‘that being the limit’ is taken as ‘up to that limit’. So in our ‘till now’, where ‘till’ is the Greek τέλος, the German Ziel. In Cato we read: cupa qua fini in modiolos erit (R. R. 21. 3, K.), and: postea operito terra radicibus fini (ib. 28. 2), where radicibus fini ‘the roots being the limit’ gives us a parallel to praesente amicis. In: amphoras nolito implere nimium; ansarum infimarum fini (R. R. 113. 2) ‘don’t fill your jars too full; up to the 92 bottom of the handles’, Cato constructs fini with the genitive just as Catullus does tenus in: nutricum tenus (64. 18); and so Caesar in: pectoris fini prominentes (B. G. 7. 47. 5). It appears as a preposition in a fragment of Sallust quoted by Arusianus Messius: fini inguinum ingrediuntur mare (231, L.). I may add that Arusianus seems to cite absente populo as the link between absente me and absente nobis (213, L.).

Interesting is the parallel to praesente nobis we have in: istorum nominandi copia. From: sui nominandi copiam habent, by a change of subject we have: illorum nominandi copiam habemus. Nominandi, attracted by sui, which is singular in form but plural in meaning, becomes like absente indifferent to number. We have few examples of this idiom, most of which we find in Cicero; but the two we find in archaic Latin are so well developed, that we must regard the idiom as quite as characteristic of archaic as of classical Latin. Schmalz tries to account for it by relating it to the genitive of the Greek substantival infinitive as found in: τούτων οὐχὶ νῦν ὁρῶ τὸν καιρὸν τοῦ λέγειν (Dem. 2. 4). But what is the relation of τοῦ λέγειν to nominandi? The influence of Greek syntax on that of archaic Latin is very slight.

We have these two examples of this idiom in the Latin of the second century B.C.:

Nominandi istorum tibi erit magis quam edundi copia,

Hic apud me (Pl. Capt. 852-3),


Novarum (fabularum) qui spectandi faciunt copiam

Sine vitiis (Ter. Heaut. 29-30).

Varro gives us: principium generandi animalium (R. R. 2. 1. 3) and Lucretius: poenarum . . . solvendi tempus (5. 1225). It is not found in Caesar, Sallust, Livy, Horace, or Virgil; but Draeger quotes from Cicero the six following examples: earum autem rerum . . . infitiandi rationem (Verr. 2. 4. 104. 47), ne reiciundi quidem amplius quam trium iudicum . . . faciunt potestatem (Verr. 2. 2. 77. 31), fuit exemplorum eligendi potestas (Inv. 2. 5. 2), quarum potiendi spe inflammati (Fin. 1. 60. 18), (causa) eorum, quae secundum naturam sunt, adipiscendi (Fin. 5. 19. 7), (facultas) caedis faciendae bonorum, diripiendae urbis, agrorum suis latronibus condonandi (Phil. 5. 6. 3), with one from Cornificius: isti magistri, omnium dicendi praeceptores (Her. 4. 9. 6). In Silver Latin we find only: iocandi licentia diripiendique pomorum et opsoniorum rerumque omnium missilium (Suet. Aug. 93 98. 3). Fronto has: tantus usus studiorum bonarumque artium communicandi (Ep. ad Aur. 1. 24); and Gellius: verborumque fingendi et novandi studium (4. 15. 1) and: causarumque orandi cupiens (5. 10. 5). We read in Caesar: neque sui colligendi hostibus facultatem relinquunt (B. G. 3. 6) and: venerunt . . . sui purgandi causa (ib. 4. 13), and in Livy: vestri adhortandi causa (21. 41. 1), in which examples the gerundive agrees with the form, and not with the meaning of the pronoun. So the gerundive becomes in a measure indifferent to number, and from: sui colligendi facultatem hostibus reliquit we should get: hostium colligendi facultatem reliquit eorum duci.

That this is the true explanation of the idiom is confirmed by examples where it is not number, but gender, that is in question. We read in Plautus: quia tui videndi copiast (Truc. 370), where tui is feminine, but videndi agrees with it in form rather than meaning. Hence the gerundive becomes indifferent to gender at times, and we have: lucis das tuendi copiam (Pl. Capt. 1008), eius (= uxoris) videndi cupidus (Ter. Hec. 372), where Donatus takes no notice of the irregularity, copia placandi sit modo parva tui (fem.) (Ov. Her. 20. 74), and: crescit enim adsidue spectandi cura puellae (Prop. 3. 21. 3).

Let us turn from this rare and complicated construction to a common and simple one, the agreement of the verb with two subjects connected by καί or et. The answer seems easy to the grammarian: the verb must be in the plural. When Kipling ventured to write: ‘The tumult and the shouting dies’, he was at once assailed by our pedants. But we read in Homer:

Τρώων δὲ κλαγγή τε καὶ ἄσπετος ὦρτο κυδοιμὸς

θυνόντων ἄμυδις (Il. 10. 523-4),

a passage which Kipling may have had in mind. In Greek related pairs of words commonly take a singular verb, as we have already noticed in dealing with the Schema Pindaricum. So in:

ὅθι νητὸς χρυσὸς καὶ χαλκὸς ἔκειτο

ὲσθής τ’ ἐν χηλοῖσιν ἅλις τ’ εὐῶδες ἔλαιον (Od. 2. 338-9),

or in:

πάντα τύχη καὶ μοῖρα, Περίκλεες, ἀνδρὶ δίδωσιν (Archil. Fr. 16, B.),


ἀλλ’ εὐγενὴς μὲν ὁ κτανών τε χω’ θανών (Soph. Phil. 336)


εἰ δ’ Ὀρφέως μοι γλῶσσα καὶ μέλος παρῆν (Eur. Alc. 357),

or: ἔδοξε γὰρ δὴ ἡμῖν ἡ πολιτικὴ καὶ ἡ βασιλικὴ τέχνη ἡ αὐτὴ εἶναι (Plat. 94 Euthyd. 291 C), or: οὗ δὴ καὶ ἐκφανὴς ἐγένετο ἡ τῆς πόλεως ῥώμη καὶ ἀρετή (id. Menex. 243 C), or: ἄνεμος καὶ χειμὼν διεκώλυσεν αὐτούς (Xen. Hell. 1. 6. 35). It is not that the pair of words describe the same object, nor that the objects they designate are not often opposed; it is that they constitute a correlated pair, such as will naturally be expressed by a dual; and hence, like the dual, they are at times constructed with a singular verb.

When we turn to Latin this becomes even plainer. ‘Close union often amounts to unity’, says Gildersleeve. But in Latin close union very often gives us plurality, and ideas mutually opposed are constructed with a singular verb as well as with a plural. In archaic and classical Latin as a rule with a closely related pair of words the verb is in the singular, as in: tua fama et gnatae vita in dubium veniet (Ter. Ad. 340), (fabulae) novae novum intervenit vitium et calamitas (id. Hec. 2), cum tempus necessitasque postulat (Cic. Off. 1. 81. 23), societas hominum et communitas evertatur necesse est (ib. 3. 22. 5), ut summus furor atque amentia consequatur (Rose. Am. 66. 24); but in: (fortuna), quam nemo ab inconstantia et temeritate seiunget, quae dignae certe non sunt deo (N. D. 3. 61. 24) we have the plural. This is especially the case when the members of a pair are contrasted, as in: ius et iniuria natura diiudicantur (Leg. 1. 44. 16) or: nam vita et mors iura naturae sunt (Sall. Hist. 2. 50. 5, Kr.). But we have both: religio et fides anteponatur amicitiae (Off. 3. 46. 10) and: ni virtus fidesque vostra spectata mihi forent (Sall. Cat. 20. 2). While Cicero and Caesar as a rule have the singular with ideas so closely related, Livy has the plural very often, and Sallust and Tacitus favour the plural. So we have: caedes ac tumultus erat in castris (Liv. 10. 20. 10), but: quod passim eos simul pavor terrorque distulerant (id. 6. 42. 8) and: ubi ira et aegritudo permixta sunt (Sall. Jug. 68. 1). We read: si pax veniaque ab dis impetrata esset (Liv. 1. 31. 7), but: postquam pax et concordia speciosis et inritis nominibus iactata sunt (Tac. Hist. 2. 20. 3); and we have the same pairs now with the singular, now with the plural, in: senatus populusque Romanus intellegit (Cic. Fam. 5. 8. 2), and: si antidea senatus populusque iusserit (Liv. 22. 10. 6); but: cum senatus populusque Romanus pacem comprobaverint (Liv. 37. 45. 14), and: auctor essem senatui populoque Romano, ut eam vos habere sinerent (Liv. 36. 32. 5). And in Livy: tempus et locus convenit (1. 24. 2); but in Tacitus: ubi locus veneficii tempusque composita sunt (Ann. 4. 10. 2). And turning to names of persons: Palatium Romulus, Remus Aventinum ad inaugurandum 95 templa capiunt (Liv. 1. 6. 4), but: Tros Tyriusque mihi nullo discrimine agetur (Aen. 1. 574). We have here to do with pairs that are potential duals, and so are constructed now with a verb in the plural, now in the singular.

The opposite will also be found: when two subjects are disjoined by οὔτε or , by neque or aut, while they will usually take a singular, they will be constructed with the plural at times. So we have in Homer:

οὐδ’ ἐδύναντο

οὔθ’ ὁ τὸν ἐξελάσαι καὶ ἐνιπρῆσαι πυρὶ νῆα

οὔθ’ ὁ τὸν ἂψ ὤσασθαι (Il. 15. 416-8),

and in a fragment of Bacchylides: θνατοῖσι δ’ οὐκ αὐθαίρετοι οὔτ’ ὄλβος οὔτ’ ἄκαμπτος Ἄρης οὔτε πάμφθερσις στάσις (Fr. 36, B.), and in Euripides: καί μ’ οὔθ’ ὁ Πλούτωνος κύων οὔθ’ οὑπὶ κώπῃ ψυχοπομπὸς ἂν Χάρων ἔσχον (Alc. 360-2); and with in: ὅταν ἀδελφὸς ἢ ἀδελφή τῳ γένωνται καλοί (Plat. Leg. 838 A). So in Latin: haec si neque ego neque tu fecimus (Ter. Ad. 103), erant enim quibus nec senatus gloriari neque princeps possent (Plin. Pan. 75); and with aut in: si quid Socrates aut Aristippus contra morem consuetudinemque civilem fecerint locutive sint (Cic. Off. 1. 148. 41), or: quaerere puerum aut puellam qui supponantur mihi (Pl. Truc. 403-4). Seu . . . seu with a plural verb is recorded only in: et me seu naturalis sollicitudo seu fides sedula, non ad diligentiam modo, verum ad amorem quoque commissae rei instigent (Frontin. de Aquaed. Praef.), and tam . . . quam in: ut proprium ius tam res publica quam privata haberent (ib. 128).

When we have more than two subjects connected by et or καί, in Latin the verb usually agrees with the last of its subjects, in Greek with the nearest, as in; aetas et forma et super omnia Romanum nomen te ferociorem facit (Liv. 31. 18. 3), ζῶντι τῷ δικαίῳ . . . ἆθλά τε καὶ μισθοὶ καὶ δῶρα γίγνεται (Plat. Rep. 614 A). But in: ἡμεῖς δή, ἐγὼ καὶ Στράτιος καὶ Στρατοκλῆς . . . παρεσκευάζοντο ἅπαντες λαγχάνειν (Isaeus, 11. 15) the agreement presents difficulty. Probably we have the third plural dependent on the pair Stratius and Stratocles.

What is the construction with plurals joined with δύο? With masculine and neuter plurals the dual may be used as in: ἐξ ἧς αὐτῷ γίγνεσθον υἱεῖς δύο (Isae. 8. 10), and: δύο πυρὸς σώματα εἰς ἓν συνίστασθον εἶδος ἀέρος (Plat. Tim. 56 E). With masculines the plural is the rule, as in: τῆς δ’ ἀρχῆς αὐτῷ λοιποὶ δύο μῆνες ἦσαν (Antiphon 6. 42), while with neuters the singular is usual, as in: δύο δέ μοι τῆς κατηγορίας εἴδη λέλειπται (Aeschin. 1. 116). At times, 96 however, we have neuters plural taking a singular and a plural verb in successive clauses, as in:

καὶ δὴ δοῦρα σέσηπε νεῶν καὶ σπάρτα λέλυνται (Il. 2. 135).

Very curious is the union of a singular with a plural verb in:

τῷ δ’ οὔ τι γυνὴ καὶ νήπια τέκνα

οἴκαδε νοστήσαντι παρίσταται οὐδὲ γάνυνται (Od. 12. 42-3).

We may have here a use of a singular and a plural with νήπια τέκνα, but more probable seems the use of both singular and plural with the dual Union: γυνὴ καὶ νήπια τέκνα.



We have already spoken (pp. 64 ff.) of the development of the Schema Alcmanicum from a form of the extended elliptical dual, which was perfected in the Skt. Mitrā Varunā, and becomes a plural in Veneres Cupidinesque. We noted in Irish: doronsat sid ocus Fergal ‘they made peace and Fergal’, for ‘he and Fergal made peace’. In the Latin life of Fintan we read: venit Fintanus . . . et salutaverunt se in vicem et Lasserianus. If in the second clause we supply Fintanus from the first, we have: Fintanus salutaverunt se in vicem et Lasserianus, a construction exactly parallel to several examples of the Schema Alcmanicum. In the development of the extended elliptical dual we seem to have three stages: (1) the dual Mitrā for Mitra and Varuna, (2) the addition for clearness of Varuna to this dual, (3) the assimilation of this Varuna to the preceding Mitrā in Mitrā Varunā ‘the two Mitras, the two Varunas’. In the examples from Anglo-Saxon: wit Scilling ‘we two Scilling’ for ‘I and Scilling’, and in that cited from Irish we seem to have the second stage. The assimilation to the third stage in Greek seems to have been, not progressive as in Sanskrit, but retrogressive. Take the example we have from Alcman: Κάστωρ τε πώλων ὠκέων δματῆρες, ἱππόται σοφοί, καὶ Πολυδεύκης κυδρός (Fr. 9); this was probably before the retrogressive assimilation: Κάστορές τε πώλων ὠκέων δματῆρες κτλ., which makes the δματῆρες intelligible. But Κάστορες was assimilated in number to the following Πολυδεύκης. In: Τυδεὺς μάχην συνῆψε Πολυνείκης θ’ ἅμα (Eur. Suppl. 144) and: Αᾶσός ποτ’ ἀντεδίδασκε καὶ Σιμωνίδης (Ar. Vesp. 1410) the verb has also been assimilated to the singular, leaving now nothing of the Schema but its peculiar order of words.

Of the examples we have of this curious figure we have already cited:

ἧχι ῥοὰς Σιμόεις συμβάλλετον ἠδὲ Σκάμανδρος (Il. 5. 774),

where we still see the dual, the primary number of the figure. This dual has become plural in:


ἦ μὲν δὴ θάρσος μοι Ἄρης τ’ ἔδοσαν καὶ Ἀθήνη (Od. 14. 216),

and in:

ἔνθα μὲν εἰς Ἀχέροντα Πυριφλεγέθων τε ῥέουσι

Κώκυτός θ’ (Od. 10. 513-4).

The verb is omitted, and we have a predicate substantive in the example cited from Alcman. We have for καί or δέ in:

εἰ δέ κ’ Ἄρης ἄρχωσι μάχης ἢ Φοῖβος Ἀπόλλων (Il. 20. 138),

and we have the Schema transferred to objects in:

πέμψε δ’ Ἑρμᾶς χρυσόραπις διδύμους υἱοὺς ἐπ’ ἄτρυτον πόνον,

τὸν μὲν Ἐχίονα, κεχλάδοντας ἥβᾳ, τὸν δ’ Ἔρυτον (Pind. Pyth. 4. 178-9),

with which compare the example already cited (p 65) from Pind. Isth. 4. 17 ff.

While in Sanskrit the dual is used strictly to designate two objects, without dvā when the duality of the objects is well understood, otherwise with dvā or ubhau, in Greek its use is occasional and irregular. From the number and nature of its inflexions we concluded that this is its natural use in a language not developed into a book-language, like Gothic or Sanskrit. It uses δύω at times with the natural dual, as in: Ἀτρείδα . . . δύω (Il. 1. 16), or Αἴαντε δύω (Il. 12. 335), while the developed duals are used in Homer with or without a δύω or ἄμφω; as in: δοῦρε (Il. 13. 241), but δοῦρε δύω (Il. 3. 18); λέοντε (Il. 16. 756), but λέοντε δύω (Il. 5. 554); θῆρε δύω (Il. 15. 324), but κάπρω (Il. 11. 324), αἰετώ (Od. 2. 146), and γῦπε (Od. 11. 578). Then we have the plural with δύω in: Αἴαντές τε δύω (Il. 13. 313), δύω κήρυκας (Il. 3. 116), δύω ἵππους (Il. 8. 290), δύω νύκτας δύο τ’ ἤματα (Od. 9. 74), and Αἴαντε δύω, θεράποντες Ἄρηος (Il. 10. 228). With ἄμφω the dual is more the rule, since the plural ἀμφότεροι is in use; but we find it with the plural in the Odyssey: ἄμφω χεῖρας (8. 135), ὀφθαλμοί τε καὶ οὔατα καὶ πόδες ἄμφω (20. 365), ἄμφω φάεα καλά (19. 417) and even in the Iliad: ἄμφω . . . ὀφθαλμοί (16, 348-9). From these examples it is plain that the Greek dual does not give so clear a distinction of meaning as does the Sanskrit dual in: vedam, vedau, vedānva ‘one veda, or two, or more than two’. We have also plural adjectives joined to dual substantives, as in: ἄλκιμα δοῦρε (Il. 16. 139), ὄσσε φαεινά (Il. 13. 435).

The use of the masculine dual in: τώ γ’ ὢς βουλεύσαντε (Il. 1. 531), for Achilles and Thetis, seems to us natural enough. But when both are female, the epic still uses the masculine dual, as in: προφανέντε ἀνὰ πτολέμοιο γεφύρας (Il. 8. 378), and in: πληγέντε κεραυνῷ (Il. 8. 455), 99 where the reference is to Hera and Athena; or in Hesiod’s Works and Days:

ἴτον προλιπόντ’ ἀνθρώπους Αιδὼς καὶ Νέμεσις (199-200).

We must remember that by this time the old dual for stems had become a nominative plural, and the new dual in was not yet developed except for masculines like Ἀτρεΐδα (Il. 1. 16), αἰχμητά (Il. 7. 281), ἵππω ὠκύπετα (Il. 8. 41-2), Αἴαντε κορυστά (Il. 18. 163). So the old Epic has no feminine form for the dual, a form that develops only in Attic. We read in Thucydides: ἄμφω τὼ πόλεε (5. 23. 1), in Xenophon: ἄμφω τούτω τὼ ἡμέρα (Cyr. 1. 2. 11), but Sophocles gives us:

τὼ δ’ εὐχλόου Δήμητρος εἰς προσόψιον

πάγον μολοῦσα (O. C. 1600-1),

where Jebb prefers the μολοῦσαι of one manuscript; but though forms in are rare even in Attic, the inscriptions give τὼ στήλα, ταμία and the like, and, as the more difficult reading, μολοῦσα is to be preferred.

For the first dual Homer uses νῶϊ (νώ twice, in Il. 5. 219 and Od. 15. 475) for the nom.-acc. and νῶϊν for the gen.-dat. The verb following is always in the plural for the active; but Homer has once the first dual middle in περιδώμεθον (Il. 23. 485), where νῶϊ is not expressed. When we remember that in English ‘us’ is plural, but ‘we’ (= two) seems primarily dual, we may deduce from this syncretism of dual and plural in the pronoun a very old syncretism in the first plural of the verb, where all separate forms for the first dual seem late developments. Marstrander says: Tout compte fait, la hitt. 1re plur. -o -en me semble reposer sur la vieille forme du duel (Caractère indo-européen de la langue hittite, p. 152). The Skt. first duals ending in -va, -vas, -vahi, and -vahe, have nothing cognate in the verbs of the western branches of Indo-Germanic; and the evident significance of these endings (va = two) also indicates late development. We may conjecture that of the two surviving Greek inflexions for the first plural, -μεν and -μες, one was originally dual. When the verb has the same form for the dual and the plural, as is the case in the Greek active, it is natural that the dual pronoun should be attracted by it, and should show a plural meaning at times even in Homer. And in: νῶϊν ἀνήκεστος χόλος ἔσται (Il. 15. 217) νῶϊν evidently stands for the five divinities, Poseidon, Hera, Athena, Hermes, and Hephaestus.

In the late Epic of Quintus νῶϊ and νῶϊν are plurals, used on 100 occasion for two, just as is ἡμεῖς or ἡμῖν. In 3. 485 Quintus uses νῶϊν for two, and in 10. 31 it is a dual, though joined with a plural participle. In: λίπομεν βλαφθέντε νόημα (9. 492) we have the first plural joined with a dual participle to express two, a union easily paralleled in Homer. But in 1. 213, 369, 583, 725, &c., νῶϊ and νῶϊν are plurals. In 13. 344 Quintus uses ἑάς for ἡμετέρας, following the use of ἑαυτόν for ἐμαυτόν. Indeed by the time of Aristotle the dual had ceased to exist as a separate number; and when Aristotle or Lucian uses ἄμφω with a singular verb, it is the use of a singular verb with a neuter plural.

We found the dual joined formally with three in:

αἴ κ’ ἀποκηδήσαντε φερώμεθα χεῖρον ἄεθλον (Il. 23. 413),

though in meaning the participle is to be distinguished from the verb. In:

τὼ δὲ βάτην παρὰ θῖνα πολυφλοίσβοιο θαλάσσης (Il. 9. 182),

while there is no formal irregularity, we feel that τὼ βάτην is really of three, not of two, for Phoenix accompanied Nestor and Ulysses. There is a curious contrast between the passages; in the first, the two horses, that are to do the real work in the race, determine the number of the verb; in the second, Phoenix, the guardian of Achilles’ boyhood, who is to take a leading part in reconciling him to Agamemnon, is left out of account to all appearance in both subject and verb.

Just as in γυνὴ καὶ νήπια τέκνα (Od. 12. 42) we have a dual union, so in:

σκευή τε γάρ σε καὶ τὸ δύστηνον κάρα

δηλοῦτον ἡμῖν ὄνθ’ ὃς εἶ (Soph. O. C. 555-6),

σκευή; is felt as a collective singular. In:

αὐτανεψίω πατὴρ ἂν εἴη σός τε καὶ τούτων γεγώς (Eur. Heraclid. 211-12),

the use of γεγώς for γεγῶτε in agreement with αὐτανεψίω is due to the attraction of the intervening εἴη. In Il. 5. 487-8 we noticed the union of ἁλόντε with γένησθε, where the Scholiast explained ἁλόντε as in agreement with σὺ καὶ οἱ ἄλλοι to be supplied. Pindar gives us:

σοφὸς ὁ πολλὰ εἰδὼς φυᾷ·

μαθόντες δὲ λάβροι

παγγλωσσίᾳ κόρακες ὣς ἄκραντα γαρύετον

Διὸς πρὸς ὄρνιχα θεῖον (Ol. 2, 94-7),

‘wise he who in his nature has knowledge; but they, who have but learned, boisterous in multitude of words, are but as crows that 101 chatter idly against the divine bird of Zeus’. The Scholiast refers μαθόντες γαρύετον to Simonides and his nephew Bacchylides; but Mr. Verrall thinks it a hit at the new art of rhetoric, taking the κόρακες for Corax and Tisias, a parallel to uses like Castores or Assaraci (Aen. 10. 124). In the verses we have a parallel drawn first in general terms by the opposition of a singular to a plural, and then passing to the designation of two particular persons by a dual.

To conclude, when we see the irregularities in agreement of the dual with the plural in: νῦν μὲν γὰρ οὕτως, ἔφη, διάκεισθον, ὥσπερ εἰ τὼ χεῖρε, ἃς ὁ θεὸς ἐπὶ τῷ συλλαμβάνειν ἀλλήλοιν ἐποίησεν, ἀφεμένω τούτου τράποιντο πρὸς τὸ διακωλύειν ἀλλήλω, ἢ εἰ τὼ πόδε θείᾳ μοίρᾳ πεποιημένω πρὸς τὸ συνεργεῖν ἀλλήλοιν, ἀμελήσαντε τούτου ἐμποδίζοιεν ἀλλήλω (Xen. Mem. 2. 3. 18), we may well feel that by Xenophon’s time all feeling for the dual as a number distinct from the plural is lost; and that, while forms like νῶϊ may remain in use, they will be treated as plural forms, just as are θύραι or κρᾶναι δύο (Theocr. 5. 47).



Is the Latin for ‘mile’ mille passus or mille passuum? The question may be put in a more general form: is mille an adjective or a noun? Now, in the grammars of the Greeks and Romans the ἐπίθετον or nomen adiectivum is not yet separated from the ὄνομα or nomen substantivum. Gens is a noun; but ingens, its negative form, is always an adjective. Our rule is that mille in the singular is an adjective, as in: mille passus, but in the plural a noun governing the genitive, as in: duo milia passuum. Centum, too, is an adjective, but its plurals ducenti, trecenti, sescenti, are also adjectives, and Gellius (1. 16) cites several exceptions to the rule for mille, closing with the much debated passage in the Milo, where he reads: ante fundum Clodi . . . facile mille hominum versabatur valentium (53. 20). He further cites from Varro: plus mille et centum annorum, from Cato: mille passum, from Cicero: mille nummum (Phil. 6. 15. 5), and from Lucilius the ablatives milli passum and milli nummum.

Of these phrases mille passuum or passum seems the most common. Kühnast (Synt. Liv. 80) finds in Livy ten examples of mille passus, nine of mille passuum; Caesar and Cicero each use mille passuum once only, but the Agrimensores have only mille passus. Brix (ad Trin. 425) finds in Plautus mille nummum six times, mille passuum twice, and one example each of mille with annorum, drachumarum, medimnum and modiorum. In Terence we find mille with drachumarum (Heaut. 601) and with nummum (ib. 606). In Augustan poetry the construction is rare; it does not occur in Lucretius, Virgil, or Ovid; Horace has it once in mille ovium (Sat. 2. 3. 197). In return we find the plural used as an adjective in: sagittarios . . . tria milia numero habebat (B. C. 3. 4), quattuor milia pedes (Columella 5. 1. 10), decem amplius milia coissent homines (Flor. 3. 20. 3), tribus millibus nummis (Apul. Apol. 59), anni ad haec tempora prope duo milia sunt (Arnob. 2. 71). We may conclude that just as sahasra in Sanskrit, so mille and milia in older Latin could be used either as adjectives or nouns, but that later in both prose and verse mille was usually an adjective, milia a noun, in construction.


But Gellius cites from Lucilius: ad portam mille a porta est, sex inde Salernum ‘from gate to gate is a mile, six thence to Salernum’, where mille not mille passus, is the Latin for ‘mile’; a use of one for two like that of Pollux for Castor and Pollux. He also cites from Lucilius:

Non milli nummum potes uno quaerere centum,

‘out of one thousand sesterces you can’t get a hundred (thousand)’. His conclusion is: ita ‘unum mille’ et ‘duo milia’ certa atque directa oratione dicitur. No doubt the use of duo milia would naturally lead to that of unum mille, and we find it repeated in Macrobius (Sat. 1. 5. 7), where, however, he is plainly copying Gellius. Mille is by etymology sm̥-ghes-le, corresponding mutatis mutandis to the Skt. sa-has-ra, and cognate with χίλιοι (older *χεσλιοι). Gellius cites χίλιοι in this connexion, and we have in Herodotus χιλίην ἵππον ‘a thousand horse’ (5. 63. 3). Sm̥ is the reduced form of sem (= ἕν), so that in itself mille means ‘one thousand’; and probably that is why in Latin older than Gellius we have only one example of unum mille. To: divum promittere nemo (Aen. 9. 6) Servius notes: nemo pro nullus posuit, et est acyrologia: nam ‘divum nemo’ non possumus dicere, cum proprie ‘nemo’ sit ‘ne-homo’. Properly nemo is for ne-hemo, for which homo is a later assimilation. But Plautus wrote: nemo homo usquam ita arbitratust (Pers. 211). Nullus is for ne-unlus, a diminutive of unus; and yet Cicero writes: nulla re una magis oratorem commendari (Brut. 216. 59). Even more striking is the acyrologia that appears in Pliny’s: nullas duas . . . effigies (N. H. 7. 1. 8). If we found unum centum, it would be parallel to this; for centum, old centom, is likewise an example of the use of one for two, being for dekm̥ dkm̥tom ‘tens of tens’, as our ‘hundred’ is for hund-rede ‘rede’ or ‘tale of tens’. Our ‘thousand’ is not cognate with mille, but seems for stavas hund ‘strong in hundreds’. Cf. tauros, old stauros, our ‘steer’.

For duo milia Lucretius uses bis mille as a poetic variety in: bis mille sagittae (4. 408), as does Horace in: bis mille equos (Epod. 9. 17). Should they not have written bis milleni? I find no example of milleni older than the second century of our era. Camerarius read: millenum numero navium (Pl. Bacch. 928), but the manuscripts give mille cum. In Gaius I read: quinque hominibus singulis, millenos asses legando (2. 225). It is probable that bis milleni was not yet developed in Horace’s day, and that bis mille was the only form in use. With the multiplicatives we often find the cardinals, even when 104 the corresponding distributives are already in use. We have in Caesar: vicies centum milium passuum (B. G. 5. 13) and in Livy: decies centum milia (43. 6. 11) and in Pliny: quater mille sescentos (N. H. 36. 13). In poetry it is very common; Ennius gives us: ter quattuor corpora sancta avium (Ann. 90, M.) and lumina bis sex (ib. 344), and Cicero: bis sex ardentia signa (Arat. 568). Virgil has: ter denis navibus (Aen. 10. 213) but: ter centum annos (ib. 1. 272). Since duodecim and quattuordecim do not fit into a hexameter, the dactylic poets resort to variants with bis. Virgil uses bis seni thrice, but bis sex four times; he has bis septem twice, but never bis septeni. Horace has: bis quinque viri (Ep. 2. 1. 24) and: bis trium ulnarum toga (Epod. 4. 8). Ovid gives us: volucrum bis quattuor (Met. 12. 15), lisque decem decies inspicienda viris (Trist. 2. 94), milia qui novies distat ab urbe decem (ib. 4. 10. 4), and Martial: si dederint superi decies mihi milia centum (1. 103. 1). We have in Chalcidius phrases like: bis duo quattuor, and: bis duo bis, quod est octo. Nonius quotes twice from Varro: semel unum singulum esse; Ausonius has: ter tria multiplicanti (Edyll. 11. 2); iuris idem tribus est quod ter tribus (ib. 11. 4); ut idem congrege ter trino per ter tria dissoluatur (ib. v. 52-3). Neither Greek nor Sanskrit has a distributive numeral, showing it to be a later formation, and the correspondence of bis mille with δισμύριος and of the idiom in ter quattuor or bis septem with Skt. trisapta and tridaśa make it certain that the use of cardinals with the multiplicative is the primary one.

In Latin from the numeral adverbs bis, ter, quater, are formed by the addition of the stem suffix -no plural adjectives: bini, trini or terni, quaterni, and on this analogy quini, seni, &c., usually called distributives. Priscian tells us: alios quoque vocales . . . solent ex duabus syllabis in unam longam transire, ut . . . bis uni bini (Gramm. II. 126. 23). This is not quite valueless for the derivation of bini; for Priscian evidently felt that the suffix -no here is the same that we find in unus. Closely connected with bini is unus in the series una, bina, trina castra; and as trina is older than terna just as tris is the older form of ter (cf. agros and ager, Skt. ajras and ἀγρός) we may assume that this is the series for which the distributives were first devised. Further this series seems to me to present fewer exceptions than any other use of these numerals. Cicero indeed tells us: bina enim non dicuntur nisi de bis quae sunt numeri tantum pluralis, as cited by Servius in his note to: frenaque bina (Aen. 8. 168); where he adds that Cicero in a letter to his son blamed him for having written 105 ‘direxi litteras duas’; for litterae, when it means ‘epistle’, is numeri tantum pluralis, and bina, not duo, is used with such. But from what I have already said, you can see that duo castra was older than bina castra; and it is significant that Cicero’s boy still uses duo. Servius himself is not blameless here; he says of Virgil’s frena bina: poetice, nam duo debuit dicere. But just as litterae takes binae as being a plural of parts, a combination of many letters, so frena, like frenum and freni, means ‘a bridle’, and is a plural of parts. ‘Two homes’ is usually binae aedes, as ‘one home’ is unae aedes, but a natural plural of unus is duo, and so Ulpian writes: si quis duas aedes habeat, et alteras tradat (Dig. 8. 4. 6). Very curious is the lack of agreement in:

en quattuor aras:

Ecce duas tibi, Daphni, duas altaria Phoebo (Buc. 5. 65-6),

an example, it seems to me, of the σχῆμα πρὸς τὸ συνώνυμον; i.e. in duas altaria, where altaria means aras, duas is made to agree with aras, the word understood, and not with altaria, the word expressed. Servius agrees with the manuscripts in his citation of the verse; but when he cites it again, apparently from memory, in his Scholium to Aen. 3. 305, he has duoque altaria Phoebo, showing how ready he himself was to use duo with a plurale tantum. Caper tells us: binas tabulas dicimus, non duas (Gramm. VII. 108. 8); he is dealing with the plural of the union of two tablets, that form a note-book; just as Cicero was dealing with the plural of the union of many letters, which makes up an epistle. In: decreverunt . . . equos duo phaleratos, bina equestria arma (Liv. 30. 17. 13) we have the uses of the cardinal and the multiplicative side by side.

The multiplicative adverbs are constructed in two ways: either with a numeral they multiply as in: bis duo quattuor, or with a numeral limiting them as in: bis uno die venit. In the first we find the multiplied numeral often attracted by the multiplier into the multiplicative form, as in bis bina quattuor. We have in Macrobius: bis bina, quae sunt quattuor . . . bis bina bis, quae sunt octo . . . ter terna, quae sunt novem . . . ter terna ter, id est, ter novena, quae sunt viginti septem (Somn. Scip. 1. 6), quae si bis bina quot essent didicisset (Cic. N. D. 2. 49. 18), ternos ter cyathos . . . petet (Hor. Od. 3. 19. 14), ter bibe vel totiens ternos (Auson. Edyll. 11. 1), ter decies ternos habeat, deciesque novenos (ib. V. 90). The form terna for trina now appears, and seems clearly due to assimilation to the preceding ter in ter terna. For bis iterum is a variant in: sub Sereno bis et Serviano 106 iterum consulibus (Spart. Hadr. 3. 8), and we find it used for bis in: octonis iterum natalibus actis (Ov. Met. 13. 753). We find bis with ordinals also, as in: bis tertia ducitur aestas (Pont. 4. 10. 1), bis decimus . . . ab urbe lapis (Mart. 4. 57. 4); and such a use seems natural when we turn back to bis sex and bis septem. That bini is primarily not distributive, but multiplicative, seems certain from its use in: binos ducentos Philippos (Pl. Bacch. 1050), where binos is plainly for bis, an adjective used for an adverb, as we find in: domesticus otior (Hor. Sat. 1. 6. 128) or vespertinumque pererro . . . forum (ib. v. 113-14).

We notice how in: ter bibe vel totiens ternos the repetition of ter is avoided. In: saepe tribus lectis videas cenare quaternos (Hor. Sat. 1. 4. 86) the ter to be implied from tribus with quaternos is omitted, as we shall prove is usual with similar pairs. So in: duo quidam gladiis subcincti, ac bina iacula gestantes (Suet. Jul. 84. 3) the bis implied from duo is omitted with bina, as it is in: (Fabius et servus) iere pastorali habitu, agrestibus telis, falcibus gaesisque binis armati (Liv. 9. 36. 6) and: Thracum et Illyriorum . . . par numerus, bina milia erant (ib. 33. 4. 4). The number of consuls is so well known that in: consulibus bellum cum Hannibale et binae legiones decretae (ib. 25. 3. 3) both duo (or bini) and bis are omitted, and we have the two pairs, consulibus duobus . . . bis binae, represented by the first and last. So in: in ea (spatha) bina foramina tribus locis sunt (Cels. 8. 15), sunt bini (venti) in quattuor caeli partibus (Plin. N. H. 2. 119), liberi tres et tres liberae cum binis comitibus (Edict. Ulp. Dig. 25. 4. 1. 10), where we have a sentence parallel to Liv. 33. 4. 4, and bis is to be supplied with binis. With these omissions we have bini, terni, quaterni passing from multiplicatives to distributives.

When the multiplicative is limited by another number, we find a like omission frequent in the limiting number. We read: non semel, sed bis, neque uno, sed duobus pretiis unum et idem frumentum vendidisti (Cic. Verr. 2. 3. 179. 77), et unus rempublicam bis servavi, semel gloria, iterum aerumna mea (Sest. 49. 22), de eadem causa bis iudicium apiscier (Ter. Phorm. 406), where the limiting numeral is either expressed or clearly implied. But in: illic bis pueri die numen . . . tuum laudantes (Hor. Od. 4. 1. 25-7), dolia cum vino bis in die fac extergeantur (Cato, R. R. 26), alienus oves custos bis mulget in hora (Virg. Buc. 3. 5), semel in mense sulcos . . . fodere oportet (Cato, R. R. 43. 2), the numeral is omitted, not implied as in eadem.

Turning to corresponding multiplicative adjectives we have: ut in iugera singula ternis medimnis decidere liceret (Cic. Verr. 2. 3. 114. 48), 107 where to balance iugera singula and to account for the form of ternis we should have semel ternis medimnis. While I know semel unum, I have never met semel bini, semel terni, or the like. Semel, though logically required, seems never expressed in such unions; and the use of bina and terna here seems due to analogy with bina for bis bina and terna for ter terna in the examples we noted above. So we have: bini senatores singulis cohortibus praepositi (Liv. 3. 69. 8) singulas binae ac ternae naves circumsteterant (B. G. 3. 15), vix singulis aetatibus binos oratores laudabiles constitisse (Cic. Brut. 333. 97).

But in such clauses even the balancing number, like singuli, is very often omitted, as in: si inermes cum binis vestimentis velitis ab Sagunto exire (Liv. 21. 13. 7), cum bina iugera agri plebi dividerentur (id. 6. 36. 11), binae tunicae in militem exactae (id. 9. 41. 7), Graeci stipati, quini in lectulis, saepe plures (Cic. in Pis. 67. 27), nihil columbis fecundius . . . pulli nascuntur bini (Varro, R. R. 3. 7. 9), neque ea (vere) minus binis arandum, ter melius (ib. 1. 27. 2), (dentes) triceni bini viris adtribuuntur (Plin. N. H. 7. 16. 15. 71), compendi feci binos panes indies (Pl. Pers. 471), nam ex eis praediistalenta argenti bina statim capiebat (Ter. Phorm. 789-90), where statim is for quotannis. So in: quotannis annui bini reges creabantur (Nep. Hann. 7. 4), neque . . . quicquam . . . findi in bina secando (Lucr. 1. 532-3), bina die siccant ovis ubera (Buc. 2. 42), immane est vitium dare milia terna macello (Sat. 2. 4. 76). In such a phrase as: multibus singulis binas hastas dedit, both singulis ‘one at a time’ and binas ‘a pair’ seem still multiplicative in force, and the distributive idea comes from their union with dedit. But when singulis is omitted, as in: militibus binas hastas dedit, with the omission of singulis this distributive force at once passes to binas, now ‘two apiece’.

In: bini consules cum binis consularibus exercitibus ingrediebantur fines nostros (Liv. 23. 42. 9) this distributive force is present neither in bini nor in binis; bini consules is merely ‘the pair of consuls’, being no longer the shortened multiplicative for bis bini; and binis is evidently for duobus, being attracted by the preceding bini. In a letter of Pollio’s to Cicero we read: binis tabellariis in duas naves impositis (Cic. Fam. 10. 33. 3), where binis is again used loosely for ‘a couple of’ and the following duas has not yet been attracted to binas. To:

ipse uno graditur comitatus Achate

Bina manu lato crispans hastilia ferro (Aen. 1. 312-13),

Servius’s note is: bina si ad utrumque referas, bene dixit bina; si ad 108 Aeneam tantum, antiquus mos est bina pro duobus poni, sicut et duplices. The antique fashion was rather, as I have shown, to put duo for bini in all its uses. Either of Servius’s alternatives is possible here, and I feel like supplying uterque with crispans, or making crispans a singular in agreement with Aeneas et Achates, the dual implied in the preceding verse; else bina will merely be the usual pair of spears. Virgil has many uses of the multiplicative for the cardinal with the additional idea of union or connexion of some kind, so prominent in its primary use in: una castra.

In: necte tribus nodis ternos, Amarylli, colores (Buc. 8. 77) this idea of connexion is very plain; as it is in: quid ternas (litteras)? a-m-o (Pl. Merc. 304), quia boves bini hic sunt in crumina (Pl. Pers 317), binas aut amplius domos continuare (Sall. Cat. 20. 11), Xerxis et imperio bina coisse vada (Prop. 2. 1. 22), habent quaedam voces binas formas ut cervus cerva (Varro, L. L. 8. 47), tigna bina . . . inter se iungebat (B. G. 4. 17. 3), uti . . . terna Medusaei vincirem guttura monstri (Ov. Met. 10. 21-2), Stygius canis, qui trina vasto capita concutiens sono regnum tuetur (Sen. H. F. 783-5). In: aequora bina suis oppugnant fluctibus Isthmon (Ov. Her. 4. 105) we have a picture of the two seas eager to unite and to beat down the narrow isthmus that blocks their union. Interesting is the return of trina in a use so closely related to that of trina castra, where it was the only form in use. In: caedit binas de more bidentes (Aen. 5. 96) the de more seems to justify binas; but we have: centum lanigeras mactabat rite bidentes (Aen. 7. 93). In: bina . . . pocula (Aen. 9. 263-4) the bina seems justified by their common history that Virgil gives us; but in the pocula bina novo spumantia lacte, and the crateres duo pinguis olivi (Buc. 5. 67-8) we have in duo merely a poetic variety for bina, making Servius’s note: bina duo intelligible, as the use here of the distributive and the cardinal is parallel to that in bis bina and bis duo.

So in: septem ingens gyros, septena volumina traxit (Aen. 5. 85), septem numero, septenaque tela coniciunt (10. 329-30), centum cui bracchia dicunt, centenasque manus (10. 565-6), optet nunc bracchia centum centenasque in bella manus (Stat. Theb. 10. 294) we have not constantly the idea of a connected set in the multiplicatives to distinguish them from the cardinals. Such seems the case in: quattuor principes ferro interempti, trina bella civilia, plura externa ac plerumque permixta (Tac. Hist. 1. 2. 1-2), trinos soles . . . saepius videre (Plin. N. H. 2. 99), a Romanis quoque trinis bellis . . . lacessiti (Justin, 41. 1), an binis verbis respondeam (Apul. Apol. 103), trina elementa (Auson. Edyll. 109 11. 19), trina aenigmata (ib. 38), per trinas species trigonorum (ib. 50). Hence we understand the use already noticed of singulus for unus in Gellius and Augustine.

When objects are conceived of as thus connected and forming a set, it is not strange to find a collective singular taking the place of the plural; as we see in: terno consurgunt ordine remi (Aen. 5. 120), centenaque arbore fluctum verberat (ib. 10. 207-8), quae (dicta) trino iuvenis foro tonabas (Stat. Silv. 4. 9. 15), cessat centeni moderatrix iudicis hasta (ib. 4. 4. 43), amphora centeno consule facta minor (Mart. 8. 45. 4), hac mihi bis seno numeratur tessera puncto (ib. 14. 17. 1), gurgite septeno rapidus mare submovet amnis (Lucan, 8. 445), centeno gutture niti (Pers. 5. 6), (scrobes) non altiores quino semipede (Plin. N. H. 17. 11 (16). 80), bis bina cervice (Sedul. Carm. Pasch. 3. 90), ternaque te . . . Gratia . . . adflavit (Claudian, Laud. Ser. 88-9), trinum dicendi genus est (Auson. Edyll. 11. 66), and: et totiens trino cornix vivacior aevo; quam novies terni glomerantem saecula tractus vincunt aeripedes ter terno Nestore cervi (ib. 11. 12-14). We have binum used as a substantive in: (tetras intra se) bis binum tenet (Mart. Capella, 2. 107), and: bis binum locum tenet in ordine (Cassiodor. Inst. Div. 4).

This singular use of the multiplicative seems to take the place of the ordinal in: denaque luciferos luna movebat equos (Ov. Her. 11. 46), being ‘the connected set of ten moons’ used for ‘the tenth moon’. We found binos used for bis in Pl. Bacch. 1050; for: bisque septenos greges deplanxit una (Sen. Herc. Oet. 1850-1) we might have expected: binos greges septem natorum, &c. The irregularity seems due to a poetic hypallage. In: ubiquomque denis hastis corpus transfigi solet (Pl. Most. 358) corpus seems the careless use of an abstract for corpora singula. To: septenos orbibus orbes impediunt (Aen. 8. 448-9) Servius’s note is: septem scuta facta in unitatem connectunt; i.e. septenos limits, not the word orbes, but the phrase orbibus orbes. In:

Te duo diversa domini de parte coronant

Binaque serta tibi binaque liba ferunt (Ov. Fast. 2. 643-4),

the offering does not consist, as one might expect from the union of duo with the distributive bina, of four garlands and four cakes, but is rather a fourfold offering of a garland and a cake from each dominus.

In: tribunis septem . . . obviam exercitui missis quina nomina principum seditionis edita sunt (Liv. 28. 26. 5) it seems plain from: haud plus quam quinque et triginta (sec. 2) and: vicit sententia lenior (sec. 3) that from tribunis septem septies we must understand quina 110 nomina. Again in: (Tusci) in utrumque mare vergentes incoluere urbibus duodenis terras (ib. 5. 33. 9), from utrumque we must understand duas with terras, and from duas supply bis with duodenis. In: tres equitum numero turmae, ternique vagantur ductores (Aen. 5. 560-1) the number of each turma is quite clear from the: pueri bis seni quemque secuti (v. 561), and terni is evidently for tres; as the three leaders are named and described immediately after. But the first impression conveyed by terni is inevitably that it is for ter terni, giving an immediate idea of over a hundred horse.

Strangest of all is the use of the distributive in:

tum pendere poenas

Cecropidae iussi, miserum! septena quotannis

Corpora natorum (Aen. 6. 20-22).

From other accounts it is plain that it was seven youths and seven maidens that the Athenians were bound to send to Crete. If in Cecropidae, really an old dual, a noun masculine in meaning but feminine in ending for the Romans, we have a duality of meaning ‘the fathers and mothers of Athens’, then from it we may imply bis with septena; but this is far from plain. Conington’s note is: ‘the story mentioned seven youths and seven maidens, but Virgil has chosen only to name the former’. He was probably guided to this not unlucky guess by the masculine inflexion in natorum. We have seen how a pair finds expression by the name of either of the pair; e.g. the Dioscuri either by Castor or by Pollux. So here the seven pairs are evidently expressed by one of each; whether it is the youth or the maid is not so clear; but the poets favoured the use of Pollux rather than Castor for the Dioscuri. We have to reserve the solution of the puzzle in: ibant octonis referentes Idibus aera (Hor. Sat. 1. 6. 75) for a later chapter.

Bini, terni, quaterni, quini are styled distributives in our Latin grammars, and in Greek we find no corresponding class of numerals. But the Glossaries translate bina by δισσά, a multiplicative, as do the Itala and Volgata (v. Thesaurus 2. 1902), confirming the theory I have developed above, and showing that the Romans themselves felt that these numerals were really multiplicatives. Evidently the distributive force that seems to us essentially characteristic of them arises from the ellipses I have indicated.



Familiar to all students is the sense construction, the σύνταξις πρὸς σύνεσιν, where a collective noun singular in form, but conveying a sense of multitude or plurality, is constructed with a plural verb. We see it in:

αὐτὰρ ὀπίσσω

ἡ πληθὺς ἐπὶ νῆας Ἀχαιῶν ἀπονέοντο (Il. 15. 304-5),

and: ἅμα δὲ ἕῳ γιγνομένῃ καὶ ὁ ἄλλος στρατὸς ἀπέβαινον (Thuc. 4. 32. 2); in: civitati persuasit ut de finibus suis cum omnibus copiis exirent (Caes. B. G. 1. 2), and: inde pars per agros dilapsi, pars urbes petunt finitimas (Liv. 5. 40. 6). We have also thus constructed pronouns singular in form, but implying a plural, as in: ἔβαν οἶκόνδε ἕκαστος (Il. 1. 606) and: τότε μὲν δὴ ταῦτ’ εἰπόντες ἀπῆλθον ἑκάτερος ἐπὶ τὰ προσήκοντα (Xen. Cyr. 5. 2. 22); in: loquere, uter meruistis culpam (Pl. Men. 779), ubi quisque vident, eunt obviam (id. Capt. 500-1), and: eodem die uterque eorum ex castris stativis . . . exercitum educunt (B. C. 3. 30. 3).

At times we have the verb in agreement with the subject in form, while an adjective or participle in the predicate, that should stand in agreement with the subject, or a following verb constructed with it, is pluralized by the meaning. So we have: τὸ δὲ στράτευμα ἐπορίζετο σῖτον ὅπως ἐδύνατο ἐκ τῶν ὑποζυγίων κόπτοντες τοὺς βοῦς καὶ ὄνους (Xen. Anab. 2. 1. 6), παντί τε τρόπῳ ἀνηρέθιστο ἡ πόλις καὶ τὸν Περικλέα ἐν ὀργῇ εἶχον (Thuc. 2. 21. 3), multumque remittit qui revocent (Aen. 10. 839-40), nec supplex turba timebat iudicis ora sui, sed erant sine vindice tuti (Ov. Met. 1. 92-3), ut quis ex longinquo revenerat, miracula narrabant (Tac. Ann. 2. 24. 6).

Far more common is the parallel construction with τὶς in Greek, as we see it in: ἦ ἄρα δή τις, ὅσαι θεαί εἰσ’ ἐν Ὀλύμπῳ (Il. 18. 429), and: ἧσσόν τις ἐμοὶ πρόσειοι, δυσχερὲς ποιούμενοι (Thuc. 4. 85. 6), ἦ μάλα τις θεὸς ἔνδον, οἳ οὐρανὸν εὐρὺν ἔχουσιν (Od. 19. 40), αὐχμηρός γέ τις ὢν καὶ ἀπὸ παντὸς περιουσίαν ποιούμενος θησαυροποιὸς ἀνήρ· οὓς δὴ καὶ ἐπαινεῖ τὸ πλῆθος (Plat. Rep. 554 A). So, too, we find τὶς 112 in a dependent clause, joined with a plural pronoun in the principal sentence, in: ἐάν τις φανερὸς γένηται κλέπτων ἢ λωποδυτῶν ἢ βαλαντιοτομῶν . . . τούτοις θάνατός ἐστιν ἡ ζημία (Xen. Mem. 1. 2. 62), ἢν δὲ παρὰ ταῦτα ἀδικεῖν τις ἐπιχειρῇ, τούτοις Κῦρός τε καὶ ἡμεῖς πολέμιοι ἐσόμεθα (Xen. Cyr. 7. 4. 5).

So to a plural antecedent is joined a singular relative, which acquires a collective sense from a subjoined τὶς, as in:

ἄλλους μὲν γὰρ παῖδας ἐμοὺς πόδας ὠκὺς Ἀχιλλεὺς

πέρνασχ’, ὅν τιν’ ἕλεσκε, πέρην ἁλὸς ἀτρυγέτοιο (Il. 24. 751-2),

and in:

οἵτε καὶ ἄλλους

ἀνθρώπους πέμπουσιν, ὅτις σφέας εἰσαφίκηται (Od. 20. 187-8),

and: πάντας ἑξῆς, ὅτῳ ἐντύχοιεν, . . . καὶ γυναῖκας κτείνοντες (Thuc. 7. 29. 4), and: ᾧτινι ἐντυγχάνοιεν . . . πάντας ἔκτεινον (Xen. Anab. 2. 5. 32). So in archaic Latin we find at times a singular relative constructed with a plural antecedent, as in: fugitant omnes hanc provinciam, quoi obtigerat (Pl. Capt. 156-7) or in: viro, quoius mos maximest consimilis vostrum, ei se ad vos adplicant (Ter. Heaut. 392-3), where quoius is attracted in number to viro, though ei is its proper antecedent.

Often the idea of plurality in the pronoun leads to the assumption of a plural form by the pronoun itself, though it is primarily and properly singular, as in: ἕκαστοι ἔχοντες ξύλα καὶ οὗτοι(Herod. 2. 63), οὐκ ἂν ποιέοιεν τούτων οὐδέτερα (id. 7. 103. 4), palmas ponto tendens utrasque (Aen. 5. 233), primo impetu simul utraque cornua et Numidae et Carthaginienses pulsi (Liv. 30. 8. 7). This attraction is most easily traced in negatives, that imply the opposite. So we have: ἐκείνης τῆς νυκτὸς οὐδεὶς ἐκοιμήθη, οὐ μόνον τοὺς ἀπολωλότας πενθοῦντες, κτλ. (Xen. Hell. 2. 2. 3), where the οὐδείς implies πάντες with which πενθοῦντες is really in agreement. So with ἤρασσον in: εἶτ’ ὀνείδεσιν ἤρασσον ἔνθεν κἄνθεν οὕ τις ἔσθ’ ὃς οὔ (Soph. Ajax 724-5). But οὐδείς, which in older Greek is used only in the singular, and in Homer only in the neuter singular except in the phrase: τὸ ὃν μένος οὐδενὶ εἴκων, in later Greek develops a plural οὐδένες. So we read: ἄρχουσι δὲ οὐδένες ἔτι αὐτῶν ἀλλ’ αὐτονόμους ἀφιᾶσι (Xen. Lac. Rep. 3. 1), οὔτε γὰρ ἦν πρεσβεία πρὸς οὐδένας ἀπεσταλμένη τότε τῶν Ἑλλήνων (Dem. de Cor. 23), οὐδένων εἰσὶ βελτίους (id. Olyn. 2. 17) with which compare: οὐδενὸς βελτίους (Plat. Prot. 324 D).

So we find μηδένες developed, too, in: φοβούμενοι εἰ μηδένες ἄλλοι 113 ἢ αὐτοὶ πολεμήσοιεν τοῖς Λακεδαιμονίοις (Xen. Hell. 5. 4. 20). In Latin the use of nulli for nullus is very common; e.g. ut nulli supersint de inimicis (Cic. Marcell. 21. 7), ante Iovem nulli subigebant arva coloni (Virg. Geo. 1. 125), quis est enim, qui nullis officii praeceptis tradendis philosophum se audeat dicere? (Cic. Off. 1. 5. 2). In English the authorities for ‘none of us were present’ seem quite as weighty as those for ‘none of us was present’, though ‘none’ is short for ‘no one’; and the plural seems the result of attraction of the verb by the plural ‘us’. So in the οὐδένας τῶν Ἑλλήνων quoted from Demosthenes. Herodotus uses not οὐδείς, but οὐδαμός (= oude-sm̥-mos), and usually makes it plural, as in: οὐδαμοὶ Ἰώνων (1. 18), οὐδαμοῖσι τῶν νῦν σφέας περιοικεόντων (1. 57). So with μηδαμός in: μηδαμοῖσι ἄλλοισι Ἰώνων (1. 143), μηδαμοὺς τῶν προσοίκων (1. 144), where the plural seems the result of attraction by the following genitive.

So in: οὐ δεινὰ πάσχειν δεινὰ τοὺς εἰργασμένους (Eur. Or. 413) and: δῆλα γὰρ ὅτι σύμμαχοι βασιλέος γινόμεθα (Herod. 9. 11) δεινόν and δῆλον seem to have been attracted to δεινά and δῆλα by the following plurals δεινά and σύμμαχοι. So, too, in: οὐκ ἂν ποιέοιεν τούτων οὐδέτερα (Herod. 7. 103). We have similar attraction, though not now through dependent genitives, but through preceding plurals, in: δεῖξαι αύτοῖς ὅτι οὐκ Ἴωνες τάδε εἰσίν (Thuc. 6. 77) and: οὓς οὐ παραδοτέα τοῖς Ἀθηναίοις ἐστίν, οὐδε δίκαις καὶ λόγοις διακριτέα . . . ἀλλὰ τιμωρητέα ἐν τάχει καὶ παντὶ σθένει (Thuc. 1. 86. 3); just as we have attraction for gender in: τὸν πολλὸν τοῦ χρόνου (Herod, 1. 24), πολλὴν τῆς χώρας (Xen. Cyr. 3. 2. 2), οὗτος παρ’ ἐμοὶ τὸ οὔνομα τοῦτο δίκαιός ἐστι φέρεσθαι (Herod. 1. 32 fin.). We seem to have a like attraction in the common phrase: ἀδύνατά ἐστι ταῦτα ποιεῖν. The only case of a like attraction in Latin I have met with is the ‘nota tibi’ in:

Frater ut Aeneas pelago tuus omnia circum

Litora iactetur odiis Iunonis iniquae,

Nota tibi (Aen. 1. 667-9),

where the figure, formally Greek, does not follow exactly any of the patterns we have quoted. Probably the agreement is with the pericula Aeneae implied in: frater ut omnia circum litora iactetur.

But if attraction by the following ταῦτα is the primary reason for the use of ἀδύνατα noticed above, evidently the feeling soon develops that ἀδύνατα is a stronger and more emphatic form for ἀδύνατον, and we have here a use parallel to the Latin use of quia for quod. This we seem to have in: δήμου τε αὖ ἄρχοντος ἀδύνατα μὴ οὐ κακότητα 114 ἐγγίνεσθαι (Herod. 3. 82), Ἀπόλλων τάδ’ ἦν (Soph. O. T. 1329), κοὔτοι γυναικὸς οὐδαμῶς ἡσσητέα (id. Antig. 678), where attraction plays no part in the change. Indeed we have the opposite in: μέγα ποιεύμενος ταῦτα (Herod. 3. 42) and: εὐπετές τε αὐτοῖσι ἔφη ταῦτα γίνεσθαι (id. 9. 90), where in the use of the invariable predicate we have another mode of emphasis. We seem to have this use in the common form of question: τί οὖν ταῦτά ἐστιν; (Xen. Anab. 2. 1. 22), and: τί οὖν δή ἐστιν ἅττα εἶπεν ὁ ἀνὴρ πρὸ τοῦ θανάτου; (Plat. Phaedo 57 A). But in such uses of nullis as we have in: te adhuc a nullis nisi ab Siculis potuisse cognosci (Cic. Div. Caecil. 28. 9), intellegetis enim nullis hominibus quemquam tanto odio, quanto istum Syracusanis, et esse et fuisse (Verr. 2. 2. 15. 5), multaque saecula postea sic viguit Pythagoreorum nomen ut nulli alii docti viderentur (Tusc. 1. 38. 16) we have attraction on the same lines as in Greek.

To: multumque remittit qui revocent (Aen. 10. 839-40), Servius’s note is: hoc pro ‘saepe’. I prefer to translate here: ‘he sends many a one to recall him’; but in such examples as: multum sunt in venationibus (B. G. 4. 1) or: in agmine atque ad vigilias multus adesse (Sall. Jug. 96. 3) Servius’s translation seems quite suitable. We have the opposite in: memini, tametsi nullus moneas (Ter. Eun. 216), Philotimus non modo nullus venit, sed ne per litteras quidem . . . certiorem facit me quid egerit (Cic. Att. 11. 24. 4), ab armis nullus discederet (ib. 15. 22), a use that seems to me to take its simplest shape in: ex quo suffragia nulli vendimus (Juv. 10. 77-8). We seem to have a parallel use of non nullus in:

Sic ego tam sancti custode reludor amoris;

Ex illo felix non mihi nulla fuit (Prop. 2. 29. 41-2).

The editors have all chosen to read nox with the worse manuscripts instead of non, which is the reading of all of any value. The resultant meaning is impossible in view of hesterna nocte in the first verse; and so as a rule they try to divide the poem into two poems, and in this, seeing that the poem naturally falls into two parts, the revelry of the evening and the repentance of the morning, they have some apparent success. But Cynthia’s success in convincing her lover of her chastity was surely far more likely to render her propitious than cruel to him; and if we follow the best manuscripts, with ex illo implying lusu from reludor, we may translate: ‘as a result of that sport of hers my mistress has turned quite propitious to me’, felix fuit being a poetic expansion for favit.

In Latin the verb is usually in the plural when it has two or more 115 subjects joined by et, as in: et Q. Maximus, . . . et L. Paullus, . . . et vester Gallus, et M. Cato . . . eis temporibus fuerunt (Cic. Fam. 4. 6). But we often find, also, two subjects connected by cum with a plural verb. Gellius quotes from an oration of Cato: si sponsionem fecissent Gellius cum Turio (14. 2. 26), and we read in Terence:

Syrus cum illo nostro consusurrant (Heaut. 473).

Nepos has: Demosthenes cum ceteris . . . in exsilium erant expulsi (Phoc. 2); but Cicero gives us only one example: Sulla cum Scipione . . . leges inter se condicionesque contulerunt (Phil. 12. 27. 11); neither Caesar, Velleius, nor Tacitus offers an example. Virgil has: cana Fides et Vesta, Remo cum fratre Quirinus iura dabunt (Aen. 1. 292-3); and Ovid:

Litora cum plausu clamor superasque deorum

Implevere domos (Met. 4. 735-6),


Ilia cum Lauso de Numitore sati (Fast. 4. 54).

Sallust has it repeatedly, as in: at Romae Lentulus cum ceteris . . . constituerant (Cat. 43. 1), Bocchus cum peditibus . . . aciem invadunt (Jug. 101. 5). Livy offers several examples, as: ipse dux cum aliquot principibus capiuntur (21. 60. 7). But in both we find examples with the singular, as: ibi Iugurtha cum plurimis erat (Sall. Jug. 101. 6) and: illa . . . iusta acies cum ducibus, cum ordinibus media urbe in forum processit (Liv. 26. 46. 7). Interesting is Ovid’s verse: fors eadem Ismarios Hebrum cum Strymone siccat (Met. 2. 257), recalling Virgil’s: septenos orbibus orbes (Aen. 8. 448).

The corresponding construction with μετά is rare; I have noted: Δημοσθένης μετὰ τῶν συστρατηγῶν . . . σπένδονται Μαντινεῦσιν (Thuc. 3. 109. 2), Ἀλκιβιάδης ἐκ Σάρδεων μετὰ Μαντιθέου . . . εὐπορήσαντες νυκτὸς ἀπέδρασαν εἰς Κλαζομενάς (Xen. Hell. 1. 1. 10), πολυτελῶς Ἀδώνια ἄγουσ’ ἑταίρα μεθ’ ἑτέρων (Diph. apud Athen. p. 292 D). It is not found with σύν, nor with λαβών or ἔχων, to my knowledge; but Servius evidently thinks that in: ipse uno graditur comitatus Achate (Aen. 1. 312) we have a variety for: ipse cum Achate uno gradiuntur, making the bina in v. 313 short for bis bina. Rare is the use of a plural verb in Latin with two subjects joined by an, as we find it in: Roma an Carthago iura gentibus darent, ante crastinam noctem scituros (Liv. 30. 32. 2), and: tuque dubitas Cimberne Annius an Veranius Flaccus imitandi sint tibi? (Suet. Aug. 86. 3).



We find Plautus constructing the plural of the imperative with the singulars quis and aliquis, as in: Simoni me adesse quis nuntiate (Pseud. 1284), aperite aliquis (Merc. 131), heus aliquis actutum huc foras exite; illinc pallium mi ecferte (Merc. 910-11), aperite atque Erotium aliquis evocate ante ostium (Men. 674). These are evidently to be connected with the syntax in:

Ibitis Aegaeas sine me, Messala, per undas (Tib. 1. 3. 1),


ᾗ περ ἂν ὑμεῖς

νῆα παρὰ γλαφυρὴν ιθύνετε, φαίδιμ’ Ὀδυσσεῦ (Od. 12. 81-2),

which give a starting-point for the use of vous in French and you in English as a singular.

When we recall how readily we pass from the use of the indicative with quin to that of the imperative with it, as we see them in: i: quid stas, lapis? Quin accipis? (Ter. Heaut. 831-2), and in: quin tu uno verbo dic (And. 45), we are not surprised to find the indicative coupled with the imperative here, too, in: datin soleas, atque me intro actutum ducite? (Pl. Truc. 631). In: heus foras exite hue aliquis; duce istam intro mulierem (Epid. 398-9) we have the second plural exite coupled with the second singular duce. We have aliquis joined with the second singular of the present subjunctive in: exoriare aliquis nostris ex ossibus ultor (Aen. 4. 625); but as a rule aliquis or quis is joined with the third singular here, as in: aliquis dicat (Ter. And. 640).

But we read in Plautus: si quis me quaeret, vocatote aliqui (Stich. 67) ‘if any one asks for me, you shall call me, some of you’; and when we compare this: vocatote aliqui with: aliquis foras exile cited above, we see at once that its singular will be: vocato aliquis, and understand why the Romans used vocato for the third person as well as for the second. From: hoc tu facito cum animo cogites (Ter. Ad. 500) we get: hoc aliquis facito cum animo cogitet, where facito has passed to the third person. That dato was primarily second person, seems plain from the structure; it is the second singular imperative da + tod, the ablative singular neuter of the definite article, which is still plain in isto, and it means ‘give from this on’. Its transference from the second to the third person seems a natural consequence from the 117 usual construction of aliquis with the third person. Its use for the third person seems old; we read in a law referred to as ‘of Romulus’: si nurus, sacra divis parentum estod (Bruns, Fontes, p. 7. 13). The second plural datote is formed from the second singular dato on the analogy of date, and the third plural from the third singular dato, following the analogy of dat to dant in the indicative.

That this transfer of the second person to the third is pro-ethnic is clear from its presence in Greek and Sanskrit, and in both it seems due to similar causes. Parallel to dato in form is the Sanskrit dhattāt, a form which is usually second singular, as in: draviṇe ’ha dhattāt ‘bestow riches here’, but which is also used as a third singular in: bhavān prasādam kurutāt ‘may your honour do this favour’. It is also used occasionally for the second plural, the second dual, and the first singular (Whitney, 571 A and B).

In Greek we have a form parallel to dato in φερέτω, which is usually a third singular. But we read:

ἴτω τις εἰσάγγελλε Τειρεσίας ὅτι ζητεῖ νιν (Eur. Bacch. 173-4),

‘go, some one, announce that Tiresias is seeking him’, where ἴτω τις is an exact parallel to vocato aliquis. So, too, in:

στειχέτω τις ὡς τάχος

ἐλθὼν δὲ θάκους τοῦδ’ ἵν’ οἰωνοσκοπεῖ

μοχλοῖς τριαίνου κἀνάτρεψον ἔμπαλιν

ἄνω κάτω τὰ πάντα συγχέας ὁμοῦ,

καὶ στέμματ’ ἀνέμοις καὶ θυέλλαισιν μέθες (ib. 346 ff.).

The use of τὶς for σύ in:

τοῦτ’ εἰς ἀνίαν τοὔπος ἔρχεταί τινι (Soph. Aj. 1138),

and in: κακὸν ἥκει τινί . . . δώσει τις δίκην (Ar. Ran. 552-4) makes it easier to understand the use of τὶς with the second person. Its transference to the third is intelligible from such uses as: καὶ μηδείς γε ὑμῶν ἔχων ταῦτα νομισάτω ἀλλότρια ἔχειν (Xen. Cyr. 7. 5. 73), or: εἴ τινες ὑμῶν . . . μήτ’ ἀπογνώτω μηδὲν μήτε καταγνώτω πρὶν ἀκούσῃ (Aeschin. Ctes. 60).

We have a parallel to the use of aliquis with the second imperative in Attic, where τὶς, or πᾶς, or πᾶς τις is used with the second singular of the present imperative; as in: ἴσχε· θάρσει πᾶς . . . παῖε, παῖε πᾶς τις ἄν . . . ἴσχε πᾶς τις . . . ἴσχε πᾶς δόρυ . . . ἕρπε πᾶς κατ’ ἴχνος αὐτών (Eur. Rhes. 681 ff.), or:

ἄγε νῦν ἄγε πᾶς (Ar. Pax 512),


χώρεί δεῦρο πᾶς ὑπηρέτης·

τόξευε, παῖε, σφενδόνην τίς μοι δότω (Ar. Av. 1186-7),


(where we may note the use of δότω as a second singular), and in: φύλαττε πᾶς (ib. 1190). We may conjecture that the use of ἄν with παῖε is due to the association between the imperative and the optative with ἄν in the apodosis of conditional sentences, where for: εἰ τοῦτο ποιοῖς, ἁμαρτάνοις ἄν we may say: εἰ τοῦτο ποιεῖς, ἁμάρτανε. We have often the infinitive used for the imperative, courteously giving a bare signification of the act to be done, as in: ἔκδοτε καὶ τιμὴν ἀποτινέμεν (Il. 3. 459), or: νεκρὸν Ἀχαιοῖσιν δώσω πάλιν, ὣς δὲ σὺ ῥέζειν (Il. 22. 259), and in return we have the imperative used for the infinitive in: ὥστε μὴ λίαν στένε (Soph. El. 1172), or: ἐγὼ νὴ Δί’ ἐρῶ καὶ γράψω δέ, ὥστε, ἂν βούλησθε, χειροτονήσατε (Dem. 129. 83 B). So, if ἁμάρτανε is used for ἁμαρτάνοις ἄν, we may get ἁμάρτανε ἄν, ἁμάρτανε being felt as equivalent to ἁμαρτάνοις. Speijer notes that the imperative in Sanskrit is used in dubitative questions (Skt. Syn. p. 276); and we read in Plato: τί οὖν; ὃ πολλάκις ἐρωτῶ, κείσθω νόμος ἡμῖν καὶ τύπος ἐκμαγεῖόν τε τρίτον τοῦτο, ἢ πῶς δοκεῖ; (Legg. 801 D). Behind the indefinites τὶς and quis lie older interrogatives; so that in παῖε πᾶς τις ἄν we have ‘let all strike; who in that case will?’; or in: ἴτω τις ‘you shall go; who will?’ or in: quis nuntiate ‘who of you announce?’ or in: vocatote aliqui ‘ye shall call; who of you will?’.

Often in Greek we find the second singular imperative joined with a following plural imperative, or indicative, or subjunctive, or with two substantives joined by καί as its subjects. We read: ἀλλ’ ἄγε μίμνετε πάντες (Il. 2. 331), but: ἀλλ’ ἄγεθ’ οἱ ἓξ πρῶτον ἀκοντίσατ’ (Od. 22. 252), εἰπέ μοι, τί φειδόμεσθα τῶν λίθων, ὦ δημόται (Ar. Acharn. 319); φέρε, στήσωμεν ἡμέων αὐτῶν βασιλέα (Herod. 1. 97), but: φέρετε, τούτους ἀνευρόντες συγχέειν πειρᾶσθε αὐτούς (id. 4. 127), ἴδ’ οἷον, ὦ παῖδες, προσέμειξεν ἄφαρ τοὔπος τὸ θεοπρόπον ἡμῖν (Soph. Trach. 821-2), εἰπέ μοι, ὦ Σώκρατές τε καὶ Ἱππόκρατες, ὡς τίνι ὄντι τῷ Πρωταγόρᾳ ἐν νῷ ἔχετε χρήματα τελεῖν (Plat. Prot. 311 D), εἰπέ μοι, ἔφη, ὦ Σώκρατές τε καὶ ὑμεῖς οἱ ἄλλοι . . . πότερον παίζετε ταῦτα λέγοντες; (Euthyd. 283 B).

It is in similar use in Latin with age and cave, as in: age igitur intro abite (Pl. Mil. 928), age eamus ergo (ib. 78), age, alter istinc, alter hinc, adsistite (Rud. 808), cave dirumpatis (Poen. 117), facite illic homo . . . siet, . . . cave quisquam, &c. (Men. 992-4), agedum conferte nunc cum illis vitam P. Sullae (Cic. Sull. 72. 26), age nunc iter expediti latronis cum Milonis impedimentis comparate (Mil. 55. 21), age nunc illa videamus, iudices (Rose. Am. 105. 36), age sis nunc de ratione videamus (Tusc. 2. 42. 18), agedum, inquit, dictatorem . . . creemus (Liv. 2. 29. 11), procedat agedum ad pugnam (id. 7. 9. 8), mittite 119 agedum legatos circa omnes Asiae urbes et quaerite (id. 38. 47. 11), huc age adeste (Sil. 11. 169). The dum affixed to age we shall consider later; it is suffixed to agite, in: agitedum clamorem . . . tollite (Liv. 3. 62. 4), or: agitedum . . . obvios sternite (id. 7. 33. 10); and agite is used just like age in: agite et cuncti laetum celebremus honorem (Aen. 5. 58).

Most grammarians are inclined to associate this irregularity with the use of these imperatives as interjections, a species of adverb and so indeclinable, as Monro puts it. But we have ἄγετε and φέρετε so used in Greek as well as ἄγε and φέρε, and so with agitedum and agite in Latin as well as agedum and age. The form of the so-called second singular of the present imperative seems worth consideration. As we see it in voca, vide, fer, ἄγε, φέρε, it is the root or stem of the verb and nothing more; there is nothing in it to mark person or number, tense or mood. When -to(d) is added, as in vocato, we get a tense mark and a future imperative; and then voca is partially relegated to the present. And when -te is added, as in vocate, we have a second plural, and then voca is partially relegated to the singular. But that for age and cave in Latin this is not wholly the case, we have just shown; there the old form is often used for the plural as well. To begin with, this root form was probably used as a general imperative covering all persons and both numbers and tenses, and merely designating in the simplest form the act to be done, as does the infinitive when used as an imperative, such infinitives, for example, as we find used as imperatives in κλέψον and κλέψαι. When the forms vocato and vocate were evolved, this old and general form was not at once strictly confined to its later function of a second singular, but in verbs that were in very common use, like ago or ἄγω, it still continued to be used as a general imperative.

When the plural was once evolved, the command was at times represented as directed to the leader of a band, or to some one of their number, as in: νῆα . . . ἰθύνετε, φαίδιμ’ Ὀδυσσεῦ (Od. 12. 82), or: προσέλθετ’, ὦ παῖ, πατρί (Soph. O. C. 1104), or:

χωρεῖτε τοίνυν, ὦ Διόνυσ’, εἴσω. τί δαί;

ἵνα ξενίσω σφὼ πρὶν ἀποπλεῖν (Ar. Ran. 1479-80),

or: ἴτ’, ἔφη, ὑμεῖς, ὦ Ἡριππίδα, καὶ διδάσκετε (Xen. Hell. 4. 1. 11). It is extended to the indicative in: ὦ τέκνον, ἦ παρέστον; (Soph. O. C. 1102), or: Ἀντίνο’, οὔπως ἔστιν ὑπερφιάλοισι μεθ’ ὑμῖν δαίνυσθαι (Od. 2. 310-11), or: Τιμόσθενες, ὕμμε δ’ ἐκλάρωσεν πότμος Ζηνὶ γενεθλίῳ (Pind. Ol. 8. 19-20), or: καθῆσθαί μοι δοκεῖ εἰς τὸ Θησεῖον πλεούσαις (Ar. 120 Eq. 1311-12). We have the opposite in: ξεῖνοι, μὴ δῆτ’ ἀδικηθῶ σοι πιστεύσας καὶ μεταναστάς (Soph. O. C. 174-5). But from the majority we derive the use of the vocative for the first only in naming a number of persons addressed, as in:

Ζεῦ πάτερ, Ἴδηθεν μεδέων, κύδιοτε, μέγιστε,

Ἠελιός θ’ ὃς πάντ’ ἐφορᾷς καὶ πάντ’ ἐπακούεις (Il. 3. 276-7),

a construction found in Sanskrit as well.

The use of the imperative as an interjection seems rather a result of its lack of concord than the cause of it. In Latin some imperatives, that have passed to other functions, are interesting here. Take vel, the 2nd sing. pres. imperative of volo, which seems to have passed to an adverb in: Heus, pax, te tribus verbis volo. Vel trecentis (Pl. Trin. 963-4), where it seems to me that we can still recover something of its primary force. In: vel tu me vende vel face quid tibi libet (Pers. 398) it has passed to a conjunction, when it is regarded as another form of ve, the Greek and Sanskrit va, cf. a and ab. So eme, the imperative of emo, is shortened to em in: em, serva (Ter. Ad. 172) and: em tibi hominem (Pl. Capt. 540), and regarded as an interjection. So both esto and ἔστω are used as adverbs.

Still more interesting is the adverb that has taken on the form of an imperative. Herodian gives us δεύρω as the old form, which Dindorf adopted in Il. 3. 240; for in this form it is clearly parallel to ἄνω and κάτω. It is made up, according to Prellwitz, of δε -υ -ρω, where the δε is that in οἴκαδε and in quamde, being a reduced form of the -do in quando and in donec, the Latin cognate of our ‘to’. The υ is the υ in οὗτος, being a reduced form of αὖ ‘again’. For the shortening to the usual form δεῦρο, compare ἅμᾰ with the Doric ἁμᾶ. It is found coupled closely with the imperative, just as were the invariable imperatives with which we were just dealing; compare δεῦρ’ ἴθι (Il. 3. 130) with βάσκ’ ἴθι (Il. 2. 8). It is felt as equivalent to ἄγε in: δεῦρ’ ἐς τοὺς φύλακας καταβήομεν (Il. 10. 97) and in: δεῦρο παρὰ Σωκράτη (Plat. Theaet. 144 D). Consequently when it is coupled with ἄγετε we find it attracted to: δεῦτε ἄγετε (Il. 7. 350). Then the use of δεῦτε is extended, as in: δεῦτε φίλοι (Il. 13. 481), and it is coupled with the singular, as in: δεῦτ’ ἄγε (Od. 8. 11), where it takes the place and plays the part of the significant imperative, which ἄγε is used to introduce. This last use led Buttmann to regard it as a contraction for δεῦρ’ ἴτε and Monro to claim δεῦρο as the second singular imperative, of which δεῦτε was the second plural.



This use is plainest and most easily traced in Greek. There we have it used alone, as in: αὖθι μένειν (Il. 10. 65), or co-ordinate with an imperative, as in: ἔρχεσθον κλισίην . . . Ἀχιλῆος . . . ἀγέμεν Βρισηΐδα (Il. 1. 322-3), and with a future indicative in: εὐνάσω ἑξείης· σὺ δ’ εῢ κρίνασθαι ἑταίρους (Od. 4. 408). It is used in a prohibition in: μὴ δή μοι ἀπόπροθεν ἰσχέμεν ἵππους (Il. 17. 501), and is repeated in this use in: ἀλλ’ εὖ οἱ φάσθαι πυκινὸν ἔπος ἠδ’ ὑποθέσθαι καί οἱ σημαίνειν (Il. 11. 788-9). It is joined with an imperative in the third person in: τεύχεα συλήσας φερέτω κοίλας ἐπὶ νῆας, σῶμα δὲ οἴκαδ’ ἐμὸν δόμεναι πάλιν (Il. 7. 78-9) and evidently stands for δότω, and its subject is in the accusative in: Ζεῦ ἄνα, Τηλέμαχόν μοι ἐν ἀνδράσιν ὄλβιον εἶναι (Od. 17. 354). In the general prayer: Ζεῦ πάτερ, ἢ Ἀἴαντα λαχεῖν (Il. 7. 179) it seems used for the optative; and in αἲ γάρ, Ζεῦ τε πάτερ . . . οἷος Νήρικον εἷλov . . . ἐφεστάμεναι καὶ ἀμύνειν ἄνδρας μνηστῆρας (Od. 24. 376-81) it is of the first person. Its use is not confined to Homer; we read: γυμνὸν σπείρειν, γυμνὸν δὲ βοωτεῖν, γυμνὸν δ’ ἀμάειν (Hes. Opp. 391-2), ὦ Ζεῦ, ἐκγενέσθαι μοι Ἀθηναίους τίσασθαι (Herod. 5. 105); πείθεσθε οὖν . . . καὶ παραστῆναι (Thuc. 6. 34. 9), οἷ δ’ ἱκάνομεν, φάσκειν Μυκήνας τὰς πολυχρύσους ὁρᾶν (Soph. El. 8-9), ἀκούετε λεῴ, τοὺς γεωργοὺς ἀπιέναι κτλ. (Ar. Pax 551), οἱ ἀπόστολοι . . . τοῖς . . . ἀδελφοῖς τοῖς ἐξ ἐθνῶν, χαίρειν (Acts 15. 23). A survey of the few examples I have selected shows how vain it is to try to account for this use of the infinitive through the influence of a previous imperative, as Monro tries to do for Homer. In a very large number of examples there is no imperative preceding, and where it does, it may come long before, as in the example quoted from Thucydides.

But this use of the infinitive is very old, and certainly was usual among the primitive Indo-Germanics. It is present in the German branch, and is in constant use in Germany to-day in such phrases as Schritt fahren, Rechts gehen. Whitney speaks of the use of the infinitive in Sanskrit without a copula that has quite nearly the value of an imperative, as in: tān nāi ’vaṁ kartavāi (R. V.) ‘that must not be done so’ (Gram. 982 D); and adds ‘The infinitives in -dhyāi and 122 -ṣaṇi . . . are those in which the imperative value is most distinctly to be recognized’. For Zend, Reichelt (Awest. El. 703) gives me many examples of the use, mostly for the imperative, but in one case in union with an optative, and with optative meaning as in Homer. Any account of its use must be founded on the nature and structure of the infinitive itself rather than on a chance connexion with a preceding imperative. And such is the account Apollonius Dyscolus gives us of this idiom, as we shall see.

It is natural to think that so old and usual an idiom must have been present in early Latin, as well as in Greek. There are but two direct examples of its use that I can cite; though it seems to me that its presence is one of the most prominent facts in the language, when we try to account for the modal uses of the Latin infinitive. In Valerius Flaccus we read: tu socios adhibere sacris (3. 412), usually set down as a Hellenism. But in: quod superest, sufferre pedes properate laborem (Prop. 3. 21. 21) the Itali and Baehrens change sufferre to sufferte. Servius gives us: si vobis audentem extrema cupido certa sequi (Aen. 2. 349-50), but adds that some read audendi, others audenti; sed neutrum procedit. Ergo ‘audentem’ legendum est. The most difficult reading of the three is audendi; and while it is hard to see how it arose out of audentem, it seems easy to trace the easy reading audentem from audendi. A scribe finding it hard to translate audendi here, as does Servius, might write audenti, as we are told was the case, and make it depend on sequi, as though it were for obsequi. But it is not for obsequi here, and so audenti would be corrected to audentem, a reading easy to understand. But if we restore audendi, we must punctuate: si vobis audendi extrema cupido certa, sequi ‘if you have a settled desire to dare the worst, follow’; where sequi is the infinitive for the imperative. I have no recollection of the idiom in Plautus, which goes to show that it had become very rare. But in: hic tamen hanc mecum poteras requiescere noctem (Buc. 1. 79) the easiest explanation is to make requiescere an imperative, and treat poteras as parenthetic. When we compare the oblique in: quae se inhoneste optavit parere hic ditias potius quam in patria honeste pauper viveret (Ter. And. 797-8) with the direct in: potius quam venias in periculum . . . dividuom face (Adelph. 240-1), plainly, just as viveret is the oblique correspondent to venias, so parere corresponds to face, and is an infinitive used imperatively balancing viveret, which is a subjunctive used for an imperative.


But we have in Latin an idiom by no means rare, which when it occurs in Greek, has been always explained as dependent on this use of the infinitive for the imperative. We read: ἆ δειλοί, πόσ’ ἴμεν; τί κακῶν ἱμείρετε τούτων; (Od. 10. 431); where the person of ἱμείρετε forbids our taking ἴμεν as the first plural, and: ὦ βασιλεῦ, κότερον λέγειν πρὸς σὲ τὰ νοέων τυγχάνω, ἢ σιγᾶν ἐν τῷ παρεόντι χρόνῳ (Herod, 1. 88), where λέγειν or ἴμεν seems a repetition of the infinitive by the speaker in reluctant response to a command usually implied. It is much more common in Latin, where in: mene incepto desistere victam (Aen. 1. 37) it is evidently an indignant protest by Juno against the command imposed on her by fate. In her anger she repeats incredulously the command conveyed to her, as though she had not heard aright, just as does Croesus in the example cited from Herodotus; and so the idiom indicates plainly that the command was conveyed by the infinitive, not the imperative. So, too, in: servom antestari? vide (Pl. Curc. 623), tantum laborem capere ob talem filium (Ter. And. 870), quid enim sedere totos dies in villa ista? (Cic. Att. 12. 44. 2). We find the idiom in the second part of:

Mene salis placidi voltum fluctusque quietos

Ignorare iubes? Mene huic confidere monstro? (Aen. 5. 848-9),

where Virgil explains it in the sense I have indicated by supplying iubes with it in the first part. But in the great majority of examples of the idiom it passes to the expression of an exclamation at a shocking misfortune, as in: mene Iliacis occumbere campis non potuisse! (Aen. 1. 97-8) or: huncine solem tam nigrum surrexe mihi! (Sat. 1. 9. 72-3). That the interrogation and not the exclamation is the older force here seems plain when we have this form transferred to indirect, as we find it in: existimabant . . . postremo quid esse levius aut turpius quam auctore hoste de summis rebus capere consilium? (B. G. 5. 28, fin.), or: sententia quae censebat reddenda bona . . . nam aliter qui credituros eos, non vana ab legatis super rebus tantis adferri? (Liv. 2. 4. 3-4). For plainly esse or credituros here has been transferred from such uses of the infinitive indirect as we have just given. Naturally the subjunctive is also used in such cases, as in: docebat etiam . . . id eis eripi quis pati posset? (B. G. 1. 43), or: legati . . . rogaverunt . . . eane meritos crederet quisquam, &c. (Liv. 7. 20. 3-5).

In: poscere fata tempus (Aen. 6. 45-6) have we a complex or a 124 simple sentence? Those who regard Skt. davane, or Greek δοῦναι as the primary and typical form of the infinitive, and treat it as the dative of a noun, ‘for the giving’, will regard it as a simple sentence. But we have also the Skt. kartum, the Latin creatum, apparently an old accusative, ‘for the making’, and this gives what is termed the infinitive proper in both Sanskrit and Umbro-Oscan. We have, thirdly, a form without any case-ending, and showing the infinitive stem only, as in δόμεν, or the Ionic ἰδέεν (probably old ἰδέϝεν), giving us the usual Greek infinitive ἰδέῖν; and this may well be the oldest of the three. It is true that, if we except first supines, Latin infinitives seem all of the first class, the datives. But in Greek active forms like ἀγέμεν or μένειν seem most usual when used as imperatives; in Homer they are in ending the same as the German forms schreiten, fahren; and gerade stehen is closely parallel to αὖθι μένειν.

To return to our question: Is not: melius non tangere (Hor. Sat. 2. 1. 45) a complex sentence, being composed of two: non tangere. Melius erit? And so with: poscere fata. Tempus est. There cannot be much doubt that: me licet unda ferat (Prop. 2. 26. 44) is a complex sentence; what of: me licet undam ferre? Is it not for: me undam ferre. Licet? We compare with: me undam ferre ‘let the wave bear me’ the Homeric: Αἴαντα λαχεῖν ‘let Ajax win’, or Hesiod’s γυμνὸν σπείρειν. When we compare these with: οὺ δ’ ἐυ κρίνασθαι, it is evident that the accusative with the infinitive is not primary, but a later development to be paralleled with that in: volo te venire, which again is complex for: te venire, volo ‘do thou come, it is my wish’. And so: quid vis faciam? (Ter. Eun. 1054) seems complex for: quid faciam? Quid vis? a use of three for four. We have the two constructions confused in: mene vis dem ipse in pedes? (Pl. Capt. 121), showing how lightly the Roman passed from one to the other.

Of course this shows us the real nature of our modal infinitive, if it is true; and to me all seems to point in this direction. Monro (Hom. Gram. 241) thinks we may connect the use of the infinitive for the imperative with the use of the infinitive to imply fitness or obligation. This, pointing to such uses as we have in: ἔστι μὲν εὕδειν (Od. 15. 392), ἀλλὰ καὶ ὥρη εὕδειν (Od. 11. 330-1), οὐ γὰρ ἔοικ’ ὀτρυνέμεν (Il. 4. 286), ἔμε δὲ χρὴ γήραϊ . . . πείθεσθαι (Il. 23. 644-5), modal infinitives all, is plainly parallel with the course we are following. But Monro thinks of explaining the use for the imperative from such constructions 125 by the ellipse of ἔστι or χρή, which is explaining the simple from the complex. He is embarrassed, because he sees this mode of explanation does not help with the German use in Schritt fahren. Is it not more reasonable to try to explain the complex from the simple sentence?

When we turn to Apollonius Dyscolus (Synt. III. 14), we read: ‘I believe that Homer, following his custom of avoiding imperative modes of expression, introduces the infinitive for the imperative, as a general form of the verb, into which, as I have just shown, all special moods can be converted’. In III. 13 he had shown that the infinitive is a general form lying at the base of all moods; so that in περιπατῶ we have ὡρισάμην περιπατεῖν, in περιπατοῖμι, ηὐξάμην περιπατεῖν, in περιπάτει, προσέταξα περιπατεῖν. His ‘general form of the verb’ corresponds closely with our usual definition of the infinitive as the form of the verb which gives its simple meaning, without reference to number or person. This form, called in Latin infinitivum, in Greek ἀπαρέμφατον ‘with no παρεμφάσεις or accessory meanings’, would naturally be, not the case form δόϝεναι, which is a dative, but the simple stem form δόμεν or ἔμμεν or ἰδεῖν; which may approximate very closely in meaning to the root form da, or voca, or fer, that we found to be the old universal forms for command, out of which dato, or vocate, or ferunto developed. Just what additional meaning the stem-forming suffixes -μεν or -ϝεν bring to the root, I do not know; and until we know this, we cannot accurately determine the relation of da or δός to δόμεν; but they seem to have been very much alike.

Brugmann (Gr. Synt. 170) thinks that a satisfactory explanation of the union of πρίν with the infinitive, already fully developed in Homer, has not yet been found. Monro’s (Hom. Gram. 236) difficulty with πρὶν δόμεναι (Il. 1. 98) ‘before the giving’ seems to be his feeling that πρίν should take an ablative, not a dative. But in: πρὶν ἐλθεῖν υἷας Ἀχαιῶν (Il. 9. 403) ἐλθεῖν is neither a dative nor an ablative; Apollonius felt it to be, not a noun, but the general form of the verb, and we know that in Greek, even in Attic Greek, what the Greek grammarian called the πρόθεσις, the Latin grammarian the praepositio, was not a preposition in our sense of the word. Dionysius Thrax finds the πρόθεσις . . . ἔν τε συνθέσει καὶ συντάξει (18); i.e. he feels that σύν is as much a preposition in συντίθημι as in σὺν θεῷ. Far less can we treat πρίν as a preposition in our sense of the word in considering the early union of sentences which joined πρίν with the 126 infinitive. Perhaps δόμεναι is properly regarded as a noun; but is δόμεν a noun or a verb? Certainly ἄγε and fer, which seem its nearest parallels in syntax, can hardly be regarded as nouns. I suppose that roots, as such, belong to a time in the history of language when the distinction between the noun and the verb was not yet developed.

Let us take first the easier construction of πρίν with the negative, as in:

τὴν δ’ ἐγὼ οὐ λύσω πρίν μιν καὶ γῆρας ἔπεισιν

ἡμετέρῳ ἐνὶ οἴκῳ (Il. 1. 29-30),

‘her I will not release; sooner shall old age come upon her in our home’, where the adverbial use of πρίν is evident. So in Il. 18. 283, also Od. 13. 427; 15. 31. We turn to its use with the infinitive in:

οὐδέ κεν ὣς ἔτι θυμὸν ἐμὸν πείσει’ Ἀγαμέμνων,

πρίν γ’ ἀπὸ πᾶσαν ἐμοὶ δόμεναι θυμαλγέα λώβην (Il. 9. 386-7),

‘but not even so shall Agamemnon any longer persuade my mind. Sooner let him pay me back all the bitter despite’, where by taking δόμεναι as equivalent to an imperative in parataxis, we get exactly the sense we want, leading us to ‘till he have paid back’, the usual translation. So in:

οὐδ’ ὅ γε πρὶν Δαναοῖσιν ἀεικέα λοιγὸν ἀπώσει,

πρίν γ’ ἀπὸ πατρὶ φίλῳ δόμεναι ἑλικώπιδα κούρην (Il. 1. 97-8),

‘nor will he before remove the loathly pestilence from the Danaans. Let Agamemnon first give back to her father the bright-eyed maid’. And in:

ὦ Κίρκη, τίς γάρ κεν ἀνήρ, ὃς ἐναίσιμος εἴη,

πρὶν τλαίη πάσσασθαι ἐδητύος ἠδὲ ποτῆτος,

πρὶν λύσασθ’ ἑτάρους καὶ ἐν ὀφθαλμοῖσι ἰδέσθαι; (Od. 10. 383-5),

‘O Circe, what man, who would be righteous, would endure first to take meat and drink? Sooner let him view his companions face to face, and set them free’. In all four examples cited the principal clause has a negative expressed or implied, and has either a future verb, or a verb in the optative used as a future. The indicative used in the dependent clause of the first is clearly a voluntative future, coming very close to the meaning of the imperatives for which infinitives are used in the remaining three.


But for this imperative force we have subjunctives substituted in:

μήτηρ δ’ οὔ με φίλη πρίν γ’ εἴα θωρήσσεσθαι,

πρίν γ’ αὐτὴν ἐλθοῦσαν ἐν ὀφθαλμοῖσι ἴδωμαι (Il. 18. 189-90),

‘and my mother does not permit me to arm myself sooner. Sooner must I see her plainly coming’. Monro sees this parataxis in:

οὐδέ μιν ἀνστήσεις, πρὶν καὶ κακὸν ἄλλο πάθῃσθα (Il. 24. 551),

‘you will not raise him. Sooner shall you suffer some new evil’. He thinks that ‘the subjunctive was directly modelled on the existing use with the infinitive, that πρὶν πάθῃσθα simply took the place of πρὶν παθεῖν, when a more definite conditional force was wanted’. It is quite simple to substitute the subjunctive for the imperative, and for the infinitive used as an imperative; we have here a parallel to the Latin licet venire and licet venias. The ‘simple’ substitution of a subjunctive for an infinitive not so used might call for further explanation. But we can understand the transference of this subjunctive to the optative after a past tense, as in:

οὐκ ἔθελεν φεύγειν, πρὶν πειρήσαιτ’ Ἀχιλῆος (Il. 21. 580).

We find the infinitive used for the imperative with πρίν in both the principal and dependent clauses in:

Ζεῦ κύδιστε, μέγιστε . . .

μὴ πρὶν ἐπ’ ἠέλιον δῦναι καὶ ἐπὶ κνέφας ἐθεῖν,

πρίν με κατὰ πρηνὲς βαλέειν Πριάμοιο μέλαθρον (Il. 2. 412-4).

‘O Zeus most glorious, most great . . . let not the sun go down thereon, nor darkness come on. Sooner let me have cast to ground Priam’s palace’. So, too, in:

μή μοι πρὶν ἰέναι, Πατρόκλεες ἱπποκέλευθε,

νῆας ἔπι γλαφυράς, πρὶν Ἕκτορος ἀνδροφόνοιο

αἱματόεντα χιτῶνα περὶ στήθεσσι δαΐξαι (Il. 16. 839-41),

‘come not sooner to me, Patroclus driver of steeds, to the hollow ships. Sooner shall thou have rent about his breast the gory tunic of man-slaying Hector’. We have, to parallel this, a voluntative future in the principal, and an infinitive in the dependent clause in:

οὐ γὰρ πρὶν πολέμοιο μεδήσομαι αἱματόεντος,

πρίν γ’ υἱὸν Πριάμοιο δαΐφρονος, Ἕκτορα δῖον,

Μυρμιδόνων ἐπί τε κλισίας καὶ νῆας ἱκέσθαι (Il. 9. 650-2),

‘for not sooner will I take thought of bloody war. Sooner let noble Hector, wise Priam’s son, come to the huts and ships of the Myrmidons’. And so, too, in:

οὔ σε πρὶν κτεριῶ, πρίν γ’ Ἕκτορος ἐνθάδ’ ἐνεῖκαι

τεύχεα καὶ κεφαλήν, μεγαθύμου σοῖο φονῆος (Il. 18. 334-5),


‘not sooner will I bury thee. Sooner let me bear hither the arms and head of Hector, thy high-hearted slayer’. I might also translate ‘let me bury’; the parallel shows how closely equivalent to the infinitives cited above are the voluntative futures I have just quoted.

But the future proper and the voluntative future find expression in Greek by the same tense, and by the same forms. When such is the case, it need not surprise us that the infinitive, which is equivalent to a voluntative future, should likewise assume the meaning of the future proper. In:

μὴ πρὶν παύειν χεῖρας ὁμοιΐου πολέμοιο,

πρὶν κατὰ Ἰλιόφι κλυτὰ τείχεα λαὸν ἐέλσαι

Τρωϊκόν, ὅς κε φύγῃσι (Il. 21. 294-6),

‘hold not thy hand from hazardous battle. Sooner shall thou have pent the fleeing Trojan host within Ilion’s renowned walls’, we have still the imperative force in both clauses. But in:

οὐδέ τις ἔτλη

πρὶν πιέειν, πρὶν λεῖψαι ὑπερμενέϊ Κρονίωνι (Il. 7. 480-1),

‘nor did any endure sooner to drink. Sooner must he make libation to sovereign Zeus’, we have in πιέειν a future proper, but in λεῖψαι an imperative. In:

οὐδέ τι γυῖα

πρὶν κάμνει, πρὶν πάντας ἐρωῆσαι πολέμοιο (Il. 19. 169-70),

‘nor does a man at all feel weary in limb, before all give back from the battle’, we have a gnomic present in the principal, and a future in the dependent clause, though in that future the older voluntative sense seems not wholly lost. In:

πρὶν γάρ τοι πολύμητις ἐλεύσεται ἐνθάδ’ Ὀδυσσεύς,

πρὶν τούτους τόδε τόξον ἐΰξοον ἀμφαφόωντας

νευρήν τ’ ἐντανύσαι διοϊστεῦσαί τε σιδήρου (Od. 19. 585-7),

‘for, lo! the many-counselled Ulysses will come hither, before these men, for all their handling of the polished bow, will have strung it and shot the arrow through the iron’, where we note that the principal clause is not negative; we have a future proper in the principal clause, with two infinitives in the dependent no longer voluntatives, but futures proper. And so in:

στῆτ’ αὐτοῦ, καὶ λαὸν ἐρυκάκετε πρὸ πυλάων

πάντῃ ἐποιχόμενοι, πρὶν αὖτ’ ἐν χερσὶ γυναικῶν

φεύγοντας πεσέειν, δηΐοισι δὲ χάρμα γενέσθαι (Il. 6. 80-2),

‘stand there, and ranging the host every whither rally them before the gates, ere yet they fall fleeing into their women’s arms, and become a joy to the foe’, we pass to a use of πρίν with a future that implies prevention. 129 In:

οὐδ’ ὅσα φασὶν

Ἴλιον ἐκτῆσθαι, εὖ ναιόμενον πτολίεθρον,

τὸ πρὶν ἐπ’ εἰρήνης, πρὶν ἐλθεῖν υἷας Ἀχαιῶν (Il. 9. 401-3),

‘what wealth men say the well-peopled city of Ilios was possessed of in yore in peaceful days, before the sons of the Achaeans came’, where the future is related to a state of things now past, and itself designates an event now past, though in the future at the time marked by the verb in its principal clause; of course the future sense is obscured somewhat for the hasty reader. In:

ἀλλ’ ἕπευ· οὐ γὰρ ἔτ’ ἔστιν ἀποσταδὸν Ἀργείοισι

μάρνασθαι, πρίν γ’ ἠὲ κατακτάμεν ἠὲ κατ’ ἄκρης

Ἴλιον αἰπεινὴν ἑλέειν κτάσθαι τε πολίτας (Il. 15. 556-8),

‘but follow close; for no longer may we wage war with the Argives in guise aloof; sooner must we either slay them, or they will capture lofty Ilios from its summit down and slay the citizens’, where we have an imperative joined with a future by ἠέ . . . ἠέ. In:

ἀτὰρ οὐ μὲν σφῶΐ γ’ ὀΐω

πρίν γ’ ἀποπαύσεσθαι, πρίν γ’ ἢ ἕτερόν γε πεσόντα

αἵματος ἆσαι Ἄρηα (Il. 5. 287-9),

‘but I deem ye twain will not sooner cease. Sooner else must one or other have fallen and glutted Ares with his blood’, we have a future joined with an imperative. In:

μηδὲ πρὶν ἀπόπαυε τεὸν μένος, ἀλλ’ ὁπότ’ ἂν δὴ

φθέγξομ’ ἐγὼν ἰάχουσα, τότε σχεῖν ἀκάματον πῦρ (Il. 21. 340-1),

‘nor sooner stay thy rage. Only whensoever I shall cry aloud, then check the unwearying flame’, we have πρίν with an imperative coordinated with τότε and an infinitive, a construction that should make it plain that such infinitives are used as imperatives. In:

ἀλλ’ ὄμοσον μὴ μητρὶ φίλῃ τάδε μυθήσασθαι

πρίν γ’ ὅτ’ ἂν ἑνδεκάτη τε δυωδεκάτη τε γένηται (Od. 2. 373-4),

‘but swear not to tell those things sooner to my mother. Whenever it will be the eleventh or twelfth morning (then shalt thou tell her)’, and in:

ἦ τοι ἔφην γε

οὐ πρὶν μηνιθμὸν καταπαυσέμεν, ἀλλ’ ὁπότ’ ἂν δὴ

νῆας ἐμὰς ἀφίκηται ἀΰτή τε πτόλεμός τε (Il. 16. 61-3),

‘yet in truth I said I would not sooner cease from my wrath. But whenever to my own ships would come the war-cry and the battle (then I would cease)’, evidently the clauses parallel to τότε ἔχειν, as 130 involving repetitions of verbs already expressed, have been omitted, and must be supplied in determining the value of πρὶν μυθήσασθαι or πρὶν καταπαυσέμεν.

I read in Sallust: nam idem velle atque idem nolle, ea demum firma amicitia est (Cat. 20. 4). But in Martial’s well-known epigram beginning: vitam quae faciant beatiorem . . . haec sunt (10. 47) his list, starting with res . . . relicta, non ingratus ager, &c., ends with:

Quod sis esse velis, nihilque malis;

Summum nec metuas diem nec optes,

where from the example in Sallust one would expect infinitives, not subjunctives. But when I recall the equivalence of licet ire and licet eas, the use of the subjunctive for the infinitive used as an imperative is clear at once. So in:

Omnibus idem animus, scelerata excedere terra,

Linqui pollutum hospitium et dare classibus Austros (Aen. 3. 60-1),

the explanation of excedere, linqui, dare, as imperatives that define animus is much clearer to me, than their definition as explicative infinitives setting forth that purpose.

With this use of the infinitive in view I understand its use with iubeo and impero. With iubeo in poetry I have (1) the use with the dative of the person bidden and the accusative of the command in: non haec miserae sperare iubebas (Catull. 64. 140), (2) the accusative with the infinitive parallel to Αἴαντα λαχεῖν in: haec ubi nos praecepta iubent deponere dona (Aen. 6. 632), (3) with the subjunctive equivalent to the imperative in: magna dicione iubeto Carthago premat Ausoniam (Aen. 10. 53-4), (4) with the subjunctive with ut in: tu deinde iubeto ut certet Amyntas (Buc. 5. 15) ‘Thy bidding shall be: Would that Amyntas may compete’. Of these four constructions prose has adopted the second as the easiest and plainest. With impero in poetry we have four constructions closely parallel: (1) flectere iter sociis imperat (Aen. 7. 35-6), (2) tolli miserabile corpus imperat (Aen. 11. 59-60), (3) letoque det imperat Argum (Ov. Met. 1. 670), (4) Apollo mihi ex oraclo imperat ut ego illic oculos exuram (Pl. Men. 840-1). The use of the mere subjunctive with impero seems late and rare, as imperator was not primarily the commander-in-chief, but rather the quartermaster (indu-parare), he who gets into the camp the provisions for the soldiers. Here, too, prose has selected the syntax best fitting his dignity, the subjunctive with ut, the form adapted to request rather than command.



Postgate (I.-G.F. 4. 255) has shown that the omission of se in a construction like: summasque minatur deiecturum arces (Aen. 12. 654-5) is probably due to analogy with: altero (gladio) te occisurum ait (Pl. Cas. 692-3) ‘with one sword he says he will kill thee’. Here occisurum is not yet the future participle, but is the old invariable future infinitive occisu-esom ‘to be for the slaying’, out of which the future participle was later developed. Probably this future infinitive was originally constructed with verbs like spero and promitto in phrases such as: spero venturum ‘I hope to come’ and: promittebat daturum ‘he promised to give’, where daturum would be exactly parallel to ποιήσειν in ἔφη ποιήσειν. Nether spero nor promitto was originally a verb of saying; but polliceor is, being a compound of liceor ‘I bid at auction’. Through it would arise a syntactic syncretism between spero and promitto on the one hand, and aio and dico on the other; we should have from this: aio te occisurum, following: aio te errare; and with spero, following aio te victum esse, spero hostium copias ibi occupatas futurum (cf. Quadrig. apud Gell. 1. 7. 9). Indeed with spero, when the subject of the infinitive was other than that of the principal verb, that subject was expressed, as we see in: est quod speremus, deos bonis bene facturum (Quadrig. ap. Gell. 1. 7. 9). From a syncretism of this syntax with: dico illos profectos esse we should get: dico illos profecturos esse, and: spero deos bonis bene facturos esse, the regular classical constructions. But we get also, according to Gellius, dixerunt omnia ex sententia processurum esse (Val. Ant. ap. Gell. 1. 7. 10), where we have the influence of the older form swaying the resultant form to processurum. Indeed Gellius read in Cicero: fiducia . . . quocumque venerint, hanc sibi rem praesidio sperant futurum (Verr. 2. 5. 167. 65), where all our editors read futuram. Gellius supports his reading by a citation from C. Gracchus: credo ego inimicos meos hoc dicturum, and from Laberius: non putavi hoc eam facturum. That his authority is worth more than the testimony of our manuscripts, I think there can be no doubt.


In like fashion from a syncretism of: aio facturum with: aio illum profectum esse would issue aio profectum, cf. testor . . . vitavisse (Aen. 2. 432-3), and nos abiisse rati (Aen. 2. 25); and from aio (me) profectum esse, formed after aio (illum) profectum esse, would result by attraction: aio profectus esse, as in: ait fuisse navium celerrimus (Catull. 4. 2), or: uxor invicti Iovis esse nescis (Od. 3, 27. 73), or: vir bonus et sapiens dignis ait esse paratus (Ep. 1. 7. 22). Following ait fore bonus, where fore is like esse a present infinitive in form, we should have ait venturum transformed to ait venire and so we should have sperat venire, or: hoc sperem Italiam contingere caelo (Aen. 5. 18), or: operam dare promittitis (Pl. Trin. 5).

At first sight: aio profectus esse does not look like Latin; but when we compare: quas hodie adolescens Diabolus ipsi daturus dixit (Pl. Asin. 634), visura et quamvis numquam speraret Ulixen (Prop. 2. 9. 7), non aliter caeco nocturni turbine Cori scit peritura ratis (Stat. Theb. 7. 791-2), quoad summos illi promitterent honores habituri mihi (Apul. Met. 7. 14), we see examples of the syntax in all periods of older Latinity. If Latin allows of dicturus dixit as well as se dicturum dixit, we may expect to find corresponding irregularities in oblique. And I note:

quam sedem somnia volgo

Vana tenere ferunt, foliisque sub omnibus haerent (Aen. 6. 283-4),

where we have the figure of three for four, i.e. tenere-ferunt, haerent, for tenere-ferunt-haerere-ferunt; and I find in Tacitus: validissimos equitum incurrere latus, Stertinium cum ceteris turmis circumgreditergaque invadere iubet; ipse in tempore adfuturus (Ann. 2. 17. 1) and: at si auxilia et socii adversum abscedentes legiones armarentur, civile bellum suscipi: periculosa severitas, flagitiosa largitio; seu nihil militi, seu omnia concedentur, in ancipiti res publica (Ann. 1. 36. 2-3).

Plautus’s daturus dixit seems to have been extended by analogy to the perfect infinitive at times, as in: pulsata indignis saepe queror manibus (Prop. 1. 16. 6), sensit medios delapsus in hostes (Aen. 2. 377), gaudent perfusi sanguine fratrum (Geo. 2. 510), quo nunc Turnus ovat spolio gaudetque potitus (Aen. 10. 500). Probably the analogy of these led Virgil from: intentique ora tenebant (Aen. 2. 1) to: defixique ora tenebant (Aen. 8. 520); or from: infert se saeptus nubila (Aen. 1. 439), or: sese tulit obvia (Aen. 1. 314), to: tectusque tenet se (Aen. 10. 802).

In: quia, qualiacumque leguntur ista, salutator scribere non potuit (Mart. 1. 70. 17-18) salutator is evidently short for: si salutaturus usque 133 adesset. This use of the nomen agentis for the future participle seems pro-ethnic. Mr. Postgate (Class. Rev. 5. 301 and I.-G.F. 4. 252) has given reasons, as we have seen, for thinking venturus a late development from the invariable infinitive venturum. To the question: How did the idea expressed by the future participle thus evolved find expression before it was developed? the answer seems plain: By the nomen agentis. We find it so used in prose as well, as in: ipse Hannibal, qua turris mobilis . . . agebatur, hortator aderat (Liv. 21. 11. 7), and it occurs repeatedly in Tacitus, as in: rector iuveni et ceteris periculorum praemiorumque ostentator (Ann. 1. 24. 3), neque legatus aut tribunus moderator adfuit (Ann. 1. 49. 3). Turning to Sanskrit we find the cognate nomen agentis used with or without the verb ‘to be’ as a periphrastic future: data (= Latin dator) is ‘he will give’, and dataras (= Latin datores) ‘they will give’, but datāsmi (= Latin dator sum) is ‘I shall give’ and datāsthas (= datores estis) ‘you will give’ (Whitney, Skt. Gram. 942-4). The obvious solution of this coincidence in syntax seems to be the reference of both uses to a pro-ethnic use of dator in the sense of the later daturus.





When we turn from Caesar or Cicero to Virgil or Horace, we feel at once the marked difference between the diction of prose and that of poetry. At first we are inclined to regard prose as much the simpler and less ornate; for we are apt to forget that ornatus is a later form of ordinatus. We do best here to begin by ascertaining and comparing the primary meanings of the words ‘prose’ and ‘verse’.

‘Prosa’ is an assimilated feminine of prorsus, short for provorsus ‘straightforward’. A similar assimilation gives us pessum for persum and rusus for rursus. Versus, older vorsus, is the past participle of vorto ‘I turn’, which has become a noun of the fourth declension, as have adventus, auditus, conventus, gestus, gressus, haustus, and many others. So numerous are the past participles thus converted, that, though the older nouns of the fourth are mostly feminine or neuter, our grammarians speak of nouns of the fourth in -us as masculine, and treat the feminines as exceptions. But a comparison with Greek makes it easier to see why nouns like nurus and socrus should end in -us. These past participles seem first to have appeared among nouns as neuters of the second declension. Servius (ad Aen. 10. 689) notes of monitis, that while Persius (1. 79) used monitus as the acc. pl., we have not a dative either singular or plural of the fourth. And we find words like senatus and tumultus partly of the second, partly of the fourth; i.e. the transference to the fourth is not yet complete. The word ‘versus’ indicates that the sentence does not go straight on as in prose, but, after completing a definite series of long and short syllables, turns back to repeat itself, or to complete another of a definite series. Often the closing syllables of the first verse have their sound reflected in that of a following verse; and this we call ‘rhyme’. This effect, characteristic of modern verse, is almost as common in the prose as in the verse of the Greeks and Romans. As an example of rhyme in Latin prose I might cite Augustine’s comment on the words of Peter: ‘bonum est nos hic esse’; in monte requiescere cupiebas; descende laborare, praedica veritatem, habe caritatem; et sic pervenies ad aeternitatem, ubi invenies securitatem.


Even more significant for our purpose of distinction than the root-words in prosa and versus are the endings -a and -us. The words are adjectives primarily, and the nouns to be supplied with them are, for prosa, oratio, and for versus, sermo. Sermo is the talk we hold with our fellows in the parlour or on the street, easy and unconstrained, without careful choice of words or constructions. Oratio is the style of speech we strive to attain when we ascend the rostra to instruct or entertain a select audience. ‘The applause of listening Senates to command’, you must have at your command a style of speech at once clear, dignified, and elegant; every word should at once convey to your hearers its proper meaning; and all your words must be chosen and connected in a way that will enhance the respect of your hearers for you and for themselves. To master such a style is the most difficult, and perhaps the highest, of literary achievements; and any degree of excellence in it comes later among Greeks and Romans than the development of poetic diction. Homer composed his verse so long before Herodotus, the father of prose, that the latter found it difficult to determine his date; and Fabius Pictor, who wrote the oldest history of Rome, finding no prose style developed in Latin, wrote it in Greek, though in his day Plautus was already delighting the Romans with his comic verse and Naevius thrilling them with his Saturnians.

Of all peoples we know, the Romans were the most successful in prose; their language was more clear and direct, if less flexible, than Greek; and at an early date they introduced into their prose certain conventional rules, some of them artificial, which tended to enhance its clearness and dignity. These usually took the shape of distinctions and prohibitions. ‘You must use a preposition with the ablative of place, but not with that of time’ is a rule very conducive to clearness of meaning. ‘You must not use the infinitive to express purpose’ is a rule that every Roman orator and writer strove to observe in oratio, while he disregarded it in sermo and versus. Indeed, in the oldest examples of this use of the infinitive, as in: volo venire, or: statim redire constituit, the orator was unmindful of this rule; for in them the sense of purpose had been obscured, and the relation had become modal.

In prose you must use the infinitive with its subject expressed after verbs of hoping or promising; while Virgil writes: hoc sperem Italiam contingere caelo (Aen. 5. 18) or: cur mea dicta neget duras demittere in aures? (4. 428). With verbs of perceiving and of telling 139 in prose you will use the infinitive, and not ut with the subjunctive; and nothing will tend to deface your Latin composition so much as a tendency to write: dico ut erres, or: video ut venias. But we have Horace writing: vides ut alta stet nive candidum Soracte (Od. 1. 9. 1-2) and Pliny: dixi utrique parti ut postulationum suarum libellos darent (Ep. 10. 81. 5). True, these are not the idioms forbidden, though they are related, and the avoidance of them in good prose is too invariable not to be the result of purposed choice. As we have already shown with iubeo and impero, of a number of constructions admitted freely in poetry one was selected for prose, evidently to promote clearness of expression; and the rest were rejected, among them at times the old and genuine mode of expressing the idea. For ‘in good time’ Plautus uses temperi, tempore, in tempore, ad tempus, and suo tempore, of which the old locative temperi seems the genuine mode; Terence has only tempore, in tempore, and ad tempus. That this selection was largely the work of a learned and cultured coterie, the circle of the Scipios and the Aemilii, seems probable. Plautus is the favourite of the common people, Terence the protégé of the circle of the Scipios; and of all archaic Latin writers he stands nearest to Cicero in style.

Of all the self-imposed conventions that mark off Latin prose style from that of Latin verse, that forbidding the use of the infinitive to indicate purpose or result seems most striking. Horace will write: omne cum Proteus pecus egit altos visere montes (Od. 1. 2. 7-8), or: metus tradam protervis in mare Creticum portare ventis (Od. 1. 26. 1-3), or: niveus videri (Od. 4. 2. 59), or: durus componere versus (Sat. 1. 4. 8); and this use of the infinitive to mark the purpose of an act or the result of a condition seems proper to its primary form. It was evidently just as common in Latin sermo as it is in Greek prose; and its banishment from Latin prose is one of the strangest tours de force in literary syntax. In poetry we have the opposite, e.g. the subjunctive with sino, that in prose requires an infinitive, in: vacua sine regnet in aula (Geo. 4. 90) and: sine vivat ineptus (Ep. 1. 17. 32).

Poetic style has its conventions too, though they commonly lead to the disregard of any fixed rule. It feels the majesty of what is ancient in phrasing as well as in subject-matter, and so tends to favour archaic diction. The passive voice is a late development in language; and it has adopted as its past participle a form of adjective that was formed as freely from noun stems as from verb stems, and 140 was once active as well as passive in meaning. We have alatus ‘provided with wings’, as well as amatus ‘affected with love’. In poetry we find this participle used at times in an active sense, like the Skt. gatas ‘having come’, in: vestigia . . . titubata solo (Aen. 5, 331-2), cessatum ducere somnum (Ep. 1. 2. 31), quo sanguine cretus (Aen. 2. 74). That such participles as the Skt. gatas were once common in Latin, seems clear from such substantives as adventus and conventus, which were once past participles active. We find in poetry ignarus active in: non ignara mali (Aen. 1. 630) but passive in: ignarum Laurens habet ora Mimanta (10. 706); and caecus is active in auri caecus amore (1. 349), but passive in saxis . . . caecis (3. 706). So we have pendĕre for pendēre in Lucretius 1. 361 and trahere for trahi (id. 1. 397).

In Plautus we find the subjunctive present frequently used for the unreal present, and the imperfect subjunctive for the past unreal. Terence writes: tu si hic sis, aliter sentias (And. 310) ‘if you were I, you would think differently’; and Ennius: nam si curent, bene bonis sit, male malis, quod nunc abest (Fab. 210, M.); and even Cicero: eius igitur mortis sedetis ultores, cuius vitam si putetis per vos restitui posse, nolitis (Mil. 79. 29). So Plautus has the imperfect: utinam te di prius perderent, quam periisti ex patria tua (Capt. 537), where classical Latin uses the pluperfect. In this archaic Latin agrees with Homeric Greek; for siem, later sim, is the Greek optative εἴην, and Homer uses the optative for the prodosis and apodosis of the unreal, where later Greek uses the imperfect or aorist indicative with εἰ and ἄν. We read in Virgil:

Continuoque ineant pugnas et proelia temptent,

Ni roseus fessos iam gurgite Phoebus Hibero

Tingat equos noctemque die labente reducat (Aen. 11. 912-14),

and with the imperfect in:

tu quoque magnam

Partem opere in tanto, sineret dolor, Icare, haberes (Aen. 6. 30-1).

Doubtless this use of the present and of the imperfect for the imperfect and pluperfect subjunctive in expressing the unreal, does, as Madvig tells us, give a vivid force to the diction, but in producing that effect Virgil is not devising a new syntax, but resorting to a construction characteristic of archaic Latin, a construction not yet quite obsolete even in the prose of his day. It may seem strange that this old use of the pres. subj. for the unreal present is never joined in syntax with the impf. subj. used later in the same sense, especially 141 in poetry where variety is aimed at. And we find them so joined in: nam si primordia rerum commutari aliqua possent ratione revicta, incertum quoque iam constet, &c. (Lucr. 1. 592-4); but editors change possent of the manuscripts to possint. Again in: nisi terminet alterum eorum, simplice natura pateat . . . nec mare nec tellus . . . possent horai sistere tempus (1. 1012-16) we have them co-ordinated in the apodosis; but editors assume a loss of two verses between v. 1013 and v. 1014. Cf. also: est combustus . . . volarit (Prop. 2. 30. 29-30).

Andrew Lang tells us how Sir Walter Scott, in writing down the Ballad of Otterbourne from Mrs. Hogg’s dictation, ventured to change a word to save the grammar. The old lady dictated thus the words of the dying Douglas:

For I have dreamed a dreary dream

Ayont the isle of Skye;

I saw a dead man won a fight,

And methinks that man was I.

But Scott for ‘won’ wrote ‘win’. From a comparison of Latin with Greek oblique it becomes plain that primarily direct narration was transferred to oblique with the smallest possible change. So here the Douglas in expressing in sub-oblique what he saw in his dream, says ‘a dead man won a fight’. The old lady was using antique syntax; and was justified in her remark to Sir Walter that in changing ‘won’ to ‘win’ he had spoiled her ballad. So in Latin the indirect question, when it was not a dubitative or a rhetorical, but a real question, primarily put the verb in the indicative, just as it is in direct. So Virgil:

Vidisti quo Turnus equo, quibus ibat in armis

Aureus (Aen. 9. 269-70),

and again in:

subigitque fateri

Quae quis apud superos, furto laetatus inani,

Distulit in seram commissa piacula mortem (Aen. 6. 567-9),

and in: viden ut geminae stant vertice cristae? (ib. 6. 779), and in a passage which has occasioned difficulty:

ne quaere doceri

Quam poenam, aut quae forma viros fortunave mersit (Aen. 6. 614-5).

Far more common in Horace and Virgil than the archaic construction just mentioned is that commonly called the Accusativus Graecus, though it is not a Latin imitation of a Greek idiom, but found also in Umbro-Oscan. We have good examples of it in: inscripti nomina regum . . . flores (Buc. 3. 106-7) ‘flowers that have 142 the names of kings inscribed on them’, perque pedes traiectus lora tumentes (Aen. 2. 273) ‘having thongs passed through his swollen feet’, laevo suspensi loculos tabulamque lacerto (Sat. 1. 6. 74). This use of an accusative governed by a passive participle, as in: caeca mentem caligine Theseus consitus (Catull. 64. 207-8), or by a finite form of a passive verb, as in: Androgei galeam . . . induitur (Aen. 2. 392-3), while common in classical poetry, is not found in Cicero or Caesar, Quintilian or Pliny. It appears in Livy, and is common in the later prose of Seneca and Tacitus, which is more and more subject to the charm of Virgil. In its old Latin uses it is confined to the senses of clothing or piercing, but its use is extended later, as in: miles . . . multo iam fractus membra labore (Sat. 1. 1. 5), or: aram posuit casus suos in marmore expressam (Tac. Hist. 3. 74. 2). Owing probably to the close relation between the present and past participles of deponent verbs we have this idiom extended to the present participle, as in: flaventem prima lanugine malas . . . Clytium (Aen. 10. 324-5), or: deum cingentem viridi tempora pampino (Od. 3. 25. 19-20).

But while classical poetry as a rule inclines to archaic idiom, often when it turns to represent the chat or sport of the day, we find in it the newest fashions of Roman speech. Two friends meet on the street; unde et quo? (Sat. 2. 4. 1) is their mutual greeting. One of them asks a favour: si me amas (Sat. 1. 9. 38) or: amabo (Catull. 32. 1) is their form of request, and: amo te (Ter. Phorm. 54) or: bene facis (Ad. 601) their form of thanks. One asks after the health or welfare of a passing friend: quid agis, dulcissime rerum? (Sat. 1. 9. 4) or of an absent acquaintance: quid mihi Celsus agit? (Ep. 1. 3. 15). In taking leave: num quid vis (aliud)? (Sat. 1. 9. 6) ‘can I be of (further) service?’ paves the way for: vale ‘good-bye’. The vivid shouts of the competitors in Virgil’s boat-race: litus ama . . . pete saxa, and such phrases as: qui subiit palmae (Aen. 5. 346) ‘who came next to the winner’, or: locum tendunt superare priorem (5. 155) ‘they strive to win the lead’, or: radit iter laevum interior (5. 170) ‘he shaves an inner course on the left’, have little of the archaic about them, and are instinct with all the vim and vigour of to-day’s ball game. Curiously modern, too, is: extremam . . . imponit . . . manum (7. 572-3) ‘she gives it the finishing touch’.

But the scene to which the poet transports us is often very different from this; he tells us of the deeds and sufferings of heroes of ancient days; and strange and vague is the light through which they move. This vagueness of effect is partly attained by the omission of prepositions, 143 a class of words developed to give clearness to the relations of ideas. Dionysius tells us that the preposition is used either in composition or in syntax; modern grammarians know it only in syntax. Here it is often omitted in poetry where it would be expressed in prose. In: devenere locos laetos (Aen. 6. 638) it is omitted with the accusative of goal, as though it were the name of a town, as it is with the ablative of starting-point in: caelo venere volantes (Aen. 6. 191). So with the ablative of place in: caeloque Ereboque potentem (6. 247), or: viridi sedere solo (6. 192). In: scalis habito tribus sed altis (Mart. 1. 117. 7) we have the ablative of interval, ‘three flights up’, used for the local ablative: tertio in tabulato ‘on the third floor’. In return we find the preposition used with the ablative of cause in: mollibus in pueris aut in puellis urere (Epod. 11. 4) or: qui stupet in titulis (Sat. 1. 6. 17); or with the ablative of manner in: horridus in iaculis (Aen. 5. 37) or: in taetro tabescat odore (Lucr. 3. 581) or: corbis in imposito pondere messor eram (Prop. 4. 2. 28). The omission of the preposition used in composition, which is still more promotive of this vagueness of effect, as in: hominem paulatim cernimus ire (Lucret. 3. 526), where ire is for perire, like ‘pass’ for ‘pass away’, I shall have to reserve for a later chapter.

Different seems the reason for the use of ire for venire that we see in: nec vero Alciden me sum laetatus euntem accepisse lacu (Aen. 6. 392-3), and in: vos celsis mine primum a navibus itis? (2. 375). Here we have the use of the general term for the special: ire is the verb denoting motion in any direction used for venire denoting motion in our direction. We have the use in Plautus: ere, unde is? (Cist. 776), and in Terence: sed eccum Syrum ire video (Ad. 361). In Greek βαίνω, the cognate of venio, seems to have taken on this general meaning; though in: Ἥρη, τίπτε βέβηκας; (Il. 15. 90) the primary force of venire is clearly visible. While in Plautus revenire is the usual verb for ‘to return’, it is found only twice in Cicero and never in Virgil. Redire, the corresponding compound of ire, has taken its place. This seems to be due to the opposition we see in: itque reditque viam totiens (Aen. 6. 122). It is plain enough that such a use of the general for the special will promote the vagueness of effect aimed at by the poet; and we find uses of it in: usque ad aquam (Buc. 9. 9) for usque ad fluvium, aurave distinctos educit verna colores (Catull. 64. 90), where colores is for flores, magno discordes aethere venti (Aen. 10. 356), where 144 aether seems the larger sphere of which aer is a part, possit parvos educere natos (Aen. 8. 413), where educere ‘to rear’ is used for educare.

But far more interesting in Latin poetry seems the opposite use of the special for the general. This is the use that at once occurs to Servius, when he meets itis for venitis in Aen. 2. 375; he cites from Terence: nisi eo ad mercatum venio (Ad. 231), where venio is used for eo. So we read in Tacitus: eo furoris venere (Ann. 1. 18. 2) ‘so far did they go in their madness’. In: venit medio vi pontus (Aen. 3. 417) the prose form would be: in medium vi magna pontus iniit. Very often the name of a special tribe is substituted in poetry for the name of the whole nation, as in: non ego sanius bacchabor Edonis (Od. 2. 7. 26-7), where Edonis is for Thracibus. We have cornua the special for arma the general term in: irasci in cornua temptat (Aen. 12. 104) ‘he stirs his wrath for the fight to come’; for arma is a usual metonymy for bellum, which we find in the first verse of the Aeneid. To: pelle Libystidis ursae (Aen. 5. 37) Servius’s note is: aut re vera; aut ferae Africanae, id est, leonis aut pardi. We understand Servius’s doubt, when we recall that Pliny said there were no bears in Africa; his feeling is that ursa here is not used in its proper sense, but in a general sense for fera. So in: gaude, Crasse, nigras si quid sapis inter harenas (Prop. 4. 6. 83) the ‘sands’ of the Euphrates are sands only in this poetical sense; for its banks are alluvial. They are qualified as black partly for this reason, partly to suggest to the reader the fusca regna to which Crassus now belongs. We have here, then, a use of the special for the general, of harena for solum.

As we shall see, Servius believes that for detexit Virgil uses its opposite texit (Aen. 10. 424), just as all Latin poets for depopulari use populari; of course it will be easy to use for ‘land’ a word that usually means ‘water’. We have already in Catullus: sive quae septemgeminus colorat aequora Nilus (11. 7-8), where aequora is ambiguous; it may refer to the waters of the Mediterranean, but is more probably the black flats of Egypt. Interesting is Virgil’s use of aequora in:

Quos patre Benaco velatus harundine glauca

Mincius infesta ducebat in aequora pinu (Aen. 10. 205-6).

At first sight Mincius seems to be bearing the Mantuans on its waters to the sea (ad aequora). But their destination is not the sea, but the field of battle, where they are to meet Turnus and Mezentius; and 145 Mackail rightly translates: ‘led to battle in his advancing pine’. Only so can we understand the force of infesta pinu, modelled on infesta hasta, and presenting the ship as a dart hurled to the battlefield. And we read in the same book: sic toto Aeneas desaevit in aequore victor (v. 569), and in v. 214 Virgil calls the waters of the Tiber down which they sail: campos salis, an expression that involves the use of the special term salis for the general aquae. But in the use of aequor for land as well as water and the use of solum for mare in: vastis tremit ictibus aerea puppis, subtrahiturque solum (Aen. 5. 199) we have the beginnings of oxymoron, a figure to be considered later.

The words in most common use in prose are words which once had a substantive force of their own, a force now almost entirely lost, so that now they merely serve to indicate the relations of the objects in question. As types of such words we may select the preposition de, the conjunction et, the pronoun is, and the verb est. They are, so to speak, the windows in a prose sentence, letting in light on the relations of its parts; and words like is and est have so far lost their old substantival force that they may be likened to colourless and transparent windows through which the meaning of the sentence becomes clear in the light of common sense. We feel the need of such words when we aim at presenting our ideas clearly and definitely, which should be the especial aim of good prose. But poetry is not content with this; in it the relations of the objects in question are presented in a light coloured by the fancy of the bard; and for this there must be found to take the place of such colourless terms as ‘is’ or est words richer in representative force and comparing with them as do the richly coloured panes of some Gothic cathedral with the clear windows of our private dwellings. Or else, what also happens, these words must be coloured with light reflected from more picturesque words associated with them in poetic diction, so that they assume a colour and meaning they do not possess in prose.

We have already spoken of the frequent omission of prepositions in poetry like de or in or cum. This omission is most usual when the preposition is used in a later or derived sense, as in: unus de his hominibus, for the older: unus horum hominum, or when de is used with the sense of about, as in: de ceteris senatui curae fore. In the primary sense of ‘down from’ in space, or ‘proceeding from’ as source or cause de is in frequent use in poetry.

Conjunctions are not so readily omitted, but que and atque are in common use for et, as in: Tereaque Harpalycumque et Demophoonta 146 Chrominque (Aen. 11. 675), or: Imbrasidas, Glaucon atque Laden (12. 343), or: matres atque viri defunctaque corpora vita magnanimum heroum (6. 306-7). Que, the oldest copulative, is the most common, as we see it in: Chloreaque Sybarimque Daretaque Thersilochumque (12. 363); no doubt its short quantity makes it most convenient in a dactylic verse. But we have the usual modern form of omission appearing in: silvas armenta virosque (12. 688). Of course the omission of any conjunction is the oldest form; and we have this often, as in: das nummos, accipis uvam, pullos, ova, cadum temeti (Hor. Ep. 2. 2. 162). When et is used, it is often for aut, as in: aut corpora saltu subiiciunt in equos et strictis ensibus adsunt (Aen. 12. 288), or for nec, as in: nec prolem Ausoniam et Lavinia respicit arva (4. 236), or for atque (our ‘as’), as in: haec amem necesse est et Verianolum meum et Fabullum (Catull. 12. 17), or for cum in: et terram Hesperiam venies (Aen. 2. 781), or for nam in: et saeva Iovis sic numina pellunt (11. 901), or for quamquam in: et dona ferentes (2. 49). Strange seems the use of seu . . . seu for utrum . . . an in: dubii seu vivere credant sive extrema pati (Aen. 1. 218-19), a use followed by Livy in: haud dubius rex seu patrum seu plebis animos periclitaretur (1. 42. 3) and by Tacitus in: iuxta periculoso ficta seu vera promeret (Ann. 1. 6. 6). So we have ut for quanto in: ut melius (Od. 1. 11. 3) and for quam in: non Liber aeque (quatit), non acuta sic geminant Corybantes aera, tristes ut irae (Od. 1. 16. 9) and in: non secus in iugis . . . ut mihi devio (Od. 3. 25. 12).

The use of the pronoun ‘is’ in its usual prose sense is rare in poetry. In: cernere ne quis eos (Aen. 1. 413) and: ab ea (7. 63) it is emphasized by its position just before the caesura; and in: quae mihi reddat eum vel eo me solvat amantem (Aen. 4. 479) the eum has this emphatic position and affects the neighbouring eo. Elsewhere we have either the emphatic isque, as in: idque audire sat est (Aen. 2. 103), or it is used for talis, as in: ea frena furenti concutit (6. 100), or it stands for a genitive, as in: ea cura (3. 505), for cura eius rei perficiendae.

So for esse the poets often use verbs that denote the various forms of existence appropriate to the context; the subject does not merely exist, he is standing, or sitting, or lying, or coming, or going, or staying, or acting, or turning, or dwelling, or told of, or living, or growing. Or, if est is used, it is often for fit, or for stat, or for manet, or for potest, or for habitat, or for liceat, or with the dative for habet, or in the perfect for periit, as in: fuimus Troes, fuit Ilium 147 (Aen. 2. 325), or it is rendered emphatic by its position, as in: est omnia quando iste animus supra (11. 509).

In Greek the verb ‘to be’ is formed wholly from the root es-, and shows only an aorist, a present, and a future tense. In Sanskrit it has gone a step farther, and from the root as- (= es-) has developed a perfect. In Latin it has the full complement of moods and tenses, but only those tenses we find developed in Greek are from the root es-; the perfect tenses are formed from the root fu- to grow or become. So fuit, which we have just seen with the meaning: non iam est ‘it exists no longer’, must also have meant at times the exact opposite, as it developed from its primary sense ‘it has come to be’ to the sense ‘it is’. This force comes out clearly in some uses of fuerat for erat, ‘it had grown to be’, and hence ‘it was’. So in: natorum Tyrrhi fuerat qui maximus, Almo (Aen. 7. 532), illi fuerat Saturnia nomen (8. 358), piscosae cui circum flumina Lernae ars fuerat pauperque domus (12. 519). So we have fuerint for sint in: hic tibi ne qua morae fuerint dispendia tanti (3. 453). We have fuit in its old sense for factum est in: usque ad illam (defectionem) quae Nonis Quintilibus fuit regnante Romulo (Cic. Rep. 1. 16). Fio, the present in use for factus sum, seems an assimilation from fuio, where the i, like that in capio, is a mark of the present. Just as capio had not yet this i when from it the compound occupo was formed, so fuat is formed, not from fuio, but from an older fuo. Plautus uses fiat for esto in Amph. 770 and Most. 1038; and fit seems for est in Aen. 10. 153. The active form deficit seems for deest in: tenent Danai, qua deficit ignis (2. 505). In return we have esse used for fieri in: nymphasque e navibus esse iusserat (10. 221), sed fatis incerta feror si Iuppiter unam esse velit Tyriis urbem Troiaque profectis (4. 111), cum faber incertus scamnum faceretne Priapum, maluit esse deum (Sat. 1. 8. 3). So too in the beginning of Sat. 2. 3 to quid fiet? (v. 4) answers nil est (v. 6), clearly for nil fit. As auxiliaries for the perfect tenses of the passive at times for sum, eram, and ero we have fui, fueram, and fuero, evidently with this older meaning of ‘I have come to be’ and so ‘I am’.

In French too the verb ‘to be’ is in use in all its moods and tenses; but here in addition to tenses derived from fu- we have for erat était for estabat, a later form of stabat, and for the past participle, for which the Roman used factum usually to be supplied, the French has été for statum. Evidently in later Latin esse was joined with stare. To: Strophades Graio stant nomine dictae (Aen. 3. 210) Servius’s note is: 148 stant sunt. In: tanta stat praedita culpa (Lucr. 5. 199) stat seems for est; and in: quin ubi transmissae steterint trans aequora classes (Aen. 3. 403) transmissae steterint seems for transmissae sint. In:

Flecte vias hac qua madidi sunt tecta Lyaei

Et Cybeles picto stat Corybante tholus (Mart. 1. 70. 9-10),

one feels that Martial might have written stant and est with like propriety. So, too, in:

non tua sunt duro praecordia ferro

Vincta ec in tenero stat tibi corde silex (Tib. 1. 1. 63-4),

stat serves better than est to emphasize the idea of hardness. In:

omnino finem non esse secandis

Corporibus faciunt neque pausam stare fragori (Lucr. 1. 746-7),

esse and stare seem used for poetic variety. To: omnis in Ascanio cari stat cura parentis (Aen. 1. 646) Servius’s note is: stat modo est. In: stat magni nominis umbra (Luc. 1. 135) stat is probably poetic for restat, though it comes very near est in force. In: sedesque adstare relictas (Aen. 3. 123) Servius explains adstare as esse sub hoc; in: quis feret uxorem, cui constant omnia? (Juv. 6. 166) constant seems merely a strengthened sunt, ‘who has all things in full possession’; and in: nunc qua ratione quod instat confieri possit (Aen. 4. 115) instat seems ‘is beginning to be’. So in: vides ut alta stet nive candidum Soracte (Od. 1. 9. 1) stet is not exactly sit, but rather for exstet, as in stat ferrea turris ad auras (Aen. 6. 554), and gives the meaning of ‘towers on high’, or ‘is plain to view’; which is how the existence of Soracte impresses the spectator from the Janiculum; and it has much the same force in: stat cruor in templis (Luc. 2. 103). We have this meaning of ‘arise’, but in a figurative sense, in: stant belli causae (Aen. 7. 553) and: altis urbibus ultimae stetere causae cur perirent (Od. 1. 16. 19). We have esse for stare in: speravimus ista, dum fortuna fuit (Aen. 10. 43 and 3. 16) and:

Et sublata volantis ungue Prognes

In nido seges est hirundinino. (Mart. 11. 18. 19-20.)

The verb stare is of such importance in Ennius and Virgil that I may be excused a little fuller treatment of it. It may be considered from two points of view, as denoting either posture or position. As posture it is natural to think of stare as opposed to sedere, as we see it in: hi stant ambo, non sedent (Pl. Capt. 2). We see this opposition in: mediisque sedent convallibus arva (Luc. 3. 380) as opposed to stet . . . Soracte. But sedere, as well as stare, is opposed to ambulare, as is clear from: si non ubi sedeas locus est, est ubi ambules 149 (Pl. Capt. 12); and so sedere is often only a stronger stare. So of birds perching in:

Alitis . . .

Quae quondam in bustis aut culminibus desertis

Nocte sedens serum canit importuna per umbras (Aen. 12. 862-4),

to which Gesner notes: aves etiam pedibus nixae tamen sedere dicuntur. So we find many uses of stat parallel with those of sedet, as in: celsa sedet Aeolus arce sceptra tenens (Aen. 1. 56) or: principis angusta Caprearum in rupe sedentis (Juv. 10. 93), parallel with: ad undam stat lacrimans (Geo. 4. 356) or: nullam nisi olenti in fornice stantem (Sat. 1. 2. 30). We have it transferred to the sea in: fluminis intrastis ripas portuque sedetis (Aen. 7. 201), but: stant litore puppes (6. 901). So both are opposed to nare in: tempus fuit quo navit in undis, nunc sedet Ortygie (Ov. Met. 15. 336-7) and: cum Phoebus linquens stantem se vindice Delon (Prop. 4. 6. 27). It is opposed to motion in: his dictis sedere minae (Sil. 10. 623), sedit rabies (Stat. Theb. 10. 823), sedeant spectentque Latini (Aen. 12. 15); as is stare in: veluti stet volucris dies (Od. 3. 28. 6) and: cum placidum ventis staret mare (Buc. 2. 26) and: stant mihi cum domina proelia dura mea (Prop. 3. 5. 2). It is opposed to all activity in: meliora deos sedet omnia poscens (Geo. 3. 456) and: sedit qui timuit ne non succederet (Hor. Ep. 1. 17. 37), and in: quid agitur? statur (Ter. Eun. 271) ‘How goes it?’ ‘Nothing doing’. Sedet is used for certum est in: si mihi non animo fixum inmotumque sederet (Aen. 4. 15), as is stat in: stat casus renovare omnes (2. 750). In Aen. 4. 15 sederet is very close to esset, owing to its union with fixum inmotumque; and to: Turnus sacrata valle sedebat (9. 4) Servius notes: ut Asper dicit, ‘erat’.

Stare is not merely the opposite of cadere, as we see it in: securus cadat an recto stet fabula talo (Hor. Ep. 2. 1. 176), but of iacēre; and we see this opposition emphasized in the arrangement of the verse:

Canities inculta iacet, stant lumina flammae (Aen. 6. 300).

Iacet is often like stat merely of position, as in: patriam . . . sub sole iacentem (Geo. 2. 512), or: Sicanio praetenta sinu iacet insula (Aen. 3. 692); but, unlike stat, it is thus used of lands rather than of living persons. More commonly it is used of earth, or sea, or snow lying still and undisturbed, as in: neu segnes iaceant terrae (Geo. 2. 37), postquam iacuit planum mare (Juv. 12. 62), cum nix alta iacet (Geo. 1. 310). It may denote low-lying land, as does sedere, as in 150 Thapsumque iacentem (Aen. 3. 689), where Servius explains: paene fluctibus par; or a plain on the mountain top, as in: in vertice montis planities ignota iacet (Aen. 11. 527), where the oxymoron is purposed.

When transferred to persons iacere is used of the slain, as in: saevus ubi Aeacidae telo iacet Hector (Aen. 1. 99), or of the sick, as in: cum tristi morbo defessa iaceres (Tib. 1. 5. 9), or of those neglected and contemned, as in: pauper ubique iacet (Ov. Fast. 1. 218). We have the opposite of this in Ennius’s noble verse:

Moribus antiquis res stat Romana virisque (Ann. 426, M.),

and in:

Quo vobis mentes, rectae quae stare solebant

Antehac, dementes sese flexere viai (Ann. 204-5, M.),

and turning to trees, that stand erect on the earth’s surface:

longique cupressi

Stant rectis foliis et amaro corpore buxum (Ann. 265-6, M.).

So Virgil:

Stant et iuniperi et castaneae hirsutae;

Strata iacent passim sua quaeque sub arbore poma (Buc. 7-53-4).

We have the same opposition clearly marked in:

Cui nec arae patriae domi stant,

Fractae et disiectae iacent. Fana flamma deflagrata,

Tosti alti stant parietes (Enn. Fab. 166-7, M.).

When we compare: tantum campi iacet (Geo. 3. 343) with: fulvae nimbus harenae tollitur (Geo. 3. 111), we feel that iacet is used of the field in its normal, peaceful state. When we turn to: stant pulvere campi (Enn. Ann. 314, M.) stant seems of fields, whose dust no longer lies quiescent, but is surging and tossing to the skies. Such a field seems to Ennius no longer to lie, but to be up and standing erect.

Servius has no doubt of ihe meaning of: stant lumina flamma (Aen. 6. 300). Stant horrent is his note. Evidently Charon’s eyes bicker with flame like Lucretius’s: claraeque coruscis fulguribus . . . taedae (5. 295-6). But is he explaining the genuine reading? For Acron (ad Hor. Od. 1. 9. 1) read: stant lumina flammae. Donatus, however (ad Ter. And. 699), reads flamma like Servius. When we turn to the manuscripts we have a different story. Of the six manuscripts Ribbeck thinks worth specifying, four had flammae, and only two flamma. Of the four, two had the e of flammae crossed out; and of the two, one had e added to flamma by a later hand. So that the testimony of the best codices is strong for flammae.


Though flammae seems at first sight the more difficult reading, it is easy to understand it here, and it offers the variety of meanings in which Virgil delighted. Virgil imitating Ennius’s: stant pulvere campi gives us: iam pulvere caelum stare vident (Aen. 12. 408). Servius explains stare here as plenum esse, the explanation usually given for Ennius’s stant, but adds: alii ‘stare’ constare intellegunt, ut significet ‘pulvere caelum constat’, and this seems the right account. Virgil wishes to transfer Ennius’s stare from the earth to the sky, and finds he can do so by using it for constare in poetic style. In this sense it naturally takes an ablative of material. And so Servius adds: id est, (caelum) in pulverem versum est, et quasi totum ex pulvere est, from which we see that his sense: plenum est is closely related to constat. Plenum may take an ablative, but usually governs a genitive; and if we read with the majority of the best manuscripts: stant lumina flammae, it is natural to take stant here too as for plena sunt.

But the scholars of the time, excepting Acron, read flamma. It is quite possible that Virgil left both readings, between which he had not himself reached a final decision. The juxtaposition of iacet and stant, making stant seem for horrent, favours the reading flamma. But the genitive is much more vivid; and we find that, while scholars favoured flamma, the ordinary reader called for flammae, which is what we should expect. The codices, of course, give the text favoured by the mass of readers. The expression becomes very picturesque if we interpret: ‘his eyes are eyes of flame’. But the reader may take flammae as a predicate nominative: ‘his eyes are flames’. Virgil, unlike Horace, would surely have chosen to please the mass of his readers, and stant lumina flammae would have been his final choice.

In: stat sentibus fundus (Lucil. 5. 5, M.) Donatus (ad Ter. And. 699) seems right in taking stat for horret, which is clearly its force too in: interea stat sentibus pectus (Lucil. 5. 4, M.). Ennius uses horrere of weapons standing erect in: densantur campis horrentia tela virorum (Ann. 316, M.), and in: sparsis longis hastis campus splendet et horret (S. 14, M.) ‘glitters and quivers’, at which verse, according to Servius (ad Aen. 11. 601), Lucilius made mock, saying it should have been ‘horret et alget’ ‘quivers and shivers’. We notice Virgil’s transfer of epithet in: tum late ferreus hastis horret ager (Aen. 11. 601); what can ferreus ager be? When a field bristles with spears of iron, it may be described as a field of iron. In: stat ferri acies 152 mucrone corusco stricta (Aen. 2. 333) Servius thinks that stat is for horret, or else it is transferred from stantibus armatis; and in: stat ductis sortibus urna (Aen. 6. 22) he thinks stat sortibus is for horret sortibus, much as: (apes) pennis coruscant (Geo. 4. 73). To me it seems rather for: non iam vertitur; the lots have been drawn, and the urn no longer turns.

Nonius (422. 32, M.) quotes from the Medea of Accius (413, R.): mare cum horreret fluctibus. From this we can understand the meaning of Catullus’s nuntius horribilis (84. 10). Arrius has been exasperating the ears and minds of the Romans by his aspirations, such as chommoda for commoda and hinsidias for insidias; but he has gone east with Crassus, and now all ears have a rest. But suddenly there comes a nuntius horribilis, a bit of news making your hair stand on end: Ionios fluctus, postquam illuc Arrius isset, iam non Ionios esse sed Hionios. Evidently the waves have been so aspirated by Arrius that they are now magis solito asperatae. We have a similar confusion of vowels, when Servius explains: turrigeraeque urbes (Aen. 10. 253) as: quas terra gerit. Catullus thinks the aspiration of the Etruscan Arrius so violent and perverse as to give a new roughness to the waves of the Adriatic, which he has just crossed. It is as if the tendency to aspirate of our own ’Arry were to prove so infectious as to give a new roughness to the Straits of Dover.

We have stare used to denote position in: si propius stes (A. P. 361), quorum statuae steterunt . . . in Rostris (Cic. Phil. 9. 4. 2), et primo haud impares stetere acies (Liv. 26. 44. 3). Here it seems the opposite of ire and venire, and often equivalent to haerere and manere. Its use is close to manere in: stare loco nescit (Geo. 3. 84), where the prose would be: manere in loco nescit. We may compare: pacto stas (Liv. 9. 11. 2) and: stare conventis (Cic. Off. 3. 95. 25) with: at tu dictis, Albane, maneres (Aen. 8. 643). So in: bene apud memores veteris stat gratia facti (4. 539) bene stat is for: firma manet. Cf. also: Troiaque nunc stares, Priamique arx alta maneres (2. 56). In: mole sua stat (10. 771) and: stat sua cuique dies (10. 467) stat is for fixa est. In: in ducibus stabat spes et victoria (Sil. 17. 400) it is for defixa est, or dependet.

Like stare in: sed abi intro; noli stare (Pl. Mil. 1129) or: i: quid stas, lapis? (Ter. Heaut. 831), is haerere in: paulum adspectu conterritus haesit continuitque gradum (Aen. 3. 597), attonitis haesere animis (5. 529), Hectoris Aeneaeque manu victoria Graium haesit (11. 290). So haeret is used for fixus est in: hic terminus haeret (4. 614) and: 153 soloque inmobilis haeret (7. 250), and for manet in: inceptoque et sedibus haeret in isdem (2. 654) and: qua spe gelidis in nubibus haeres (12. 796). It is used as a passive of teneo in: haesere caeno fossisque impedimenta (Tac. Ann. 1. 65. 4) and: haeret pede pes (Aen. 10. 361), and for the passive of retineo in: vox faucibus haesit (2. 774).

Stare in the sense of position approaches nearly the sense of esse in Cicero’s: a se potius quam ab adversariis stare (Inv. 1. 81. 43), in Plautus’s: hinc stas, illim causam dicis (Men. 799) and Virgil’s: Iuppiter hac stat (Aen. 12. 565). But manere and haerere are rather for semper esse, as in: natura manet sine pondere inanis (Lucr. 1. 363) or: hi in oculis haerebunt (Cic. Phil. 13. 5. 3). In return est is for manet in: est hic, est animus lucis contemptor (Aen. 9. 205).

Clueo (older cluo = κλύω) is used in Latin like audio as a passive of dico or nomino, as in Ennius’s: nostra Latinos per populos terrasque poemata clare cluebunt (Ann. 3, M.) or Lucretius’s: per gentes Italas hominum quae clara clueret (1. 119), where however in union with the cognate clara it approaches the sense of esse. In Lucretius it seems often to have this force: as in: quaecumque cluent (1. 449), quae nondum clueant ullo temptata periclo (1. 580), inter se nota cluere (2. 351). Vivo seems for sum in: ecquis me hodie vivit fortunatior? (Ter. Eun. 1031), purus et insons (ut me collaudem) si et vivo carus amicis (Sat. 1. 6. 70), viveret in terris te si quis avarior uno (Ep. 2. 2. 157). In return we have esse for vivere in:

Ostendent terris hunc tantum fata neque ultra

Esse sinent (Aen. 6. 870).

With a predicate implying position we have stare used for esse; with a predicate implying motion ire is often so used. Servius’s note to:

sed non felicibus aeque

Tum comes auspiciis caro datus ibat alumno (Aen. 11. 32-3),

is: ibat pro ierat according to Thilo and Hagen. This is pure nonsense, and is apparently a scribe’s blunder for: ibat pro erat. So in: ibis ab excusso missus in astra sago (Mart. 1. 3. 8) we have in missus ibis a more picturesque missus eris. We have a like use of the verb ire, but not now as an auxiliary, in: nuntius ibis Pelidae genitori (Aen. 2. 547) and in: quibus ibat in armis aureus (9. 269), si non tanta quies iret frigusque caloremque inter (Geo. 2. 344), non tutior ibis Homero (Prop. 2. 34. 45), animo tam procul ibit amor (3. 21. 10), non ibo inulta (Sen. Her. Oet. 282), ite alacres et spiritus pleni (Curt. 4. 14. 25), non dabitis murum sceleri; qui vindicet, ibit (Claudian de IV Cons. 154 Hon. 109). It seems rather for fieri in: incipit res melius ire quam putaram (Cic. Att. 14. 15. 2), nec miror ista sic ire (Sen. Ep. 1. 5. 8), and: sic eat (Luc. 2. 304 and 5. 297). It is used for ferri in: postquam omnia fatis Caesaris ire videt (Luc. 4. 144); and in: solus ego in Pallanta feror (Aen. 10. 442), where Servius carelessly explains feror as ferri debeo, it is evidently for ire debeo. Just as in French we have j’ai été for je suis allé, so here we have fuisse for ivisse in: quod ferar in partes ipse fuisse tuas (Prop. 3. 9. 60). Gellius (1. 7. 16) quotes from Cicero: cum vestros portus . . . in praedonum fuisse potestatem sciatis (Leg. Man. 33. 12), where the change to potestate in the manuscripts is easy to understand. We have in Petronius: fui enim hodie in funus (42), and in Suetonius: nec prius surrexisse ac militibus in conspectum fuisse (Aug. 16. 2).

When we turn to Italian we find venire used for essere in forming the passive voice. So in Latin we have venire in periphrasis with the appropriate noun to give the passive of verbs that have no passive form in use. For the passive of suspicari we have: in suspicionem venire (Cic. ad Q. Fr. 1. 1. 14. 4), of periclitari, in periculum venire (Caes. B. C. 1. 17. 2), for odisse, ne in odium veniam (Cic. Fin. 2. 79. 24). Modelled on these we have alternatives for the passive in: in dubium venire (Liv. 3. 13. 7), for dubitari, in cruciatum venire (Caes. B. G. 1. 31. 2), in contemptionem venire (ib. 3. 17. 5). Close to this seems its use for fieri in: in proverbium venit (Liv. 40. 46. 12). Not far removed is the use for nasci in: hic segetes, illic veniunt felicius uvae (Geo. 1. 54) and: (arbores) sponte sua veniunt (Geo. 2. 11). Similar is the use of ire for crescere in Cato; donicum in semen videris ire (R. R. 161. 3).

We have venire used as a more picturesque esse in: dummodo morata recte veniat, dotatast satis (Pl. Aul. 239), non impune illa rogata venit (Prop. 1. 5. 32), non tamen haec ulli venient ingrata legenti (id. 2. 34. 81), gratior et pulcro veniens in corpore virtus (Aen. 5. 344), pelagine venis erroribus actus (6. 532), rebusque veni non asper egenis (8. 365), nam venio moriturus (10. 881), cum fletu nox vigilanda venit (Tib. 1. 2. 76), canem illum invisum agricolis sidus venisse (Sat. 1. 7. 26), tarda venit dictis difficilisque fides (Ov. Fast. 3. 350), quaerens quibus mortifera veniat (Sen. Med. 687-8), muneribus venit tegula missa tuis (Mart. 7. 36. 4). In: se satis ambobus Teucrisque venire Latinisque (Aen. 7. 470) satis venire seems for parem fore. We have venire with nouns for esse in: nam tu solacia praebes; tu curae requies, tu medicina venis (Ov. Trist. 4. 10. 118), 155 and: quaenam tot divis veniet nurus (Claudian de IV Cons. Hon. 647). As in all those uses, so in the uses of venturus for futurus its primary meaning persists, as we see in: ventures nautis prodentia ventos (Aen. 10. 99), and: idem venturos tollemus in astra nepotes (3. 158), to which Servius’s note is: venturos plus est quam si dixisset ‘futuros’: nam quasi eos iam esse significat. In: veniens Marsorum montibus Umbro (Aen. 10. 544) veniens is for qui venerat, the opposite in meaning of veniens in: veniens in aevum (Od. 3. 5. 16).

So esse is used for venire in: numero mihi in mentem fuit dis advenientem gratias pro meritis agere atque adloqui (Pl. Amph. 180), as Gellius notes (1. 7. 17), where: in mentem fuit is clearly for: in mentem venit. Not so clear is: ecquid in mentem est tibi, patrem tibi esse? (Pl. Bacch. 161), where est seems for adest, often used as a perfect of advenire, as in: vesper adest (Catull. 62. 1) or: atque utinam rex ipse noto compulsus eodem adforet Aeneas (Aen. 1. 576), where adforet is plainly for advenisset, though Servius carelessly says it is for adveniat. Plain seems esse for venisse in: ut certior fieret quo die in Tusculanum essem futurus (Cic. Att. 15. 4. 2). It seems for convenire in: senatus hodie fuerat futurus (ib. 4. 17. 4). In: fluminaque Haemonio comminus esse viro (Prop. 3. 1. 26), where most editors change esse to isse, disregarding the manuscripts, esse is probably for venisse.

When we compare the diction of Latin prose with that of Latin poetry we see at once how much greater is the role of the noun in the latter. To the frequent and constant use of verbs in Latin prose is due its vibrant, elastic strength and pliancy. When we substitute nouns for verbs we get a vague cloud-like grandeur of effect far removed from the vivid reality of prose. We might compare: tibi imperanti semper parebo with: numquam frustrata vocatus hasta meos (Aen. 12. 95), or: sic respondet with: sic ore vicissim orsa refert (Aen. 11. 123-4). Poetic diction tends to the substitution of nouns for verbs, which we find so prevalent in English prose, when compared with that of Cicero or of Caesar. So in: impudens Orcum moror (Od. 3. 27. 50) for mori moror, or: dedidicit iam pace ducem (Luc. 1. 131). Very often instead of a single verb like amplexabitur we have a verb with a noun like: dabit amplexus (Aen. 1. 687). So instead of vovit we have vota facit (11. 50), instead of regnabat, regna habebat (1. 346), instead of gemet, aget gemitus (6. 873), instead of dicit, dictis it (10. 448), instead of liceat mihi, sit mihi fas (6. 266), instead of quibus valde intereram, 156 quorum pars magna fui (2. 6), instead of nec dis placet, nec dis amicum est (Od. 2. 17. 2). Of course these follow the analogy of older phrases usual in prose like poenas dare, vela dare, iura dare, vitam agere, vota reddere, cordi esse, curae habere.

Of auxiliaries thus used dare is the most common in Virgil; to cite some examples, we have for pariet, partu dabit (Aen. 1. 274), for cecinere, cantus dedere (1. 398), for cedunt, dant locum (2. 633), for lacrimavit, lacrimas dedit (4. 370), and so dicta dabat (5. 852), fugam dedit (7. 24), cursum dedit (10. 870), saltum dedit (12. 681), ruinam dant (11. 614), excidio daturum (12. 655), milia multa daret leto (5. 806). Of verbs which have arisen from such phrases we might mention mando, vendo, and credo (cf. Skt. śrad-dha).

Less usual is a like use of ago, though in older Latin it must have been very common in such phrases; as is clear from verbs like levigo, litigo, mitigo, iurgo, and purgo. Of the phrases found with it most belong to prose diction as in: gemmas agant for geminent (Varr. R. R. 1. 30), animam agere for mori (Cic. Fam. 8. 13. 2), bellum agere for bellare (Caes. B. G. 3. 28, Herz), laborem agere for laborare (Cic. Fin. 2. 105. 32), crimen agere for criminari (Verr. 2. 4. 48. 22), curam ago for curo (Liv. 6. 15. 11). In the poets I note: maximas nugas agis (Pl. Asin. 91), vitam . . . agebat (Geo. 2. 538), agit acri remige (Aen. 5. 116). In: nullo discrimine agetur (Aen. 1. 574) we seem to have the passive of nullo discrimine habebo (10. 108).

Verbs like significo, amplifico, honorifico, magnifico, horrifico, terrifico are derived from older phrases, such as we see in: facimus meritosque novamus honores (Aen. 8. 189). So we have vota facit (11. 50), vela facit (5. 281), indicium faciet (Geo. 2. 246). We have: fecere ruinas (Lucr. 1. 740) but: dabant . . . ruinas (id. 5. 1329), dedit . . . ruinam (Aen. 2. 310); nomen . . . fecere (Geo. 4. 272) but: nomen dedit (Aen. 1. 248). In: fecere pedem (Aen. 5. 830) we have the equivalent of vela dabant (Aen. 1. 35); and faceret pretium (Mart. 1. 85. 7) is plainly for emeret, where facere is for dare. Fit sonitus (Geo. 4. 79) is the passive of sonitum . . . dedere (Aen. 10. 488), fit gemitus (Aen. 6. 220) of gemitum dat (1. 485). Parallel to: date volnera lymphis abluam (Aen. 4. 684) is: tu facito . . . sis memor (12. 438); and to: dedit esse deas (10. 235) is: nati coram me cernere letum fecisti (2. 539), where the use of facere with the infinitive instead of curare with ut cum subiunctivo gives a curiously modern effect. Servius transfers this idiom to his prose in: di . . . Aeneam huc venire fecerunt (ad Aen. 4. 45), 157 and it is the regular idiom in modern French. Different is the force of the infinitive in:

Fecerat et viridi fetam Mavortis in antro

Procubuisse lupam (Aen. 8. 630-1);

here fecerat seems for fecerat videndam, and Virgil uses procubuisse for procubantem; for the infinitive or present participle can be used with like sense after videre.

In: nunc te mea dextera bello defensum dabit (Aen. 12. 437) we have a parallel to: habeat victos (ib. 17) i.e. in ‘it will defend’ to ‘let him conquer’. The idiom with habeo is much more common in prose, as in: pecunias magnas collocatas habent (Cic. Leg. Man. 18. 7), de Caesare satis hoc tempore dictum habebo (Cic. Phil. 5. 52. 19), ut . . . scaphas ad litus adpulsas habeant (B. C. 2. 43), qui omnia . . . domita armis habeat (Liv. 7. 32. 9). In: persuasum habet, notum habet, exploratum habet, compertum habet, cognitum habet, it is virtually the French perfect with avoir. But it is rare in poetry, where the use with dare is the common one. Besides: regna . . . habebat (Aen. 1. 346) I notice: spem si quam . . . habuistis (11. 308), imperium . . . habebit (9. 449), numen habere maris (10. 221), arma Latinus habeto (12. 192) ‘Latinus shall command the army’. In: tu quoque magnam partem opere in tanto . . . haberes (Aen. 6. 30-1), as in: quorum magna pars fui (2. 6), we have poetic substitutes for interesses and intereram. We seem to have the opposite of this dissolution in: te patrios miscere iuvat cum coniuge census (Mart. 4. 75. 3), where miscere seems substituted for communes habere.

We read: sit numine vestro (Aen. 6. 266) for: per vos liceat, non opis est nostrae (1. 601) for: non possumus, and: vestro . . . in numine Troia est (2. 703) for: vos Troiam tuemini. In: longe illi dea mater erit (12. 52) longe erit ‘will be afar’ seems to be poetic for aberit ‘will fail to aid’. We note: naribus duces (Od. 4. 1. 21-2), auribus accipere (Lucr. 4. 982), addidit ore (Aen. 2. 593). In: discrimina costis per medium qua spina dabat (Aen. 10. 382-3) dabat discrimina is for discernebat, and we have a passive of this in: tenui discrimine leti esse suos (10. 511-12), which would be in prose: minimum abesse quin sui perirent. But in: (cum) esset . . . aequalis Mars utriusque diu (Mart. Ep. Lib. 29. 2) we have a poetical equivalent for: cum uterque aequo Marte diu certaret.

So in: non lacrimis hoc tempus (Aen. 12. 156) lacrimis is for lacrimare; and in: subtracta sibi quaereret arma dolor (Mart. 1. 42. 2) dolor is for Portia dolens. In: lacrimis iamque peractus eras (Mart. 7. 158 47. 6) lacrimis is for nobis lacrimantibus. While dolor is here subjective, in: postquam primus amor deceptam morte fefellit (Aen. 4. 17) amor is objective, being for: is quem primum amavit. So in: castigant . . . moras (4. 407) for: eos qui morabantur; fletus fert . . . refertque soror (4. 437) for: flentis mandata; equitum levia . . . arma (11. 512) for: equites leviter armatos. While Virgil in: medio . . . campo (Geo. 3. 466) follows the prose construction and treats medio as an adjective, in: castrorum et campi medio (Aen. 9. 230) he uses medio as a noun, and in: medio sermone (= sermone interrupto) (4. 277) and fugae medio (11. 547) we have medio in both taking the place of a past participle.

We have nouns for a subjunctive with ut in: missio saepe viris magno clamore petita est (Mart. Ep. Lib. 29. 3). So too in: suadent somnos (Aen. 2. 9). In: nescis dominae fastidia Romae (Mart. 1. 3. 3) fastidia is for: quantum fastidia, and in: temptaturum aditus et quae mollissima fandi tempora (Aen. 4. 293-4) aditus is clearly for: quanam eam adiret. So we have nouns used for infinitives after verbs of hearing or knowing in: coniugis audisset fatum cum Portia Bruti (Mart. 1. 42. 1), where coniugis fatum is for: coniugem periisse; veterem . . . agnovit amicum (Aen. 3. 82) for: amicum olim fuisse. So after verbs of hoping, promising, or refusing in: sperare salutem (1. 451), praesens . . . minatur exitium (12. 760-1), canitiemque sibi et longos promiserat annos (10. 549), cursumque recusant (12. 747), and after a modal verb in: rebellionem coeptavere (Tac. Ann. 3. 40. 1). In: ego nec tumultum nec mori per vim metuam (Od. 3. 14. 14-15) we have a noun coupled with an infinitive, as we have in: iram miserantur inanem amborum et tantos mortalibus esse dolores (Aen. 10. 758-9). In: fugae studio (Aen. 4. 400) fugae is for fugiendi, and in: quem tum vates Cassandra moveret? (3. 187) vates is for vaticinando. In: satis prospectum urbanae servituti (Tac. Ann. 1. 46. 4) we have a poetic expression for: urbi in servitutem redigendae satis esse provisum. In: exim promptum quod multorum intimis questibus tegebatur (ib. 3. 36. 1) we have Tacitean brevity for: deinde id prolatum est quod ante multi clam querendo secretum tenebant. So: intersint . . . patris lacrimis (Aen. 11. 62) is a poetic brevity for: patri lugenti lacrimando consentiant.

It seemed to me as a boy that our phrase ‘with fire and sword’ had a great advantage over Cicero’s: vi et armis (Sest. 78. 36) or even: vim et ferrum (ib. 79. 37), and Virgil has the same feeling, for he 159 writes: qui face Dardanios ferroque sequare colonos (Aen. 4. 626); and Tacitus strengthens the phrase in: igne et caedibus perfidiam ultus est (Ann. 2. 8. 4). In: hos successus alit (Aen. 5. 231) or: successu exsultans (2. 386) our schoolboy meets a familiar word that he has sadly missed in Caesar and Cicero. He wastes no time over the propriety of Virgil’s use of viam in: fata viam invenient (Aen. 3. 395), or of modi in: haud ignara modi (10. 247), and greets visus as an old friend in: rite secundarent visus (3. 36) ‘might duly aid my plans’. He feels no need of explanations about: Danai puellas (Od. 3. 11. 23), but when he comes to Virgil’s use of: meliorem animam (Aen. 5. 483), he may doubt Virgil’s eminence in theology. But in: humum semel ore momordit (11. 418), if he is fortunate enough to have this reading presented to him, he will probably master the sense without the aid of Servius. My Clarendon Press text reads simul, following Ribbeck; but of his six manuscripts cited three give semel, one semul, one simul corrected to semel, and one only simul. Servius read semel, and his note is: semel cito confestim, i.e. qui tota mortis celeritate consumptus est, quasi nihil ultra passurus. Volnerati enim solent vel terram vel arma mordere, ne dolorem eorum indicet gemitus. Evidently semel is for: semel tantum, non bis, and to Servius: humum semel ore momordit means: vita confestim excessit; showing that to the Roman terram momordit had a second meaning corresponding to our figurative sense of ‘bit the dust’.

Very familiar to us seems sub te in: nos tumidum sub te permensi classibus aequor (Aen. 3. 157), where in prose we expect te duce. The phrase is probably elliptical, and was parallel to: sub te . . . magistro (Aen. 8. 515). Sub here is rather a preposition than an adverb; while in such phrases as: grato . . . sub antro (Od. 1. 5. 3) ‘below in the pleasant grot’, or: Acheronte sub imo (Aen. 11. 23) ‘in Acheron’s lowest depth, down below’, it is still a mere adverb. Familiar too seems: sine ictu (2. 544) for nullo volnere dato. Prose favours nullo or nullis in many phrases where we have sine in verse, as in: sine nomine corpus (2. 558) for nullo honore cadaver. Quite familiar too seems: ante annos animum . . . gerens (9. 311), where there is an ellipsis of sapentiorem, the prose being: animum gerens sapientiorem quam pro annis. For this use of ante for magis quam we may compare: ante omnes stupet ipse Dares (5. 406) and: ante alios inmanior omnes (1. 347).

The simplicity of Ennius, who in: saxo cere comminuit brum (Ann. 552, M.) ‘he smashed his head with a rock’ tried by this 160 disposition of cerebrum to represent to his readers the result of the catastrophe, finds a gentler echo in: seque gregari (Lucr. 1. 452). In: Argi nempe soles subire letum (Mart. 1. 117. 9) the like disposition of Argiletum seems intended to indicate the close relation common between the initial and final words of the verse. In: quadrupedumque putrem cursu quatit ungula campum (Aen. 11. 875) we have this relation between pairs of words at the beginning and end; and in: quadrupedante putrem sonitu quatit ungula campum (8. 596), a like correspondence of pairs indicates that quadrupedante, transferred in syntax to sonitu, belongs in sense to ungula. The repetition in: repetens iterumque iterumque monebo (3. 436) serves to heighten the force of iterum, while the enhancement of the effect in: nam thermis iterumque iterumque iterumque lavatur (Mart. 2. 14. 13) recalls Virgil’s use of: alterum in alterius (Aen. 2. 667) for three persons. So with the position of inter in: latus inter et ilia (10. 778); while that of adventu and abitu in: adventu manet incolumis natura abituque (Lucr. 1. 457) and of demum in: nam solido vincunt ea corpore demum (id. 1. 486) has a like effect. Fine is the effect of the spondaic close in: ex infinito iam tempore subsidendo (id. 1. 995).



We noticed in our last paragraph how Virgil in two of his most striking verses had arranged the words so that a pair at the beginning of each should correspond to a pair at the end. This tendency to arrange words in pairs, very evident as it is, would be still more so, were it not that a pair is so often expressed by one of its members; just as we found the Dioscuri named by either Castor or Pollux. From this ellipsis arise the figures of synecdoche and metonymy, the most common in poetry, and so those whose nature and origin it most concerns us to trace.

Whether we turn to Latin prose or poetry, we find a notable tendency, growing with the growth of emotion, to the expression of objects and qualities in related pairs. Turn for a moment to Plautus, where in the Captives Ergasilus enters in triumph:

Iuppiter supreme, servas me, measque auges opes.

Maximas opimitates opiparasque offers mihi,

Laudem lucrum, ludum iocum, festivitatem ferias,

Pompam penum, potationis saturitatem, gaudium (vv. 768-771).

In Terence, Menander’s faithful imitator, we often have a threefold arrangement of ideas, in which he is especially followed by Horace. This is joined with the arrangement into pairs of which we are speaking in:

perpulisti me, ut homini adulescentulo

In alio occupato amore, abhorrenti ab re uxoria

Filiam ut darem in seditionem atque in incertas nuptias,

Eius labore atque eius dolore gnato ut medicarer tuo (And. 828-31).

When we turn to Cicero, this is prominent, but often further developed in the union of two pairs into four, as in:

Saxa et solitudines voci respondent, bestiae saepe inmanes cantu flectuntur atque consistunt: nos, instituti rebus optimis, non poetarum voce moveamur? Homerum Colophonii civem esse dicunt suum, Chii suum vindicant, Salaminii repetunt, Smyrnaei vero suum esse confirmant, itaque etiam delubrum eius in oppido dedicaverunt: permulti alii praeterea pugnant inter se atque contendunt (Arch. 19. 8). 162 Caesar in his descriptions shows the same tendency, as in:

Cum bellum civitas aut inlatum defendit aut infert, magistratus, qui ei bello praesint, ut vitae necisque habeant potestatem, deliguntur. In pace nullus est communis magistratus, sed principes regionum atque pagorum inter suos ius dicunt controversiasque minuunt (B. G. 6. 23. 4-5).

Let us turn to Virgil:

Iam gravis aequabat luctus et mutua Mavors

Funera: caedebant pariter pariterque ruebant

Victores victique, neque his fuga nota neque illis (Aen. 10. 755-7),

and to Horace:

Laurea donandus Apollinari,

Seu per audaces nova dithyrambos

Verba devolvit numerisque fertur

Lege solutis;

Seu deos regesque canit, deorum

Sanguinem, per quos cecidere iusta

Morte Centauri, cecidit tremendae

Flamma Chimaerae (Od. 4. 2. 9-16),

and to Propertius:

Tarda Philoctetae sanavit crura Machaon,

Phoenicis Chiron lumina Phillyrides,

Et deus exstinctum Cressis Epidaurius herbis

Restituit patriis Androgeona focis (2. 1. 59-62),

where the Ciceronian tendency to fourfold structure is very marked; as it is in less degree in Ovid:

At Silenus abest; titubantem annisque meroque

Ruricolae cepere Phryges, vinctumque coronis

Ad regem traxere Midan; cui Thracius Orpheus

Orgia tradiderat cum Cecropio Eumolpo (Met. 11. 90-3).

In Livy it seems far more carefully developed than in Caesar or Cicero, as we find it in: at enim pauci quidem sunt, sed vigentes animis corporibusque, quorum robora ac vires vix sustinere vis ulla possit. Effigies immo, umbrae hominum, fame frigore, inluvie squalore enecti, contusi ac debilitati inter saxa rupesque; ad hoc praeusti artus, nive rigentes nervi, membra torrida gelu, quassata fractaque arma, claudi ac debiles equi: cum hoc equite, cum hoc pedite pugnaturi estis; reliquias extremas hostium, non hostem habebitis (21. 40. 8-10). It is hardly less evident in Sallust’s less laboured periods: quae res Marium cum pro honore quem adfectabat, tum contra Metellum vehementer accenderat. Ita cupidine atque ira, pessimis consultoribus, grassari, neque facto ullo neque dicto abstinere, quod modo ambitiosum 163 foret; milites, quibus in hibernis praeerat, laxiore imperio quam ante habere; apud negotiatores, quorum magna multitudo Uticae erat, criminose simul et magnifice de bello loqui (Jug. 64. 4-5). So L. Seneca: quemadmodum radii solis contingunt quidem terram, sed ibi sunt, unde mittuntur: sic animus magnus et sacer et in hoc dimissus, ut propius quidem divina nossemus, conversatur quidem nobiscum, sed haeret origini suae: illinc pendet, illuc spectat et nititur, nostris tamquam melior interest (Ep. Mor. 4. 12 (41). 5). But in Tacitus we seem to reach the crowning development of this tendency, as we see it in: haud perinde Germanos volnera, luctus, excidia quam ea species dolore atque ira adfecit. Qui modo abire sedibus, trans Albim concedere parabant, pugnam volunt, arma rapiunt; plebes primores, iuventus senes agmen Romanum repente incursant, turbant. Postremo deligunt locum flumine et silvis clausum, arta intus planitie et humida (Ann. 2. 19). I have selected these specimens from the twelve Di Maiores of Latin letters to show how they develop the union of these pairs, now weaving them in combinations of three, now of four, as we shall be led to do in tracing their ellipses, and the figures to which they give rise.

One of the most common pairs we find in Latin is: longe lateque ‘far and wide’. Virgil has it in full three times, once in his youthful poems: longe lateque per orbem (Cir. 16), once in the Georgics: longe saltus lateque vacantes (3. 477), once in the Aeneid: longe lateque per urbes (6. 378). In his numerous uses of the phrase he expresses it usually by late, as in: loca nocte tacentia late (Aen. 6. 265), late circum loca sulpure fumant (2. 698), populum late regem (1. 21), late loca milite complent (2. 495); less often by longe, as in: resonantia longe litora misceri (Geo. 1. 358), auro ductores longe effulgent (Aen. 5. 133), longeque refulget (8. 623), Asia longe pulsa palus (7. 701). It is used by Cicero in the order: late longeque as well (Balb. 13. 5), and abbreviated in: bellum . . . tam late divisum (Leg. Man. 31. 11). But the abbreviation is more usual in poetry; for the excellence of poetic diction consists rather in its power of suggesting ideas to the mind than in a full and clear expression of them.

The same is true of the adjectives longus and latus; it is very common both with us and in Latin to express the idea of extension by one of its two dimensions. So Virgil gives us: latos Haemi pinguescere campos (Geo. 1. 492), latis otia fundis (2. 468), templa . . . latis inmania regnis (Aen. 4. 199), ingentem lato dedit ore fenestram 164 (2. 482), aeris in campis latis (6. 887). We have predicative uses of it tending to adverbial in: latos vastant cultoribus agros (Aen. 8. 8) ‘far and wide they lay waste the farmers’ fields’. So too in: Sigaea igni freta lata relucent (2. 312), alius latum funda iam verberat amnem (Geo. 1. 141), latum reget aequus orbem (Od. 1. 12. 57), latam dives habebat humum (Ov. Fast. 5. 280), patet in curas area lata meas (Ov. Her. 1. 72). In: (murus) latius quam qua caederetur ruebat (Liv. 21. 11. 9) we have a metonymy of latius for longius.

While I read in Horace: latumque per aequor (Ep. 1. 2. 20) and: effusi late maris (Ep. 1. 11. 26), I read also: dum longus inter saeviat Ilion Romamque pontus (Od. 3. 3. 37), and: meliusne fluctus ire per longos fuit? (Od. 3. 27. 43). So: ex aethere longo (Aen. 7. 288) we should translate ‘from the wide sky’, and: reboant silvaeque et longus Olympus (Geo. 3. 223) ‘both the broad woods and the wide sky return the roar’. So: longa procul longis via dividit invia terris (Aen. 3. 383) is ‘a long stretch of pathless way across broad lands sunders afar’. In:

Ut saepe ingenti bello cum longa cohortes

Explicuit legio, et campo stetit agmen aperto,

Directaeque acies, ac late fluctuat omnis

Aere renidenti tellus (Geo. 2. 279-82),

while both longa and late are abbreviated forms, the propriety of abbreviating by longa on the line of march and by late on the battlefield is evident. Longus and longe are often transferred to time, as in the common phrase: longum tempus. Servius explains: nec longe (Aen. 10. 317) as for: nec multo post, where evidently longe is one for a pair: longe post. In return for this use of longe for diu we have in Horace: diu lateque victrices (Od. 4. 4. 22-3) for longe lateque. We may compare with nec longe in Virgil: longe et multum . . . antecellet (Cic. Mur. 29. 13), where, as we shall see, the union in a pair of longe and multum explains Virgil’s use of longe for multum. So Horace’s use of temporibus for rebus in: secundis temporibus dubiisque rectus (Od. 4. 9. 35-6) is connected with his union of diu lateque for ‘far and wide’.

Cicero in his Academics gives us the following enumeration of opposites: multa pauca, magna parva, longa brevia, lata angusta (Acad. Prior. 2. 92. 29). From longa and lata let us turn to brevia and angusta. In: (libellos) quos artat brevibus membrana tabellis (Mart. 1. 2. 3) pages have two dimensions, and so brevibus must be short for brevibus et angustis. As we began with lata, let us start 165 here with angusta. Just as brevis is for parvus in the example cited, so in: angustis eiecta cadavera cellis (Sat. 1. 8. 8) we have angustis for parvis. And so in: rebus angustis animosus (Od. 2. 10. 21), angustis hunc addere rebus honorem (Geo. 3. 290), spes sibi quisque, sed haec quam angusta videtis (Aen. 11. 309), neque . . . intonet angusto pectore Callimachus (Prop. 2. 1. 40), with which compare: exiguo . . . e pectore of himself (id. 4. 1. 59). So too in: tutusque mensa capitur angusta cibus (Sen. Thyest. 452), cuncta ad rem publicam referri, qua tenui angustas civium domos (Tac. Ann. 2. 33. 3), quorum virtutibus obstat res angusta domi (Juv. 3. 165). From such a union as tenui angustas seems to have arisen the metonymy of angusta for tenuis in: angusta cantare licet videaris avena (Mart. 8. 3. 21). Angustus seems a metonymy for brevis in: angusti terminus aevi (Geo. 4. 206), angustis quod equum compescit habenis (Tib. 1. 4. 11), nec mihi solstitium quicquam de noctibus aufert, efficit angustos nec mihi bruma dies (Ov. Trist. 5. 10. 7-8), usque adeone angusta dies (Stat. Th. 1. 442), temporis angusti mansit concordia discors (Luc. 1. 98).

When we read in Horace: quia scilicet illis maiorem natura modum dedit, his breve pondus (Sat. 2. 2. 37), we are apt to notice only the coincidence with our English ‘short weight’; but the contrast with maiorem here shows that breve is for parvum. So in: privatus illis census erat brevis, commune magnum (Od. 2. 15. 13-14), cena brevis iuvat (Ep. 1. 14. 35), qua brevis occultum mus sibi fecit iter (Ov. Fast. 2. 574), instruis impensa nostra sepulcra brevi (Her. 7. 188), aude aliquid brevibus Gyaris et carcere dignum (Juv. 1. 73). When we compare Martial’s: frons brevis (4. 42. 9) with Horace’s: angusta fronte (Ep. 1. 7. 26) evidently each is for brevis et angusta. So when we compare: scis in breve te cogi (Hor. Ep. 1. 20. 8) with: in parvum quendam et angustum locum concludatur (Cic. Leg. 1. 17. 5), or: res . . . adducta in angustum (Lael. 20. 5), or: ita hac re in angustum oppido nunc meae coguntur copiae (Ter. Heaut. 669) it is clear that breve is by metonymy for angustum.

We read: longe omnes multumque superabit (Cic. Verr. 2. 5. 115. 44), dicendi consuetudo longe et multum isti vestrae exercitationi ad honorem antecellet (Mur. 29. 13), cum longe multumque praestet mens atque ratio (Fin. 5. 40. 14). Servius (ad Aen. 5. 406: longeque recusat) says longe is for valde, and he is confirmed by Nonius (1. 545. 21, M.). Here longe would seem to be for longe et multum. Servius refers us to longe (Aen. 1. 13), which he says is for 166 valde, which Charisius confirms (203. 17, K.); and whether we connect longe with contra or dives, it seems to require that meaning. To: errat longe (Ter. Ad. 65) Donatus notes: Melius est ‘longe’ simpliciter accipere, ut locale sit, quam valde. But in: longe mihi alia mens est (Sall. Cat. 52. 2), where Servius says longe is for valde, he seems right. So in: canitiemque sibi et longos promiserat annos (Aen. 10. 549) longos is short for multos et longos, though Servius says it is for multos. In: longe servet vestigia coniunx (Aen. 2. 711) Servius says longe is for valde; it cannot mean ‘from a great distance’, for we read: pone subit coniunx (v. 725) ‘my wife follows close behind’. Evidently longe is here intended as a caution against her straying: ‘let her observe closely the way I go’. So in:

miserere parentis

Longaevi, quem nunc maestum patria Ardea longe

Dividit (Aen. 12. 43-5),

Servius takes longe as valde, joining it with maestum; we must not join it with dividit, he says; for Ardea is not far from Laurentum. We may notice in support of this, that dividit is in the next verse, while maestum and longe are the first and last words of the second colon of the verse, words often in close union in sense. In: multum inter se distant istae facultates, longeque sunt diversae (Cic. de Orat. 1. 215. 49) we have multum et longe distributed, i.e. longe in the second clause and multum in the first are both for longe et multum, as I shall show.

When we turn to comparatives we find often multum joined with them, evidently for longe et multum, as we shall see. We read: non multum est maius quam illud volgare ac forense (Cic. de Orat. 3. 92. 24), where manuscripts favour multum, hercle qui multum improbiores sunt quam a primo credidi (Pl. Most. 824), multum ad agendum difficiliorem (Quint. 9. 2. 68), multum est tersior (id. 10. 1. 94), multum graviora tulisti (Ov. Trist. 5. 11. 7), multumque coitur humani generis maiore in proelia damno (Luc. 2. 225), multum felicior exit (Stat. Theb. 6. 701), multumque aliis iactantior umbris (ib. 9. 559), multum uno maiora viro (Sil. 13. 708), multum hic robustior illo (Juv. 10. 197). In many of these examples some manuscripts read multo, the usual construction in classical Latin, but for that very reason multum, where it has good manuscript authority, should be retained.

But we find longe also with the comparative, as in: pedibus longe melior Lycus (Aen. 9. 556), longe cunctis longeque beatior illa (Ov. Met. 167 4. 325), qui sum longe fortior (Phaedr. 3. 7. 6), uno (proelio) longe magis Pompeianis prospero (Vell. 2. 51 fin.), longe quam speraverat tumultuosiorem (id. 2. 74 init.), sed mihi longe magis orator probari in opere suo videtur (Quint. 10. 1. 70), utiliorem longe fore Euripidem (id. 10. 1. 67), longe maiore nisu clamavit (Petron. 9), Giton longe blandior quam ego (ib. 98), quod longe melius historici faciunt (id. 118). We can see from these examples that of the pair longe multumque multum was often taken for the comparative, being usually, however, changed to multo, the ablative of difference of measure, as fitting more closely with it. But just as we saw with longe lateque, at times it is longe, and not multum, that is retained with the comparative; and we even find this longe coupled with late in this connexion, as in: longe lateque rem meliorem facit (Dig. 4. 4. 39. 1).

With the superlative longe is usual in classical Latin. In: longi pars maxima luctus (Aen. 11. 214) Servius prefers longe to longi; our editors are right in keeping longi as the more difficult reading, but it seems elliptical for longi pars longe maxima luctus, so that Servius was right about the meaning. Multo is also used with the superlative, and pretty frequently, as in: conspectus vester multo iucundissimus (Cic. Leg. Man. 1. 1), multo omnium nunc me fortunatissimum factum puto esse (Ter. Heaut. 842), multo sopor ille gravissimus exstat (Lucr. 4. 956), postera lux oritur multo gratissima (Sat. 1. 5. 39), multo pars maxima (ib. 2. 3. 82), ea regio eis temporibus multo putabatur locupletissima (Nep. Ages. 3. 1), multo elegantissimum poetam (id. Att. 12. 4). Hand says multum is found with the superlative once only, in: mater odorati multum pulcerrima turis (Grat. Cyn. 1. 133). Again we may trace this alternation of longe and multo with the superlative to the pair longe et multum.

We read in Cicero: multis meis et magnis laboribus et periculis (Sulla 5. 2), utilitates multae et magnae (Lael. 30. 9), accedunt eodem multa privata, magna eius in me merita (Phil. 13. 7. 4), atque haec in bello plura et maiora videntur timentibus (Div. 2. 58. 27), plurima et maxima proelia commemorare possem (Mur. 33. 16), iudicio plurimis maximisque in rebus probatissimus (Verr. 2. 2. 102. 42), ex plurimarum et maximarum appetitione concluditur (Fin. 4. 34. 13), plurimis et maximis voluptatibus (ib. 2. 63. 19). We have here, too, parallel to the use of multum (or multo) and longe with the comparative and superlative, the use with interest of magni, pluris (rarely maioris), and maximi or plurimi. In: tua plurima . . . pietas (Aen. 2. 429) plurima is probably by metonymy for summa, and in: culti iugera magna soli 168 (Tib. 1. 1. 2) magna seems to be for multa et magna. For: at tibi curarum milia quanta dabit (Prop. 1. 5. 10) reminds us of: qui ab dis immortalibus tot et tantas res tacitus auderet optare, quot et quantas di immortales ad Cn. Pompeium detulerunt (Leg. Man. 48. 16). In prose we read magna pars, but in Horace: multaque pars mei (Od. 3. 30. 6); and so probably with: multa . . . aura (Od. 4. 2. 25). So magno pretio in prose, and magna pecunia (Cic. Att. 11. 3. 3), but: multa mercede (Geo. 2. 62). We all remember Iuppiter optimus maximus; in: vos haec facietis maxima Gallo (Buc. 10. 72) maxima is a metonymy for optima.

We read in prose: magnae virtutis (B. G. 2. 15), but: multa viri virtus (Aen. 4. 3), probably for: magnae et multae viri virtutes; and: magnam laudem (Cic. Off. 2. 45. 13), but: non sine multa laude (A. P. 281-2), for: multis et magnis laudibus. So in: magnaque illic imago tristium laetorumque (Tac. Ann. 2. 53. 3), where magna seems for magna et alta; ‘many and profound were his reflections on victory and disaster’. For the use of magnus for altus and the opposite is not uncommon. To: iacet altus Orodes (Aen. 10. 737) Servius notes: altus magnus, ut: sic pater ille deum faciat, sic altus Apollo (10. 875), with which compare: eris mihi magnus Apollo (Buc. 3. 104). Probably Virgil substituted altus for magnus here to attain the oxymoron: iacet altus. We may further note: magnis montibus (Catull. 64. 280-1), magnum . . . Olympum (Enn. Ann. 1, M.), magna templa caelitum (Varr. L. L. 7. 2. 81), magna eloquentia (Cic. Tusc. 1. 117. 49), magnam laudem (Off. 2. 45. 13). It is for this reason that the superlative of magna in the last three examples is usually not maxima, but summa. Probably magnus is thus used for altus in: mari magno (Enn. Ann. 491, M.), aquae magnae bis eo anno fuerunt (Liv. 24. 9. 6); and we may compare: magna . . . voce (Sat. 1. 7. 31) with summa voce (1. 3. 7-8) and with altissimos sonos (Quint, 11. 3. 23). We find the union from which this confusion proceeds in: magni cuiusdam et alti viri (Cic. Tusc. 5. 31. 10). For parvus and paucus I have noted only: in parvo tempore (Lucr. 5. 106) for brevi; and: responsum paucis ita reddidit heros (Aen. 6. 672), out of which may arise Hyginus’s: paucum tempus.

Very common is the pair we find in: exitus quidem omnium unus et idem fuit (Cic. Div. 2. 97. 47), in qua omnes sentirent unum atque idem (Cat. 4. 14. 7), ferar unus et idem (Hor. Ep. 2. 2. 200), haec ut cera liquescit uno eodemque igni (Buc. 8. 80-1), una eademque via (Aen. 10. 487), uno eodemque tulit partu (12. 847). We have the pair 169 distributed in: non semper idem floribus est honor vernis, neque uno Luna rubens nitet voltu (Od. 2. 11. 9-11), where idem is for unus et idem, and uno for uno et eodem. It is much easier to trace the expression of the pair by unus than by idem; we note this in: quia quasi una aetas erat (Pl. Capt. 20), nam parentum iniuriae unius modi sunt ferme (Ter. Heaut. 205), illa cum uno tempore audisset (Cluent. 28. 9), ille sinistrorsum, hic dextrorsum abit; unus utrique error (Sat. 2. 3. 50), rege incolumi mens omnibus una est (Geo. 4. 212), aestuat ingens uno in corde pudor mixtoque insania luctu (Aen. 10. 871). Harder to distinguish is the use of idem for unus et idem, but it seems evident in: idem omnium gemitus (Tac. Ann. 3. 1. 5). In: his amor unus erat (Aen. 9. 182) the ellipsis is involved with the idea of reciprocity, and we may solve it: hi inter se uno et eodem amore fovebant.

We read: ille volat simul arva fuga simul aequora verrens (Geo. 3. 201) and: cum simul terra simul mari bellum impellitur (Tac. Agr. 25) ‘alike by land, alike by sea’; for simul seems a shortened form of simile. In these examples we have the older form of the construction, and at times one of the pair is omitted, either the first, as in: salve, simul autem vale (Pl. Merc. 830), or the second, as in: Cymothoe simul et Triton (simul) (Aen. 1. 144). So in: (adpetitus animi) et oderit se et simul diliget (Cic. Fin. 5. 28. 10) the first is omitted, in: simul aliquid audiero, scribam ad te (Att. 8. 11 fin.) the second. But when the pair is coupled as simul atque simul, or simul et simul, in the Latin we know usually either the first or second is omitted, as in: ut, simul atque posita sit causa, habeant quo se referant (Cic. de Orat. 2. 117. 27). We know that in dissolving a complex sentence into its components we usually change their order; so we have here: habeant quo se referant simul; atque simul posita causa est.

We know that along with: simul atque simul ‘alike, thereto somehow alike’ there was also in use: simul et simul ‘alike, further alike’, but the latter is rarer. We have it in: ego hic esse et illic simitu haud potui (Pl. Most. 792), nunc operam potestis ambo mihi dare et vobis simul (id. Men. 1099). We read in Tacitus: privatam gratiam statim mereare, statim recipias (Ann. 1. 28. 7), and recognize in statim . . . statim a pair like simul . . . simul, and interchanged as we see in: simul accepi a Seleuco tuo litteras statim quaesivi e Balbo per codicillos, quid esset in lege (Cic. Fam. 6. 18 init.). I read: semen statim cum spargitur, statim obruendum est (Pallad. Apr. 3. 3) 170 I find too: (simul) cum simul ‘(alike) what time alike’ in: ad portum hinc abii mane cum luci simul (Pl. Merc. 255), which seems short for: ad portum hinc abii mane simul, cum luci simul sol oriebatur. We have here the conjunction cum in juxtaposition with the locative luci, ‘when at dawn’. But in Latin the locative is fused very early with the ablative; and we have here a starting-point for the use of cum as a preposition, ‘when at dawn’ passing into ‘along with dawn’; just as we have it in: intro abi tu cum istac simul (Pl. Cist. 770), where istac passed from the nom. ‘when she’ to the ablative ‘with her’. And so in: quas (res) tecum simul didici (Cic. Acad. Post. 1. 3. 1) for; quas res tu (simul didicisti) cum simul didici. Of course in like fashion with simul we must treat its opposite: secus . . . secus.

It is easy to substitute ut ‘when’ for cum in Latin, and so for simul cum we get simul ut, as in: omne animal, simul ut ortum est, se ipsum diligit (Cic. Fin. 2. 33. 11). Cum and ut are both joined with primum; and it is not strange to find simul and simul atque also joined with it, as in: simul ac primum ei occasio visa est (Cic. Verr. 2. 1. 34. 13), simul primum magistratu abiit (Liv. 6. 1. 6), unde simul primum me dimisere Philippi (Hor. Ep. 2. 2. 49). It would be natural to expect simul ubi too; and when the best manuscripts give us: simul ubi conspexit (Liv. 4. 18. 7), though it seems a ἅπαξ γεγραμμένον, I feel like adopting the reading.

But the omission of the second simul has given rise to the idea that in: quo simul mearis (Od. 1. 4. 17) simul is for simul atque. But it is rather an ellipse such as we see in: simul inflavit tibicen, a perito carmen (simul) agnoscitur (Cic. Acad. Prior. 2. 86. 27), or: hic simul argentum repperit, cura sese (simul) expedivit (Ter. Phorm. 823), or: nostri simul in arido constiterunt, in hostes impetum (simul) fecerunt (B. G. 4. 26). Neither Servius, nor Acron, nor Porphyrio has thought it necessary to explain the syntax.

We have aeque . . . aeque in: aeque pauperibus prodest, locupletibus aeque (Hor. Ep. 1. 1. 25), aeque discordiam praepositorum, aeque concordiam subiectis exitiosam (Tac. Agr. 15). We find aeque et in: nisi aeque amicos et nosmet ipsos diligamus (Cic. Fin. 1. 67. 20); and aeque . . . que in: (quod) aeque neglectum pueris senibusque nocebit (Hor. Ep, 1. 1. 26). Like simul it is joined with atque, cum, and ut, and owing to its close connexion with the comparative it is also joined with quam, as in: nihil aeque eos terruit quam praeter spem robur et colos imperatoris (Liv. 28. 26. 14). Just as in: quae 171 (amicitia) incepta (simul) a parvis cum aetate adcrevit simul (Ter. And. 539) the pair simul . . . simul is represented by the second term, so in: (et) haec amem necesse est et Veraniolum meum (Catull. 12. 16-17) et seems for et . . . et, and assumes the force of atque ‘as’ in translation. The Itali changed this et to ut, which change Baehrens and Robinson Ellis were inclined to follow.

We read: nam qualis quantusque cavo Polyphemus in antro (Aen. 3. 641) and: nam aut ipsius rei natura qualis et quanta sit quaerimus (Cic. Tusc. 3. 56. 23). We find: quali fide, quali pietate existimatis esse eos? (Cic. Font. 21. 10), but: qui tanta virtute atque integritate fuit (ib. 29. 13); and: haud equidem tali me dignor honore (Aen. 1. 335), but: deae donis et tanto laetus honore (8. 617). In: ac primum quanta innocentia debent esse imperatores, quanta deinde in omnibus rebus temperantia, quanta fide, quanta facilitate, quanto ingenio, quanta humanitate! quae breviter qualia sint in Cn. Pompeio consideremus (Cic. Leg. Man. 36. 13) we shall best understand the meanings of quanta and qualia, if we think of quanta as short for quali et quanta, and of qualia as for qualia et quanta. So in: Poenorum qualis in arvis saucius ille gravi venantum volnere pectus tum demum movet arma leo (Aen. 12. 4-6) and in: qualis Hyperboreis Aquilo cum densus ab oris incubuit (Geo. 3. 196), while in: qualis mugitus, fugit cum saucius aram taurus (Aen. 2. 223) and in: qualis ubi ad terras abrupto sidere nimbus it mare per medium (12. 451) qualis seems a metonymy for quantus. We read: conservare urbes tantas atque tales (Cic. N. D. 3. 92. 39); and see that in: Eupolin Archilochum, comites educere tantos (Sat. 2. 3. 12) tantos is short for tantos et tales. So with tales in: si duo praeterea tales Idaea tulisset terra viros (Aen. 11. 285). We have the pairs in distribution in: qui tanti talem genuere parentes? (1. 606).

We read in Cicero: quam brevi tempore quot et quanti poetae . . . exstiterunt! (Tusc. 4. 5. 2) and: ad horum omnium iudicia tot atque tanta (Pis. 97. 40). We have already cited: tot et tantas res . . . quot et quantas (Leg. Man. 48. 16). So in: quis umquam tantis opibus, tantis rebus gestis fuit, qui se populi Romani . . . patronum dicere auderet? (Phil. 6. 12. 5) and: heu quianam tanti cinxerunt aethera nimbi? (Aen. 5. 13) tanti seems for tot et tanti. Still more frequent are such uses of quanti, as in:

Quantae tum scindunt hominem cuppedinis acres

Sollicitum curae, quantique perinde timores! (Lucr. 5. 45-6).

We read in Statius: O quantae pariter manus laborant! (Silv. 4. 3. 49) 172 and: et quantos (annos) ego Delium poposci! (ib. 152); and in Valerius Flaccus; heu socii quantis complerunt litora monstris! (3. 261), and in distribution: quot mihi post lacrimas, post quanta piacula patrum serus ades! (2. 563), where quanta is for quot et quanta. In: at tibi curarum milia quanta dabit! (Prop. 1. 5. 10) and: quanti tum iuvenes, quantae sprevere pudorem spectandi studio matres! (Claudian, de III Cons. Hon. 126) quanti is for quot, a metonymy evidently resulting from such uses as we cited above. In Sidonius Apollinaris we read in distribution: suffragia tot sunt, quanta legit mundus (Carm. 2. 21-2).

This use seems to have found a further support in: et quantum est hominum venustiorum (Catull. 3. 2), where we have for quot a collective quantum. This is common in the Comedy, as in: o mihi quantum est hominum optimorum optime (Pl. Capt. 836) and: o omnium quantum est qui vivont homo hominum ornatissime (Ter. Phorm. 853). Propertius joins it with tam multa in:

Tam multa illa meo divisa est milia lecto

Quantum Hypanis Veneto dissidet Eridano (1. 12. 3-4).

We have the opposite in Lactantius: vocavit discipulos quaerens quantos secum cibos gestarent (4. 15. 17), where quantos cibos is for quantum cibi. In Tertullian we have: nec tamen tantos inveniunt verba discipulos, quantos Christiani factis docendo (Apol. 50 fin.), where we have quanti for quot; and tanti for tot in: si Tiberis ascendit in moenia . . . statim: Christianos ad leonem! Tantos ad unum? (ib. 40 init.). And I read in Ducange that in mediaeval Latin the pair tanti . . . quanti is used for tot . . . quot; quot is found only in such unions as quot diebus, quot mensibus, while tot is nowhere used.

In French tant is the regular word for ‘so many’. While the ordinal is quantième and quantes is joined with fois at times for ‘how many times’, ‘how many’ is combien (= quam bene), But combien is often ‘how much’, and its use for ‘how many’ is probably parallel to the development from quantum of quanti for quot. The use of bene in: bene doctum leads quite naturally to its use in: bene nummatum (Hor. Ep. 1. 6. 38), equivalent to multis nummis praeditum. In this development it finds support in a parallel use of bonus. We recall: bonam atque magnam cenam (Catull. 13. 3), and: nam hic quoque bonam magnamque partem ad te adtulit (Ter. Eun. 123), sed bona magnaque pars servabat foedera casti (Lucr. 5. 1025), equestris 173 quoque ordinis bona magnaque pars (Val. Max. 2. 9. 7). For this pair we have magnam in: cum . . . magnam partem noctis vigilasses (Cic. Div. 1. 59. 28), but bonam in: bonam partem sermonis (de Orat. 2. 14. 3); and bonam spem (Lael. 23. 7), but: cum spe magna (Rab. Post. 5. 2) and: sine magna spe (Tusc. 1. 32. 15). So too in: bona pars hominum (Hor. Sat. 1. 1. 61), vocis accedet bona pars (Od. 4. 2. 46), melior quoniam pars acta diei (Aen. 9. 156).



I had thought of treating synecdoche and metonymy in separate chapters; but I find it impossible. Harkness in his school grammar defines synecdoche as the use of the part for the whole, or of the whole for a part, or of the special for the general, or of the general for the special. Metonymy, he says, is the use of one name for another naturally suggested by it; and he adds: by this figure the cause is often put for the effect, or the effect for the cause, the property for the possessor, the place or age for the people, the sign for the thing signified, &c. But I find that synecdoche is only another name for the expression of a pair by one of its terms, and that metonymy, the expression of one of the terms by the other, is an almost constant consequence of synecdoche. When we name the Dioscuri by Pollux in Od. 3. 29. 64, we have synecdoche; when we name Castor by Pollux in Geo. 3. 89, we have metonymy. The steps by which metonymy issues from synecdoche are not wholly clear to me; but the use of Castores for Castor and Pollux evidently assumes that we may call Pollux Castor.

My experience is that the two figures are often confused. Let us turn to: contigimusque manum, qua concidit Ilia tellus (Aen. 11. 245) ‘the hand, by which the land of Troy fell’. We should at once say, we have here a metonymy of tellus for urbs. Servius’s note is: ἐμφατικῶς dixit pro ‘urbs Ilia’, nam terra non concidit, sed civitas Ilium. But behind this use of tellus for urbs, we probably have terra et urbs parallel to Virgil’s: urbes arvaque (Aen. 1. 549-50). If this is right, then we have here tellus for the urbs et tellus which constitute the civitas: a synecdoche, not a metonymy. But we may imagine a student, not versed in the use of tropes, who would emend tellus to turris, just as Ribbeck, influenced by a note of Servius, changed arva to arma. Servius’s note is to urbes, and is: arma latenter minatur.

We find Bentley emending in like fashion, and with even less justification, through a failure to understand such figures. Horace writes:

Tutus bos etenim rura perambulat;

Nutrit rura Ceres almaque Faustitas (Od. 4. 5. 17-18).


What grace or favour, he asks, has the repetition rura . . . rura? is it not rus merum? Already Tanaquil Faber, from a like feeling, had changed rura to prata in v. 17. But there the sense is plain and sound. The sore spot is in: nutrit rura Ceres. Quid enim, amabo, est ‘nutrire rura’? Rura sunt arva, prata, campi, saltus, &c., pro varia terrarum forma et situ. Quis vero mortalium nutriri dixit res eiusmodi? One can appreciate here the prosaic common sense of Bentley in its hopeless task of reading a poet, even so careful and discreet a poet as Horace. Even by trope can anything be said to be nursed, he asks, except what is capable of growth and increase, as e.g. trees, crops, fruits, or as hate, love, war, fire, and the like? Just fancy, I might add, the farmer nursing his two-acre field, in the hope that it will grow to a hundred acres! Quid multa? In rura there is an erasure in the codex Graevii, above which ru- appears. Rura is evidently the work of a second hand; and Bentley has little doubt that Horace wrote here farra, not rura; and in his text I read: nutrit farra Ceres. Neither Acron nor Porphyrio explain rura, but Porphyrio indicates the text thus: nutrit r. C. a. Faustitas; and from r. we may assume that he read rura, not farra. But, as we shall see, rura here may well be a poetic shortening for the Ciceronian terrae fruges, and so a metonymy.

Gellius (18. 5) relates how during a summer holiday at Puteoli, in company with a rhetor, Antonius Julianus, he went to hear an ἀναγνώστης give readings from Ennius in the theatre. After the reading, as Julianus was leaving the theatre, he expressed his opinion that if the reader had had a teacher that cost him anything, he would not have read:

Denique vi magna quadrupes ecus atque elephanti

Proiciunt sese (Enn. Ann. 249-50, M.)

but quadrupes eques, as Ennius wrote. And to the wondering questions of some standing by, what quadrupes eques could be? he replied: I could wish that you had read Ennius with care as did Virgil, who imitating this verse of Ennius wrote equitem for equum in:

Frena Pelethronii Lapithae gyrosque dedere

Impositi dorso atque equitem docuere sub armis

Insultare solo et gressus glomerare superbos (Geo. 3. 115-17).

Furthermore, that equitare is used both of the rider and of the horse as he steps beneath the rider, is clear from:

Quis hunc currere ecum nos atque equitare videmus,

His equitat curritque; oculis equitare videmus,

Ergo oculis equitat (Lucil. Inc. 70-2, M.).


Not content with these confirmations of Ennius’s usage Julianus went on to examine a copy of his poem, Lampadionis manu emendatum, and found eques, not ecus, in the passage in question.

Macrobius too (Sat. 6. 9. 8-11) quotes the verses in the same words, and in a discussion of Virgil’s verses he makes Servius say: ‘You have this question arising from your disregard of the old style of speech. For since our age has revolted from Ennius and the old writers, we are ignorant of much that would not escape us, if the style of the ancients were familiar to us. For the old writers, just as they called the man seated on the horse eques, so they gave the name eques to the horse that bore the man’. Probably there are no verses of the ancients whose text is better accredited than the verses of Ennius and of Virgil quoted by Gellius.

Of the verses of Virgil Julianus says (Gell. l.c.): in quo loco ‘equitem’, si quis modo non inscite inepteque argutior sit, nihil potest accipi aliud nisi ‘equum’. It is a censure that falls upon all our modern editors since Heyne, who notes: cur eques non insultare dicatur, qui et incedere, decurrere. The rider is said to advance, who causes his horse to advance; why should he not be said to paw the ground, who causes his horse to do so? Thrice and four times happy editors! who understand Virgil so much better than did the Romans themselves. But they have not ventured on an explanation of Ennius’s quadrupes eques. I should add Servius’s note to Virgil’s equitem: equitem equum: pro equo rectorem posuit. Et aliter: hic ‘equitem’ sine dubio equum dicit, maxime cum infert ‘insultare solo’, et sub armis id est insidente armato. Could anything be more explicit? Virgil calls the horse ‘equitem’ and the armed rider ‘arma’.

Let us compare Conington: ‘An old gloss . . . gave equitem the sense of equum on the strength of a doubtful passage in Ennius (Ann. 7. Fr. 9), an anomaly which, if justified, would only produce a platitude. Here, as in Hor. Epod. 16. 12: “Eques sonante verberabit ungula”, the rider is evidently said to do what the horse does. So “sub armis” points to the weight on the horse’.

I question not merely the clearness of Conington’s note, but its truth of intention. Why does he speak of a doubtful passage in Ennius? It is attested by Gellius, Nonius, Macrobius, and Servius. And it is not the only passage where Ennius uses eques for equus. I read:

It eques et plausu cava concutit ungula terram (Enn. Ann. 485, M.),

‘the horse proceeds, and his hollow hoof shakes the earth with its 177 beat’, where the connexion favours the translation of eques as for equus. Let us turn to the verses cited by Conington from Horace:

Barbaras heu cineres insistet victor et urbem

Eques sonante verberabit ungula (Epod. 16. 11-12),

‘the victorious barbarian, alas! will trample the ashes, and the horse with sounding hoof will beat the city’. Is not he inepte argutior, who will translate here: ‘The horseman with sounding hoof, &c.’? About his inscitia we shall see in a moment.

The use of eques for equus need not have puzzled our English critics so much. We all know the old English use of horse for horsemen. Murray (Vol. V, p. 394) quotes Hall’s Chronicles: King Henry with a few horse in the night came to the Tower of London, and Robertson’s America 1. 157: The body consisted only of two hundred foot, twenty horse, and twenty Indians. And with this collective use is to be connected the use of ‘hoss’ in the Western States for cavalryman, as seen in the invitation ‘Step up this way, old hoss, and liquor’.

Where in English we use horse for horseman—the stranger use, it seems to me,—in Latin we have equitem for equum and equum for equitem. For eques and equus constitute a pair; and for this pair eques is very often used, especially in the plural equites, which is the cavalry and included equi as well. Livy always uses magister equitum; but I read in Ennius:

Vel tu dictator vel equorum equitumque magister

Esto vel consul (Ann. 346, M.).

We have here equorum equitumque, which Livy shortens to equitum, just as Horace shortened longus et latus sometimes to longus, sometimes to latus. So for this pair we have usually equites, but at times equi, as I shall show. Later in the singular comes eques for equus, and equus for eques. In the use of equites for equi et equites we have synecdoche, in that of eques for equus or vice versa, metonymy.

Equi is used for equites by Cicero in: cum his, viris equisque, ut dicitur, si honestatem tueri et retinere sententia est, decertandum est (Off. 3. 116. 33), fore ut omnes inflammati odio, excitati dolore, armis viris equis Dec. Bruto subveniamus (Phil. 8. 21. 7), nunc cum confecta sunt omnia, dubitandum non est, quin equis viris (obviam Caesari eamus) (Fam. 9. 7). Livy has: quacumque ibant equis virisque longe ac late fuso agmine inmensum obtinentes loci (5. 37. 5); Nepos gives 178 us: equis, armis, viris, pecunia totam locupletavit Africam (Ham. 4), and Florus: aderant Rhodii, nauticus populus, qui navibus a mari, consul a terris omnia equis virisque quatiebat (2. 7. 8). Equis virisque is evidently a colloquial expression for ‘with horse and foot’, felt as equivalent to summis viribus, and in it equi is for equi et equites. Parallel to it is the use of equites virique for horse and foot in: terrebant ex adverso hostes omnem ripam equites virique obtinentes (Liv. 21. 27. 1), the common use of equites.

Virgil gives us: utque acres concussit equos, utque impulit arma (Aen. 8. 3) where Mackail translates: ‘when he spurred his fiery steeds, and clashed his armour’. But this hardly seems in agreement with the preceding or following verses; and Servius’s note is: hoc ad equites pertinet. The pair: equi virique occurs repeatedly; did Virgil shorten this to equi? We shall show that arma is often for bellum; does Virgil intend a second meaning here ‘he thrilled horse and foot to keen ardour, and urged on war’?, a verse now quite in harmony with the context. Moreover in: semper equos atque arma virum pugnasque canebat (Aen. 9. 777), where the obvious translation is: ‘ever he sang of steeds, of arms of heroes, and of battles’, following the uses above we should have as a second meaning: ‘ever he sang of armies, of the wars of heroes and their battles’, a statement of the epic theme far more in consonance with the verses of Homer and Virgil.

In the mouth of one of his favourite soldiers of the tenth legion Caesar puts the words: plus quam pollicitus esset Caesarem facere: pollicitum se in cohortis praetoriae loco decimam legionem habiturum ad equum rescribere (B. G. 1. 42 fin.) ‘he was enrolling them in the cavalry’. The phrase seems a bit of soldiers’ slang, and is explained if we take equum as collective for equos used as above for equites. In: equites sequi iubet sese iterque accelerat, . . . At illi itinere totius noctis confecti subsequi non poterant (B. C. 2. 39 fin.), while equites seems for horsemen, illi (equites) seems rather for the horses. In: testes equestrium fratrum in lacu, sicut ostenderant, statuae consecratae, qui anhelis spumantibus equis atque fumantibus de Perse victoriam eadem die, qua fecerant, nuntiaverunt (Min. Fel. 7. 3) editors have corrected anhelis to anheli, as being naturally of the riders, while spumantibus was of the steeds. But the manuscripts give anhelis; and probably the phrase anhelis spumantibus equis is an example of the use of three for four, being short for: anhelis equitibus spumantibus equis, a shortening that becomes easy when equites is used for equi.


There can be no doubt that equus is used for eques in Horace’s verses:

Sive quos Elea domum reducit

Palma caelestes pugilemve equumve

Dicit (Od. 4. 2. 17-19),

where Porphyrio’s scholium begins: utrum Castorem et Pollucem significat, an generaliter. Of course Pollux is pugil, but Castor is eques, not equus. Then in Propertius we have:

Ite agite, expertae bello date lintea prorae

Et solitum armigeri ducite munus equi (3. 4. 7-8),

where his armigeri equi, like Virgil’s bipes equus (Geo. 4. 389), makes it clear that he has Ennius’s quadrupes eques in mind. Equi is evidently collective here for equitum, ‘lead the wonted service of the armed horse’. But in v. 17 of the same poem we have:

Tela fugacis equi et bracati militis arcus,

where equi seems for equitis. We have also in his elegies:

Quot sine aqua Parthus milia currat equus (4. 3. 36),

where Scaliger corrected equus to eques, relying on a passage of Dio Cassius which tells of the dry climate of the Parthians and how they could ride many miles without stopping to slake their thirst. Most scholars of the seventeenth and eighteenth centuries adopted his emendation; and Burmann thought he had read eques in the Codex Livineius. Lachmann showed this was not so, and rightly restored equus, the manuscript reading; for equus rhymes with Parthus. But if equus is used for eques, all difficulty vanishes, and emendation is unnecessary.

Gellius was a friend of Fronto and of Herodes Atticus; and it was probably not long after A.D. 150 that he had the adventure at Puteoli of which he speaks. Macrobius was consul in Africa in A.D. 400. It seems clear that the literati at both these dates regarded the use of eques for equus as a curious and antique literary idiom that in their day had passed out of popular use, and was known only to scholars. It is true that we read in Martial: equitis . . . culus aheni (11. 21. 1), but Martial is a scholar and a literary virtuoso. If this idiom were really remote from popular use from the second to the fifth century, it is curious to find it flourishing in popular use in the sixth. I suppose the solution is that the literary circles during these centuries paid little or no attention to popular usage of Latin.

Gregory of Tours brings his story of the Franks down to the year A.D. 591; and its Latinity seems to justify Dr. Pfister’s statement that he wrote in the vernacular of his day. In the third book of his Historia Francorum, in the latter part of its tenth chapter, I note that 180 he uses: assumptis equitibus—stratis equitibus—sumptis reliquis equitibus secum—relictis equitibus—tum motis equitibus discesserunt: in these examples, five in all, eques is plainly for equus. In: audiunt pedibulum equitum currentium it may be claimed that equitum is for horses and riders; when they heard the tramp they suspected this, but did not know. We have also in the chapter: custos equorum [bis]—cum equos ad claudendum adduxeris—deducat equos—cum equi urinam proiicerent—five uses of equi in its proper sense. I noted also: ascensis equitibus (ch. 18); and Bonnet (p. 284) gives me: misit pueros suos cum equitibus et plaustris (2. 24), where Migne reads equis, and ascenso equite (2. 12), He calls the use a synecdoche, but it is rather a metonymy. He tells me (p. 740. 1) that Gregory uses five terms for ‘horse’: equus, eques, caballus, sonipes, and cornipes; that he uses equus sixteen times, and that his uses of eques and caballus together come to a little more (p. 205. 3). We have evidently here an idiom not confined to literary Latin, but usual in the popular Latin of over eight centuries, from Ennius to Gregory of Tours. While the literary expert of Puteoli wondered at Ennius’s quadrupes eques, probably every muleteer in the street would have understood it.

Parallel to equi et equites seems currus et equi, a pair we find in: sustineas currum ut bonus saepe agitator equosque (Lucil. inc. 155, M.), currus et quattuor ausus iungere equos (Geo. 3. 113-14), diversos ubi sentit equos currumque referri (Aen. 12. 495), currumque et equos et lora regebat (ib. 624), egit equos volucremque currum (Hor. Od. 1. 34. 8), iret alter consul sublimis curru multiiugis, si vellet, equis (Liv. 28. 9. 15). We have the terms distributed in: frustra retinacula tendens fertur equis auriga, neque audit currus habenas (Geo. 1. 513-14), and in Aen. 12. 350-2, where currus in v. 350 and equis in v. 352 plainly stand for the pair. Equi is for currus et equi in: cum (sol) invectus equis altum petit aethera (Geo. 3. 358) and: quattuor hic invectus equis (Aen. 6. 587) et saepius. So too with equi for currus, as in: iunctos conscendebat equos (Aen. 12. 735-6). Following this use Horace gives us: impositus mannis (Ep. 1. 7. 77) ‘riding in a gig’; and in v. 98, following this use, he has pes as the diminutive of decempeda. We have currus for currus et equi in: domitantque in pulvere currus (Aen. 7. 163), currus effreno impetu effugit aciem (Sen. Agam. 944), and currus for equi in: infrenant alii currus (Aen. 12. 287), si verbere saevo Palladia stimulet turbatos aegide currus (Luc. 7. 569-70), et proculcantes moderantum funera currus (Stat. Theb. 10. 741), 181 stimulabat in aequore currum (Sil. 16. 366). From the use of currus for equi, it is easy to explain the use of currūs for a single chariot, found in three of the four examples quoted; it is extended to other uses of currus, and thence to bigae, trigae, quadrigae. In Greek ἵπποι is used for the chariot; perhaps the plurals ὄχεα and ἅρματα must be treated in the same way.

In dealing with the language of a people like the Romans, a pair of words like bellum et arma will attract notice. We find it very often, as in: non belli atque armorum fuerunt (Cic. Marcell. 14. 5), nec bello maior et armis (Aen. 1. 545), ad bellum atque arma incitantur (Liv. 1. 27. 3), belloque et armis (Tac. Hist. 4. 52. 3), bello cogendus et armis (Stat. Theb. 12. 165). In his note to the first words of the Aeneid: arma virumque cano, Servius tells us: per arma autem bellum significat, et est tropus metonymia. It is hardly necessary to multiply examples of this trope, common in prose as in verse. We have synecdoche in: silent leges inter arma (Cic. Mil. 10. 4), Ennius . . . numquam nisi potus ad arma prosiluit dicenda (Hor. Ep. 1. 19. 7), ab externis armis otium fuit (Liv. 3. 14. 1); metonymy in: nec post arma ulla rebelles Aeneadae referent (Aen. 12. 185), a fine Actiaci belli ad ea arma quis Servius Galba rerum adeptus est (Tac. Ann. 3. 55. 1). In return we have bellum or bella for arma in: positis . . . bellis (Aen. 1. 291, Hor. Ep. 2. 1. 93), poterat deponere bellum (Ov. Met. 8. 47), rapiendi tempora belli (Luc. 5. 409), spargatque per aequora bellum (id. 2. 682), ut . . . bellum pacis caritate deponerent (Tac. Hist. 2. 37), sumpsere universi bellum (Agr. 16), with which compare: nam bellum atque arma . . . sumenda sunt (Sall. Or. Phil. 2).

Arma is usually derived from the root ar-, to join or fit (cf. ἀραρίσκω), and primarily meant the defensive armour fitted to the body; so Varro derives it ab arcendo (L. L. 5. 115). Servius in a note to: arma viri (Aen. 4. 495), which he identifies with the ensem relictum of v. 507, says: arma gladium dicit abusive; proprie enim arma sunt quae armos tegunt, hoc est scutum; and he cites: at Lausum socii exanimem super arma ferebant (10. 841). But Festus, though he adopts the same derivation, says: arma proprie dicuntur ab armis, id est humeris, dependentia, ut scutum, gladius, pugio, sica: ut ea, quibus procul proeliamur, tela (dicuntur) (3, M.). We read in Cicero: arma esse nominibus suis, alia ad tegendum, alia ad nocendum (Caecin. 60. 21). If arma were primarily defensive, the name was extended later to weapons of attack as well. In: quaerere conscius arma (Aen. 2. 99) it seems to denote means of assailing Sinon as well as of defending 182 Ulysses. Servius (ad Aen. 8. 249): omni quod iaci potest telum vocatur ἀπὸ τοῦ τηλόθεν; and again (ad 9. 507): ostendit telum vocari omne quod iacitur. But in: at non hoc telum, mea quod vi dextera versat (9. 747) Virgil uses telum for ensem (cf. v. 749); so that we have here a further extension of the meaning of telum. We read: rex impius aptat tela (Stat. Theb. 11. 500). When we meet: arma ac tela (dicuntur) pro bello (Cic. de Orat. 3. 167. 42) we have at once the clue to this confusion, which arises from the use of arma or tela for arma ac tela. So we have arma used in: tum demum arma movet leo (Aen. 12. 6); and we have cornua used for arma (= bella) in: taurus . . . irasci in cornua tentat (12. 103-4), where with bella we supply futura, ‘he essays to rouse his wrath for wars to come’.

We read of Chloreus: spicula torquebat Lycio Gortynia cornu (Aen. 11. 773); but in v. 774: aureus ex umeris erat arcus. Had he two bows? or is the arcus hanging from his shoulders a metonymy for pharetra? for what is the use of a bow of gold? We find the union arcus et pharetra in: arcum et pharetram et sagittas sumpsero (Pl. Trin. 725) and: arcus plenaeque pharetrae (Ov. Pont. 1. 2. 83). When we compare: umeris de more habilem suspenderat arcum venatrix (Aen. 1. 318) with: virginibus Tyriis mos est gestare pharetram (v. 336), we may take arcum as a metonymy for pharetram in v. 318, or take both as by synecdoche for arcum et pharetram.

Very usual is the union auro et ostro, as in: regali conspectus in auro nuper et ostro (A. P. 228), ostroque insignis et auro (Aen. 4. 134), vestes auroque ostroque rigentes (11. 72), auro volitant ostroque superbi (12. 126). We have an older form in: emit auro et purpura (Pl. Most. 286), and: purpurei cristis iuvenes auroque corusci (Aen. 9. 163). Ostro seems for auro et ostro in: victor ego et Tyrio conspectus in ostro (Geo. 3. 17) and: ut regius ostro velet honos leves umeros (Aen. 7. 814); and auro for auro et ostro in: ut se ferret in auro (Aen. 11. 779).

In: orantes veniam (Aen. 1. 519) venia is ‘pardon’, but the Trojans had not wronged Dido or her people. Servius explains it as pacem, and adds: propter incendium navium; from v. 525 it is clear that the Tyrians were threatening to burn their ships. We have a second note from Servius: venia quidem pro culpa petitur, sed nunc beneficium; aliqui tamen veniam pro impunitate accipiunt, &c., from which it is plain that the ancients too found difficulty in ‘veniam’. But we have the pair in: pacem veniamque impetrare a victoribus (Liv. 37. 45. 7) and: pacem ac veniam peto (Cic. Rab. 5. 2). So we 183 may assume that veniam is here a metonymy for pacem, and that in: tu modo posce deos veniam (Aen. 4. 50) and: pacemque per aras exquirunt (v. 56) we have synecdoche for pacem et veniam. So in: veniamque rogantes (11. 101) and: pacem me exanimis . . . oratis? (v. 110), which describe the same act. In: votis precibusque iubent exposcere pacem (3. 261) probably pacem is a metonymy for veniam.

To: iustae quibus est Mezentius irae (Aen. 10. 716) Servius’s note is: quibus quasi odium est. Livy gives us: efferati odio iraque (5. 27. 10), giving me the clue to Virgil’s metonymy. I read: ut erat recens dolore et ira (Tac. Ann. 1. 41. 5) and: necdum etiam causae irarum saevique dolores exciderant animo (Aen. 1. 25). And so I understand that dolor in: duris dolor ossibus ardet (Aen. 9. 66) and in: illa furens acrique accensa dolore (11. 709) is for: dolor et ira. We have this pair distributed in: quos iustus in hostem fert dolor et merita accendit Mezentius ira (8. 500-1), where we have a double metonymy, of dolor for ira, and ira for dolor, parallel to that in: auras suspiciens hausit caelum (10. 898-9). In: necdum antiquum saturata dolorem (5. 608) we have a metonymy of dolorem for iram that recalls the association of ira with dolor and odium in Horace’s verses:

qui non moderabitur irae

Infectum volet esse dolor quod suaserit et mens,

Dum poenas odio per vim festinat inulto (Ep. 1. 2. 59-61)

where to prevent undue repetition he has brought in a new metonymy of mens for ira, following the analogy of θυμός. He uses the same metonymy in: compesce mentem (Od. 1. 16. 22).

But with amor, odium and ira is present another association, that of opposites. We recall Terence’s: amantium irae amoris integratiost (And. 555), Catullus’s: odi et amo (85. 1), Horace’s: oderunt peccare boni virtutis amore (Ep. 1. 16. 52), and Virgil’s: duri magno sed amore dolores polluto (Aen. 5. 5-6), with which we may compare Ovid’s: spreto totiens iratus amore (Met. 7. 375) and Seneca’s: sed magnus dolor iratus amor est (Her. Oet. 451-2). With these in view we understand the use of amor in: sic omnes amor unus habet decernere ferro (Aen. 12. 282), and the use of ira for amor by metonymy in: subit ira cadentem ulcisci patriam et sceleratas sumere poenas (2. 575-6), where ira seems to be: amor patriam ulciscendi ex ira ortus. In: et quisquis amores haud metuet dulces, haud experietur amaros (Buc. 3. 109-10) amaros (amores) would seem to be for curas amaras.

We meet another interesting case of association of opposites, when we compare: vitamque volunt pro laude pacisci (Aen. 5. 2 30) with: 184 letumque sinas pro laude pacisci (12. 49). Heyne explains: scil. pactione et damus aliquid et accipimus; utrumque pacisci designare potest; ideoque et vitam pacisci quis et mortem potest, illam reddendam, hanc ferendam. In: dum vernat sanguis, dum rugis integer annus (Prop. 4. 5. 59) annus seems used for aetas, and sanguis for vita, as it is in: quibus integer aevi sanguis (Aen. 2. 639-40) and: laudem ut cum sanguine penset (Ov. Met. 13. 192). We see the association of sanguis with vita in: cum (Epaminondas) una cum sanguine vitam effluere sentiret (Cic. Tusc. 2. 59. 24). Hence the metonymy of vita for sanguis in: est animus nobis effundere vitam (Ov. Her. 7. 181). But while sanguis seems for vita in: poenas cum sanguine poscunt (Aen. 2. 72), it is rather for mors in: nec soli poenas dant sanguine Teucri (2. 366).

Sanguis is still more closely associated with caedes, which we find used for it in: respersum iuvenem fraterna caede (Catull. 64. 181), alta tepefaciet permixta flumina caede (id. 360), semperque recenti caede tepebat humus (Aen. 8. 196), virgo caede madentes ultima caelestum terras Astraea reliquit (Ov. Met. 1. 149). So in: iamque aderit multo Priami de sanguine Pyrrhus (Aen. 2. 662) we have in: multo de sanguine a picturesque metonymy for: atroci de caede. Interesting is the union of caedes and sanguis in: pugnatum ingenti caede utrimque, plurimo sanguine (Liv. 2. 64. 5); when we compare it with: multa utrimque volnera, multa passim caedes est (id. 4. 28. 7) and with: multa utrimque cadunt, plures volnera accipiunt (id. 7. 8. 1), it seems that ingenti caede is for multis occisis, and plurimo sanguine for plurimis volneratis. While caedes primarily meant a blow or stroke, as we see in: ut ilex . . . per damna per caedes ab ipso ducit opes animumque ferro (Od. 4. 4. 57-60), as a rule caedes is equivalent to: ictus mortiferus, as we see it in: telumque in caede reliquit (Sil. 7. 614). While we read: pugnam caedesque petessit (Lucr. 3. 648) and: en age, miles, in pugnam et caedes (Sil. 15. 445), it is clear from: me sequimini ad caedem, non ad pugnam (Liv. 5. 44. 7) and: iam non pugna, sed caedes erat (Curt. 4. 15. 32) that the Romans admitted no confusion in a matter so vital; and we find that committere caedem (Ov. Her. 14. 59) is entirely different in meaning from committere pugnam.

From formosus we conclude that ‘beauty’ is the primary meaning of forma; the meaning it presents in: di tibi formam (dederunt) (Hor. Ep. 1. 4. 6), spretaeque iniuria formae (Aen. 1. 27), eximia forma pueros (Cic. Tusc. 5. 61. 21), (virgines) quasdam forma excellentes 185 (Liv. 1. 9. 11). How does it come to have the usual meaning? We find it in union with species in: speciem ac formam . . . gerit eius imago (Lucr. 4. 51), tum fingit formam quandam et speciem deorum (Cic. N. D. 1. 37. 14), quorum in adulescentia forma et species fuit liberalis (Cael. 3. 6), quanta religione fuerit eadem specie atque forma signum illud (Verr. 2. 4. 129. 58). In consequence we have forma taking on the meaning of species et forma, sometimes with an inclination to forma as in: formaque ante omnes pulcer Iulus (Aen. 5. 570), quarum quae forma pulcerrima Deiopea (1. 72). But sometimes it inclines to species, as in mortalem eripiam formam (9. 101) or: aspicite, o cives, senis Enni imaginis formam (Enn. Test. XLII, M.); and this class gives us the usual meaning of the word. Species too, primarily the look (*specio) or appearance, takes the meaning of species et forma, inclining to forma, in: est specie alia magis alia formosa et illustris (Cic. de Orat. 3. 55. 14). So too in: species auri frondentis (Aen. 6. 208), agro bene culto nihil potest esse . . . specie ornatius (Cic. Sen. 57. 16), speciem haberet honesti (Off. 3. 2. 7). And it seems used by metonymy for forma in: si fortunatum species et gratia praestat (Ep. 1. 6. 49) and: ducit te species (Sat. 2. 2. 35). Perhaps the most noteworthy metonymy of forma for species is that in use in formal logic. We read in Quintilian: speciem, quam eandem formam (Cicero) vocat (5. 10. 62).

Contrary to Varro’s view that facere comes from facies, we have every reason to think that facies comes from facere; it denotes the human face as prepared for human intercourse by washing, combing, shaving. Here we may notice the significance of the Latin word for ‘razor’, novaculum ‘the little renewer’; cf. faciemque novat (Ov. Met. 15. 255). We read: facies homini tantum, ceteris os aut rostra (Plin. N. H. 11. 37. 51), uretur facies, urentur sole capilli (Tib. 1. 9. 15), cura dabit faciem, facies neglecta peribit (Ov. Ars 3. 105). We have it paired with forma in: (metalla) in quamlibet formam et faciem decurrere rerum (Lucr. 5. 1263), conlaudato formam et faciem (Pl. Mil. 1027), quod tibi non facies solave forma dabit (Ov. Ars 2. 108), formam quidem ipsam . . . et tamquam faciem honesti vides (Cic. Off. 1. 15. 5). So we have facies for forma in its secondary sense in: nec pingues unam in faciem nascuntur olivae (Geo. 2. 85), curvata in montis faciem . . . unda (4. 361); and in its prior sense of ‘beauty’ in: insignis facie (Aen. 9. 336) and: non tibi Tyndaridis facies invisa Lacaenae (2. 601), where Servius explains facies as pulcritudo. In: forma quoque hinc sous debet filumque videri 186 (Lucr. 5. 573) forma seems to be for facies and filum for os ‘the edge’. When Varro tells us: proprio nomine dicitur facere a facie; qui rei quam facit imponit faciem (L. L. 6. 78), he has especially in mind the shaping of coins by stamps or moulds. In that case in: ex eis (silicibus) formae fiunt, in quibus aera funduntur (Plin. N. H. 36. 22. 49) forma is short for forma et facies, as seen in: formam et faciem honesti (v. supra). Moulding seems primarily connected with coins, as we see from: utendum . . . plane sermone ut nummo, cui publica forma est (Quint. 1. 6. 3), and the formae denariae there stamped are, of course, facies deorum imperatorumve.

Out of the unions species et forma and forma et facies may arise the use of facies for species noticed by Servius in: non ulla laborum . . . nova mi facies inopinave surgit (Aen. 6. 103-4). We have it likewise in: ne qua . . . hostilis facies occurrat (Aen. 3. 406-7), ad istam faciem est morbus qui me . . . macerat (Pl. Cist. 69), diversa omnium quae umquam accidere, civilium armorum facies (Tac. Ann. 1. 49. 1), with which compare: velut in urbe victa facies (ib. 1. 41. 1). Species is commonly opposed to res, as in: speciem prae te boni viri feras (Cic. Off. 2. 39. 11) or: scurrantis speciem praebere (Hor. Ep. 1. 18. 2), and facies takes its place in this sense too in: publici consili facie (Tac. Hist. 2. 54. 3) and: facie maioris vivere census (Juv. 7. 137). But I have found no such pair as facies et species. Perhaps the metonymy has arisen from the influence of the two pairs named above, as we are further led to conjecture from the following paragraph.

We find the pair in: nil extra numerum fecisse modumque (Hor. Ep. 1. 18. 59), and we have numerus for modus in; in numerumque exsultant (Lucr. 2. 631). We have the pair in: prodest quorum in locum ac numerum pervenire velis, ab eis ipsis illo loco ac numero dignum putari (Q. Cic. Petit. Cons. 4. 1), and numerus for locus in: hostium se habiturum numero confirmat (B. G. 6. 6). We have the pair in: hominem ornatissimum loco, ordine (Verr. 2. 1. 127. 48), and we have locus for ordo in one of the commonest metonymies, as in: summo nati loco (Cic. Cat. 4. 16. 8), summo loco adulescens (Mur. 73. 35). But I have numerus for ordo in: carmina virgo digerit in numerum (Aen. 3. 446); and in: numero beatorum eximit virtus (Od. 2. 2. 18) numero seems for mimero et ordine. But I have not found the pair numerus et ordo, though it has no formal reason against its use, such as we see in the rhyming species et facies. It may exist, but if not, the metonymy seems the result of 187 the complex of pairs: numerus et locus, locus et numerus, locus et ordo.

One of the most puzzling problems for the teacher in Latin syntax is the use of: in tempore for time and of: loco for place in direct opposition to the rule and to common usage with other temporal and local terms. Here we read: tempus et locus convenit (Liv. 1. 24. 2), tanta vis est et loci et temporis (Cic. Off. 1. 144. 40). Hence in: epistulae offendunt non loco redditae (Fam. 11. 16. 1) we may assume that loco is short for: tempore et loco and in: interea loci numquam quidquam facinus feci (Pl. Men. 446-7) loci is for temporis. And so in: tamen is ad id locorum talis vir (Sall. Jug. 63. 6). But in: in loco ego vero laudo (Ter. Heaut. 537) and: dulce est desipere in loco (Od. 4. 12. 28) in loco is short for: in loco et tempore; as is in tempore in: in tempore ad earn veni (Ter. Heaut. 364) and: ni . . . pedites equitesque in tempore subvenissent (Liv. 33. 5. 2). And just so in: quo cum consul ad tempus . . . venisset (38. 25. 3) ad tempus is short for ad locum et tempus, the union giving the idea we convey by ‘fitting’. We read: apis Matinae more modoque (Od. 4. 2. 28), Carneadeo more et modo disputata (Cic. Tim. 1), si humano modo, si usitato more . . . peccasset (Cic. Verr. 2. 2. 9. 3), and in: tempus secum ipsa modumque exigit (Aen. 4. 475) we have tempus short for locum et tempus and modum for morem et modum.

When we turn to manus, we recall Propertius’s benediction on the earliest Roman artist we can name:

At tibi, Mamuri, formae caelator ahenae,

Tellus artifices ne terat Osca manus (4. 2. 61-2).


Aut certe tabulae capient mea lumina pictae

Sive ebore exactae, seu magis aere manus (3. 21. 29-30),

he is evidently using manus for artes. I read in Petronius: nam et Zeuxidos manus vidi (83), and in Martial: Mentoris haec manus est, an, Polyclite, tua? (8. 51. 2), and in Statius: vidi artes, veterumque manus variisque metalla viva modis (Silv. 1. 3. 47). From this union and that in Martial’s:

Argenti genus omne comparasti

Et solus veteres Myronos artes,

Solus Praxitelus manum Scopaeque (4. 39. 1-3),

we get the metonymy of manus for artes. Virgil has it in: quale manus addunt ebori decus (Aen. 1. 592), if it is not rather a 188 synecdoche for artes et manus. But in: et recidiva manu posuissem Pergama victis (4. 344) manu has gathered additional force from the context and stands for ars bellica. Petronius introduces his verses on the civil wars thus: etiam si nondum recepit ultimam manum (118 fin.), where Servius explains: translatio a pictura, quam manus complet et ornat extrema. But Virgil shapes the figure thus: extremam Saturnia bello imponit regina manum (Aen. 7. 572-3); and in older Latin literature manus is far less associated with art than with arms. This association is not plain at first glance in: pacem orare manu, praefigere puppibus arma (10. 80), where one thinks of manu as for manum tendendo, recalling: manus ad Caesarem tendere (B. G. 2. 13. 2) and: supplexque manus ad litora tendit (Aen. 3. 592). But though manu and pacem are the last and first words of the first colon, probably the connexion of manu, the last word of the first, with arma, the last word of the second colon, is still closer, and Gesner is right in supplying armata with manu.

The association of manus with arma is plain in the Laws of the Twelve Tables in: si telum manu fugit (Tab. VIII) and: manu fustive si os fregit (ib.). Cicero gives us: non exercitu amisso nudus inservorum ferrum et manus incidisset (Tusc. 1. 86. 35), and Tacitus: ut rem ad mucrones ac manus adducerent (Agr. 36). We have the union distributed in: cum . . . pugna iam in manus, iam ad gladios . . . venerat (Liv. 2. 46. 3). Still more common is the union found in: senatus iure optimo vim et manus intulisset (Cic. Cat. 1. 21. 8), praesidioque contra vim et manum comparando (Sest. 92. 42). We have such pairs as ferrum et manus, gladii et manus, vis et manus often shortened to manus with the meaning of weapons or blows, as in: manu cum hoste confligere (Cic. Off. 1. 81. 23), cum tribunus plebis populo concitato rem paene ad manus revocasset (Cluent. 136. 49), neque umquam ad manum accedere licebat (Nep. Eum. 5), ut paene uno tempore et ad silvas et in flumine et iam in manibus nostris hostes viderentur (B. G. 2. 19 fin.), qui tecta manu defendere possint (Aen. 12. 627), ubi ad manum venisset hostis (Liv. 2. 30. 12), libertique etiam ac servi patrono vel domino, cum voces, cum manus intentarent, ultro metuebantur (Tac. Ann. 3. 36. 1). In: ne qua manus se attollere nobis a tergo possit (Aen. 9. 321) manus seems a metonymy for vis.

In: quibus acer Eryx in proelia suetus ferre manum (Aen. 5. 402-3) Servius explains ferre manum as contendere. Ferre is probably poetic for conferre; and conferre manum is for boxers what conferre 189 arma is for nations in: neque validiores opibus ullae inter se civitates gentesque contulerunt arma (Liv. 21. 1, 2). When we read of the boxers: inmiscentque manus manibus, pugnamque lacessunt (Aen. 5. 429) and of armies: ubi miscuerint manus (Tac. Ann. 2. 15. 3), we naturally ask for the connexion between manus and pugna. Pugnus is the hand clenched to strike; Plautus gives us the old neuter form in: nam meumst ballista pugnum (Capt. 796), of which pugna is the old plural, which from its collective sense became later a feminine singular; for pugna means primarily a lot of fists or blows.

We need not then be surprised at the union manus et pugna that we find in: nonnumquam etiam res ad manus atque ad pugnam veniebat (Cic. Verr. 2. 5. 28. 11). Virgil has: conferre manum (Aen. 9. 44; 12. 345), but Lucretius: conferre manu certamina pugnae (4. 843); Livy gives us: impias inter nos conseramus manus (7. 40. 14), but: pugnam inter se consererent (32. 10. 8). We read: desiste manum committere Teucris (Aen. 12. 60), but: illam pugnam navalem . . . mediocri certamine et parva dimicatione commissam arbitraris? (Cic. Mur. 33. 15) and: pridie quam Siciliensem pugnam classe committeret (Suet. Aug. 96. 2). So in: quascumque urbes et agros manu ceperat (Sall. Jug. 5. 4) or: sunt (tibi) oppida capta multa manu (Aen. 12. 23) clearly manu is for pugna or bello. So in: in manibus Mars ipse, viri (Aen. 10. 280) in manibus is for: in pugna futura ‘in the battle now imminent’, non iam in votis (Serv.) ‘no longer in your prayers’; and Mars ipse seems for victoria ipsa. In: si bellum finire manu (11. 116) Servius explains: manu internecione; potest enim et pace finiri; here manu is for pugna in its deadliest form. So in: meruisse manu (2. 434), which Servius explains as: dimicasse, manu is clearly for pugna, as it is in: ipse manu mortem inveniam (2. 645), not for manu mea. In: si quas manus remisi cuique exegissem (Suet. Aug. 71. 3) ‘if I had exacted from each the stakes I let go’ we have a use of manus for matches or contests in sport closely allied with pugna in its primary meaning.

But you may ask: ‘Does it make no difference whether for pugna we use the singular or the plural of manus?’ We have: in manibus for in pugna (Aen. 10. 280), but manu for pugna (11. 116). Both seem for the pair manus et pugna; and for the pair Castor et Pollux we find in use Castor—Pollux—Castores—Polluces. For ira et dolor we had dolor (Aen. 9. 66), but dolores (5. 5; 10. 863). For currus et equi we have: curruque volans dat lora secundo (1. 156), but: ausus 190 Pelidae pretium sibi poscere currus (12. 350). In: sollicitos Galli dicamus amores (Buc. 10. 6) sollicitos amores is poetic for: curam et amorem; and probably in: securus amorum germanae (Aen. 1. 350) amores is for: amor et dolor, the love and sorrow of Dido. We read: egressi superant fossas (9. 314), where it is clear from: castra . . . vallo fossaque XVIII pedum munire iubet (B. G. 2. 5 fin.) or: vallo atque fossa moenia circumdat (Sall. Jug. 23), that what Nisus and Euryalus crossed was not fossas, but vallum fossamque. Just as fratres is used for frater et soror, patres for pater et mater, here fossas is for vallum et fossam. So too in: institutae fossae magno impedimento fuerunt (B. C. 3. 46. 5). So too with the fossae Cluiliae (Liv. 2. 39. 5), which is also called by Festus Cloelia fossa. True, Virgil styles the ditch fossae even when the vallum is expressed in: et fossas implere parant ac vellere vallum (Aen. 9. 506), where he follows the analogy of: muro fossisque tenetur (10. 236), where fossis may well be for vallo fossaque; this use of the plural certainly gives amplitude and majesty to his phrase.

I have tried to trace the origin of a number of the older and more usual metonymies; one remains that has occasioned much difficulty. In: nam seu mobilibus veris inhorruit adventus foliis (Od. 1. 23. 5) editors observe that at spring’s first arrival there are no leaves to greet her. Such a rendering as: ‘if through the light-hung leaves has run the shudder of the spring’s approach’ seems more in the style of Stephen Phillips than of Horace. Muretus changed veris to vitis, and Bentley to vepris; but the text cited by both Acron and Porphyrio is the text of the manuscripts. In Horace’s: grata vice veris et Favoni (Od. 1. 4. 1) we have the pair needed to make the passage intelligible; we have virtually the same union in:

It ver et Venus et Veneris praenuntius ante

Pennatus graditur, Zephyri vestigia propter

Flora (Lucr. 5. 737-9),

for Zephyrus is the Greek for Favonius. We read in Cicero: ver . . . cuius initium iste non a Favonio . . . notabat (Verr. 2. 5. 27. 10), and in Pliny: in principle (veris) Favonii hibernum molliunt caelum (2. 46). In the verse we have a spring day, not necessarily the first, when the budding leaves are stirred by the gentle Zephyr. Cf. also Hor. Ep. 1. 7. 13.

But in many cases, no doubt, no such union existed, from which the metonymy we meet can be derived. The use of one term of a pair to express the other, closely associated with it in thought, 191 became so habitual to the popular mind, and to the mind of the poet, that in case of objects associated in thought the figure was formed, though no antecedent union of the objects in speech existed. It may be a mere chance that for the metonymy in: volnera dirigere et calamos armare veneno (Aen. 10. 140) I have no union tela et volnera to show. Far less probable is the union equus et fuga, though we have the resultant metonymy in: dimitte fugam et te comminus aequo mecum crede solo (11. 706); or the union regina et sceptra for sceptra in: sceptra per Ionias fracta vehuntur aquas (Prop. 4. 6. 58). Some seem to come from a complex relation as we have seen: from the use of currus for equi we get by analogy the use of habenae for equi in: conversisque fugax aufertur habenis (Aen. 11. 713), and of currus for habenae in: nec tenui currus (Stat. Theb. 9. 656). In: classis Amisiae relicta laevo amne (Tac. Ann. 2. 8. 2) laevo amne is formed on the analogy of laeva ripa, an analogy originating from the pair ripa et amnis. Very easy by analogy are synecdoches such as we meet in: nec pinguia Gallicis crescunt vellera pascuis (Od. 3. 16. 35-6), me Castalia speculans ex arbore Phoebus (Prop. 3. 3. 13) for ex silva, and its opposite: viridemque ab humo convellere silvam (Aen. 3. 24).



Interesting is the relation of terra to tellus. Tellus is rare in prose; when it is used there it is said to denote the globe or orbis terrarum, as opposed to terrae, the separate lands. But while Cicero writes: ea, quae est media et nona, tellus (Rep. 6. 17), he also writes: terram in medio mundo sitam (Tusc. 1. 40. 17). Servius in his note to: magno telluris amore (Aen. 1. 171) has: tellurem autem pro terra posuit; cum Tellurem deam dicamus; terram elementum. Compared with terra, tellus is the rarer word even in poetry; but it is far oftener used for terra, the element, than for the goddess. We read: omnis feret omnia tellus (Buc. 4. 39), tellus . . . glandem mutavit arista (Geo. 1. 7-8), fudit equum . . . tellus (ib. 13), adgeritur tumulo tellus (Aen. 3. 63), tellus inarata (Hor. Epod. 16. 43); and for the goddess too Terra is used more often than Tellus, vide Aen. 4. 178; 6. 595, Lucr. 5. 1402, Suet. Tib. 75. 1. Indeed Cicero asks: Terra ipsa dea est; . . . quae est enim alia Tellus? (N. D. 3. 52. 20). Varro knows a masculine form Tellumo, evidently formed on the analogy of homo.

While for a separate land terra is far the more usual, we have Gnosia tellus (Aen. 6. 23), debita tellus (7. 120), Iubae tellus (Od. 1. 22. 15), Pontica tellus (Ov. Pont. 4. 9. 115), but Pontica terra (ib. 114). We have: tellure sub ima (Aen. 6. 459), but: sub terras ibit imago (4. 654); and of the upper world: producit corpora tellus (12. 900), ostendent terris (6. 869). Terra appears in prose mostly in union with mare, but Ovid gives us: mare et tellus (Met. 1. 291). Etymologically terra seems the old neuter plural of an Italic word, seen in Oscan as terum ‘a farm’, and is to be connected with torreo and tergeo, of which it seems an assimilated form of the past participle tersa (cf. ferre and *ferse). Tellus is less clear; but Walde relates it with tabula, and with the Attic τηλία, a board. Terra, then, will be the dry surface, the earth as opposed to the sea, while tellus will be the flat surface, the earth as opposed to the mountains, or the sky. Tellus is the rarer and more obscure word, found almost entirely in poetry; we may assume that it is the older word, to which terra, older tersa, was joined later as an epithet. For tellus terra, the union thus produced, 193 one only of the pair was commonly expressed, usually the second, terra, but at times tellus. But you may ask: ‘Where shall I find this union, tellus terra “the dry flat”?’ I read in Varro: Iuppiter pater appellatur, Tellus Terra mater (appellatur) (R. R. 1. 1. 5). Terra here has been usually coupled with mater in translation; but Victorius saw from the structure of the preceding phrase, with which this is evidently parallel, that it must be joined with Tellus; and in his anxiety to avoid this unusual union he changed terra to vero.

In Tellus Terra we have a type of an older kind of union of substantives not connected by et. Though both tellus and terra remained substantives, apparently because this pair soon became rare, being commonly expressed by one of its terms, in other unions of this type one of the terms commonly becomes an adjective. But in older Latin this term, later felt to be an adjective, is still regarded as a substantive. Take for example bipennis, as we see it in: ferro sonat alta bipenni fraxinus (Aen. 11. 135), to which Servius’s note is: ferro bipenni; ad epitheton transtulit nomen proprium; nam bipennis per se plenum est et securim significat. Servius feels that the noun is different from the adjective (Dionysius Thrax did not), but is surprised to find bipennis, which he feels to be a noun, here used as an adjective. Does Virgil feel it to be an adjective? From the frequent occurrence of securis bipennis in the Glossaries (vide Thesaurus ii, pp. 2001-2) we may conclude that this was the prose form of the union, for which Virgil and Horace substituted ferrum bipenne as a poetic form. We have this union distributed in: ornum . . . ferro accisam crebrisque bipennibus instant eruere agricolae certatim (Aen. 2. 627) and: ut ilex tonsa bipennibus . . . ab ipso ducit opes animumque ferro (Od. 4. 4. 57-60), where ferro is for ferro bipenni and bipennibus for ferris bipennibus. But for securis bipennis ‘the double-bladed cutter’ the second term bipennis is often used, and to Servius conveys the same meaning as the first term securis, which is also in frequent use.

We read: cum saevum cupiens contra contendere monstrum (Catull. 64. 101), but in v. 110: sic domito saevum prostravit corpore Theseus, where saevum is used for saevum monstrum. We have: proiciet truncum summisso poplite corpus (ib. 370), but: truncumque relinquit sanguine singultantem (Aen. 9. 332), where truncum is by synecdoche for corpus truncum ‘a headless corpse’. But in: illa quidem pugnat recto se attollere trunco (Ov. Met. 2. 822) truncum is not a headless corpse, it is not even a corpse. Here by metonymy truncum is used for corpus. We read: Parnasi vertice summo (Catull. 64. 390), but: 194 saxi de vertice (Aen. 2. 308); which is the older term for top, vertex or vertex summus? What does vertex mean in: rapidus vorat aequore vertex (Aen. 1. 117)? What but ‘the hurtling turn (or whirl) of the main gulps it down’? So the mountain top is vertex summus, the highest turn of its outline, shortened usually to vertex. In: non flavo retinens subtilem vertice mitram (Catull. 64. 63) vertex is used for the head, but we have: nonne ab imis unguibus usque ad verticem summum? (Cic. Rosc. Com. 20. 7), showing that in: talos a vertice pulcer ad imos (Hor. Ep. 2. 2. 4) a vertice is short for a vertice summo. This is confirmed by Cicero’s use of extremus vertex for the pole (N. D. 2. 105. 41) but cf. vertices for the poles in Rep. 6. 20.

In: postquam res Asiae Priamique evertere gentem (Aen. 3. 1) it is hard for us to see how the union: res publicas, for which res stands here, could be regarded as a pair of substantives; but we might turn to: caruit senatu; caruit publico (Cic. Mil. 18. 7) or: morem in publicum consulendi (Plin. Ep. 9. 13. 21). The agreement of publicus with res is a long step towards the establishment of the epithet as a part of speech. The omission of this is to some extent made good by the context in: nono die in iugum (summum) Alpium perventum est (Liv. 21. 35. 4), falcibus et messae ad lunam (plenam) quaeruntur ahenis pubentes herbae (Aen. 4. 513), videt Iliacas ex ordine pugnas (depictas) (1. 456), teneras turbavit ianua (patefacta) frondes (3. 449), Augusti avi memoria, (commemoratus) socer Drusus (Tac. Ann. 1. 41. 3), qui campis (extremis) adstiterant (ib. 2. 17. 4). In the phrase: ad unguem factus homo (Hor. Sat. 1. 5. 32) we are glad of the help given us by praesectum (A. P. 294) and by Pers. 1. 65 as well as by the scholia of Acron, Porphyrio, and Servius (ad Geo. 2. 277). In: numquam homini satis cautum est in horas (Od. 2. 13. 14) the pair horas singulas is so commonly used that the omission is hardly felt. In: avunculum (magnum) Augustum ferens (Ann. 2. 43. 6) Tacitus feels that the relationship of Augustus to Germanicus is known to all his hearers.

But the noun, not the adjective, is the word more commonly omitted, as we saw in Tellus Terra. At times the omission is cleared up by the context, as in: in dubiis (rebus) responsa petunt (Aen. 7. 86), in lento luctantur marmore tonsae (palmae) (7. 28), interea medium (mare) Aeneas iam classe tenebat (5. 1), caesis ut forte iuvencis (cruor) fusus humum viridesque super madefecerat herbas (5. 330), turrim in praecipiti (loco) stantem (2. 460), tranquillo (caelo) silet (5. 127), ego limis (oculis) specto (Ter. Eun. 601), ut limis rapias (Hor. Sat. 2. 5. 53), 195 perfundit gelida (ib. 2. 7. 91). Generally it is from a very common union that the noun is omitted, as: in curuli (sella), stativa (castra), molaria (saxa), pluvia (aqua), (sol) oriens or occidens, magni (preti) constat. When Tacitus writes: iuncto ponte tramittit (Ann. 1. 49. 6) the real nature of the phrase is obscured by the later meaning of iungere ‘to join’, and by such a syntax as we have in: pontes et propugnacula iungunt (Aen. 9. 170). When we compare with it: ponte Ticinum iungunt (Liv. 21. 45. 1) and: aggere aut ponte iniecto (26. 6. 2), we see that we have in Tacitus’s phrase the older meaning of iungo ‘I yoke’, derived directly from iugum, and that the phrase is shortened from flumine ponte iuncto.

In the common use of cuncti for ‘all’ we have a like ellipsis. Festus tells us: cuncti significat quidem omnes, sed coniuncti et congregati. Walde thinks it for conciti ‘called together’; in its meaning of ‘all’ it is evidently short for omnes conciti. But in a union like senatus cunctus, it is not necessary to understand omnis or totus. The loss of omnes seems to have come very early; for we have no example left of the old union. But with the parallel word universi we have: nam id genus hominum omnibus universis est adversum (Pl. Trin. 1046), where the editors write, against the manuscripts, hominibus universis. But the phrase was imitated by Apuleius in: talibus dictis universi omnes adsensere (Met. 7. 5 init.).

Accius gives us: cuncta fieri cetera imbecilla (Carm. Fr. 14), copied by Augustine in: ante cetera cuncta (Civ. D. 11. 6). Now we understand the use of cuncti in: ut ea cuncta optima (arma) levia prae illis putet (Ace. Tr. 146), haec Iovem sentire deosque cunctos (Hor. C. S. 73), e quibus unus amet quavis aspergere cunctos (Sat. 1. 4. 87), excipit Uranie, fecere silentia cunctae (Musae) (Ov. Fast. 5. 55), celsior at cunctis Bruti praetoria puppis (Luc. 3. 535), Zalaces cunctis narratur ephebis mollior (Juv. 2. 164), where cuncti is plainly for ceteri, being short for ceteri cuncti.

For in many passages we read ceteri omnes, as in: quod pol . . . ceteris omnibus factumst (Pl. Poen. 1183), cetera omnia quasi placentam facias (Cato R. R. 77), omnia sic avido complexu cetera saepsit (Lucr. 5. 470), sic praeter mundum cetera omnia aliorum causa esse generata (Cic. N. D. 2. 37. 14), cum hominum nostrorum prudentiam ceteris omnibus et maxime Graecis antepono (de Orat. 1. 197. 44), nam cetera cernet omnia (Tib. 1. 2. 57), cetera omnia ageret faceretque ut e republica duceret (Liv. 22. 11. 2), ceteris omnibus suadentibus (22. 3. 8). Cicero gives us these examples of the use of omnes for 196 ceteri omnes: duo sola recentia ponam, ex quibus coniecturam facere de omnibus possitis (Verr. 2. 5. 34. 13), qui cum omnibus potius quam soli perire voluerunt (Cat. 4. 14. 7), haurire me unum pro omnibus illam indignissimam calamitatem (Dom. 30. 11). And in: ceu cetera nusquam bella forent (Aen. 2. 438) evidently cetera is equivalent to alia.

How did it reach its usual meaning of ‘the rest’? We have the two unions, ceteri cuncti and ceteri omnes, for which if cuncti or omnes is retained, it means, not ‘all’, but ‘all the others’; if ceteri is retained, it means, not ‘others’, but ‘all the others’. There is nothing in the word ceteri to lead us to this inclusive sense; in Greek this meaning is given, not by ἕτεροι, but by οἱ ἄλλοι. We find too: ne reliquas fortunas omnes amitterent (Verr. 2. 3. 121. 52), res capitales et reliquas omnes iudicabant idem (Rep. 3. 35), showing that reliquos in: reliquos hos esse ex bello, is for reliquos omnes.

We have alii in the poets and in some prose writers used as the equivalent of ceteri; as in: ea libertas est qui pectus purum . . . gestitat; aliae res obnoxiosae nocte in obscura latent (Enn. Fab. 377-8, M.), ad fratrem modo captivos alios inviso meos (Pl. Capt. 458), si alia membra vino madeant, cor sit saltem sobrium (Truc. 855), quorum unus Homerus sceptra potitus eadem aliis sopitus quietest (Lucr. 3. 1038), inde alias animas . . . deturbat (Aen. 6. 411), obstupuere animis alii, sed Troius heros agnovit sonitum (8. 530), vinci animos, ubi alia vincantur, adfirmans (Liv. 21. 12. 6), ille potens; alii sordida turba iacent (Ov. Am. 2. 2. 30), mox desolatus aliorum discessione (Tac. Ann. 1. 30. 4), quod senatum invidia liberassem, qua flagrabat apud ordines alios (Plin. Ep. 9. 13. 21). We find the union alii omnes, of which these are shortened forms, in: quom . . . nos, Iuppiter, iuvisti dique alii omnes caelipotentes (Pl. Pers. 755), et hac re et aliis omnibus (Ter. Ad. 925), et quid factum vini, frumenti, aliarumque rerum omnium (Cato R. R. 2. 1), ecpendi longe opera ante alia omnia (Lucil. 29. 16, M.), tum Catilina polliceri tabulas novas . . . alia omnia (Sall. Cat. 21. 2), quamquam alia omnia incerta sunt (Cic. Phil. 4. 13. 5); and cuncta alia in: huic mulieri cuncta alia fuere praeter honestum animum (Tac. Ann. 13. 45. 2).

The variations consequent on this union of alii omnes will be best exemplified in examining Virgil’s treatment of ante alios omnes, used by him as a substitute for maxime omnium. We have the full expression in: petit ante alios pulcerrimus ornnes Turnus (Aen. 7. 55). He transfers it to the comparative in: scelere ante alios inmanior 197 omnes (1. 347). We have it shortened in: ante omnes stupet ipse Dares (5. 406), nobis placeant ante omnia silvae (Buc. 2. 62), ante omnes exit locus Hippocoontis (Aen. 5. 492), and formaque ante omnes pulcer Iulus (5. 570), where it is restored to the positive. It is shortened to ante alios in: effugit ante alios (5. 151), Phyllida amo ante alias (Buc. 3. 78), ante alios . . . diligit ignes (Aen. 8. 590), cara mihi ante alias (11. 537), ante alios dilectus lapyx (12. 391).

Its use is complicated with that of omnium primus or unus or solus, phrases also used as substitutes for maxime omnium. Of these we have primus (omnium) in: primus ego in patriam mecum . . . Aonio deducam vertice Musas (Geo. 3. 10); unus (omnium) in: iamque adeo super unus eram (Aen. 2. 567); solus (omnium) in: quamvis solus avem caelo deiecit ab alto (5. 542). Just as was ante omnes, so we find unus (omnium) coupled with the superlative in: Rhipeus, iustissimus unus qui fuit in Teucris (2. 426), cf. 7. 536. In: primum ante omnes victorem appellat Acesten (5. 540) and: me vero primum dulces ante omnia Musae (Geo. 2. 475) we have a coupling of primum (omnium) with ante omnes (alios), each phrase being the equivalent of maxime omnium; so that the doubling is parallel to that in quamquam. The form of ante omnes evidently favours this union with primum. Following the analogy of this union we have sola (omnium) joined with ante alias in: fida ante alias quae sola Camillae (Aen. 11. 821), and una (omnium) in: o felix una ante alias Priameia virgo (3. 321).

Last of all, and following the transfer of ante alios omnes to the comparative in Aen. 1. 347, we have ante alios omnes transformed to magis omnibus, the comparative phrase of which maxime omnium is the superlative, and with this is joined unam (omnium) in: quam Iuno fertur terris magis omnibus unam posthabita coluisse Samo (Aen. 1. 15) ‘which land alone (of all) Juno is said to have loved more than all lands, preferring it even to Samos’. The syntax in: si nondum exosus ad unum (omnes) Troianos (5. 687) is one common to prose and verse, and is probably for: omnes ad unum (et ultimum enumeratos). So in: venit summa dies (2. 324) summa seems short for summa et ultima. I have not found the pair, but the distribution in: summa summarum in illa gloria fuit . . . Asiam ultimam provinciarum accepisse (Plin. N. H. 7. 26 fin.) points to it; just as: valet ima summis mutare (Od. 1. 34. 12) points to: superis deorum gratus et imis (Od. 1. 10. 20). So with suprema et ultima, from which would spring such a use as: in te suprema salus (Aen. 12. 653). That in: supremo 198 te sole domi, Torquate, manebo (Ep. 1. 5. 3) we have no such complex idea, but the simple one of ‘high noon’, we are taught both by the plain meaning of the epistle, and by the scholia of both Acron and Porphyrio.

Difficulty has been felt with: si visurus eum vivo et venturus in unum (Aen. 8. 576), which is poetic for: si eum visurus sum et venturus in unum et eundem locum cum eo. We may compare: ante annos animum . . . gerens (9. 311) with: hosti ante exspectatum positis stat in agmine castris (Geo. 3. 348), where ante exspectatum is for the prose spe citius. Curious is the ellipsis of pluribus in: saepe decem vitiis instructor (Hor. Ep. 1. 18. 25), which is made good by changing the adjective to the comparative.

We have already noted several omissions of the past participle, where the meaning is obvious from the context. To these we may add: nos abiisse rati et vento (actos) petiisse Mycenas (Aen. 2. 25), positis novus exuviis nitidusque iuventa (novata) (2. 473), praecipites (actae) atra ceu tempestate columbae (2. 516), cum mihi se, non ante oculis tam clara (oblata), videndam obtulit (2. 589), quo magis inceptum peragat (coacta) . . . vidit . . . latices nigrescere sacros (4. 452), in medium . . . caestus proiecit, quibus (indutus) acer Eryx, etc. (5. 402), cum iam in orbem (coacti) pugnarent (Liv. 21. 56. 2).

Zumpt notices how common in Latin is the omission of facio (Lat. Gr. 771), especially in short propositions giving an opinion on a person’s actions, and in the phrase: finem facere, e.g. in: nihil per vim umquam Clodius, omnia per vim Milo (Cic. Mil. 36. 14), at stulte (Regulus) (Off. 3. 101. 27), ut et ipse nequid tale posthac (ib. 1. 33. 11), hoc quidem non belle (Hor. Sat. 1. 4. 136), quae cum dixisset, Cotta finem (Cic. N. D. 3. 94. 40). So in idioms like nihil aliud quam, as in: per biduum tamen nihil aliud quam steterunt parati ad pugnandum (Liv. 34. 46. 7), ac si nihil aliud, volneribus suis ferrum hostium hebetarent (id. 30. 35. 8). From such phrases it seems to have been transferred to facere non possum quin, and hence to possum, especially in poetry, where the ellipsis gives a vagueness of meaning and width of range dear to poetic diction, as in: potes namque omnia (Aen. 6. 117), non omnia possumus omnes (Buc. 8. 63), possunt quia posse videntur (Aen. 5. 231), furens quid femina possit (5. 6), hactenus . . . potui (11. 823), possit quid vivida virtus (11. 386). In: quid non mortalia pectora cogis? (3. 56) Virgil has transferred this ellipsis to cogo.

The use of facere with possum in poetry seems rare, and even in 199 prose it is not common, though facere non possum quin seems the more usual form of the idiom for ‘I cannot help’; and we have in Cicero: quoad eius facere poteris (Att. 11. 12. 4). Fieri non potest quin is also the common form for the idiom for ‘it must be that’, and in poetry fieri is much more usual with potest than is facere, as in: nil igitur fieri de nilo posse fatendumst (Lucr. 1. 205), quod fieri ferro liquidove potest electro (Aen. 8. 402). But the omission of factus, the past participle of fieri, is one of the most common ellipses in Latin poetry, and has led to difficulty, as we shall see. It is perhaps due to the connexion of fieri with esse through fui and futurus. Esse has no past participle, and fio is inclined to follow it. Our English verb ‘is’ has adopted ‘been’, the past participle of be, cognate with fieri.

The following are examples of the omission of factus when joined with a noun: viridem . . . frondenti ex ilice (factam) metam constituit (Aen. 5. 129), probat auctor (factus) Acestes (5. 418), hic victor (factus) caestus artemque repono (5. 484), nullo discrimine (facto) (12. 770), et modo formosa qui multa Lycoride Gallus mortuus inferna volnera (facta) lavit aqua (Prop. 2. 34. 92), post pugnam ad Trebiam (factam) (Liv. 21. 15. 6), orator publicae causae (factus) (Tac. Ann. 1. 19. 5), luce demum (facta) (ib. 1. 39. 8), globo (facto) perfringerent (2. 11. 4), haud proinde in crimine incendi (facti) quam odio humani generis convicti sunt (15. 44. 5). Its omission leads at times to the apparent use of an adverb as an adjective, as in: ignari . . . ante (factorum) malorum (Aen. 1. 198), gravibus superne (factis) ictibus (Tac. Ann. 2. 20. 3), sensit dux imparem (factam) comminus pugnam (2. 20. 4). Indeed in: dictaturae ad tempus sumebantur (1. 1. 1) we seem to have a use of two for four—for ad tempus factae ad tempus sumebantur.

The following are some of the examples I have noted of the omission of factus when joined with an adjective: insignis (factus) tota cantabitur urbe (Hor. Sat. 2. 1. 46), quod meretrice nepos insanus (factus) amica filius uxorem grandi cum dote recuset (Sat. 1. 4. 49), vivos et roderet ungues (factos rodendo) (Sat. 1. 10. 71), aspera (facta) nigris aequora ventis (Od. 1. 5. 6), deliberata morte (facta) ferocior (Od. 1. 37. 29), duplicem (factam) gemmis auroque coronam (Aen. 1. 655), iuvenem monstris pavidi (facti) effudere marinis (7. 780), fluctusque atros (factos) aquilone (5. 2), flammam . . . quae plurima (facta) vento corripuit tabulas (9. 536), postquam habilis (factus) lateri clipeus (12. 432), caeso moenia firma (facta) 200 Remo (Prop. 3. 9. 50), with which compare: quamvis firmatus animo (Tac. Ann. 1. 6. 1). The verse:

Tu satius memorem Musis imitere Philetan (Prop. 2. 34. 31),

where I follow the reading of the Neapolitanus, long since recognized as the best codex, has given the editors trouble ever since Scaliger changed memorem Musis to the preposterous: Musis meliorem. We have had: Mimnermi, Meropem, tenerum, and Butler thinks it impossible to assign any meaning to memorem. But if we supply factum, we have: memorem factum Musis ‘inspired by the Muses’.

We have noted how Propertius uses firmus, where Tacitus has firmatus. So we have adjectives, evidently with factus to be supplied, used for past participles passive in: Scipiadas duros bello (Geo. 2. 170), durum a stirpe genus (Aen. 9. 603), adsis o placidusque iuves (4. 578), et placidi servate pios (3. 266), Cererique sacrum Polyboeten (6. 484), truncos inhonesto volnere nares (6. 497), trunca manu pinus regit (3. 659), cava flumina crescunt (Geo. 1. 326), nuda genu (Aen. 1. 320), nudus membra Pyracmon (8. 425). We have the opposite use in: nudato capite (12. 312) and: sub rupe cavata (1. 310). We have: infrenis equi lapsu (Aen. 10. 750), where infrenis is plainly for infrenati. We have seen that eques is used for equus; and so we have in Livy: equites frenatos infrenatosque (21. 44. 1); and infrenus transferred to the horsemen in Virgil’s: Numidae infreni (Aen. 4. 41) ‘the Numidians who have no bridles’. On the same pattern from innumerati we get innumeri in: innumerae gentes populique (Aen. 6. 706), seminaque innumero numero . . . volitent (Lucr. 2. 1054), innumeram pecuniam circumdedisti (Tac. Ann. 14. 53. 5). Curious is the aprosdoketon in Horace’s: nos numerus sumus (Ep. 1. 2. 27) ‘we don’t count’, but are merely counted. In: aequo animoque agedum magnis concede; necessest (Lucr. 3. 962) Munro has changed magnis, the reading of all manuscripts. Giussani prefers to read gnatis, found in the margin of the Cod. Bern.; but gnatis seems not an emendation, but an explanatory note. Probably, as v. 967 indicates, magnis here is for gnatis magnis factis.

We have, substituted for past participles, adjectives of different root but like meaning in: nunc cassum lumine lugent (Aen. 2. 85) and: aethere cassis (11. 104), where cassus is for privatus; in: gravi . . . saucia cura (4. 1) and: adversa sagitta saucius ora (12. 652), where saucius is for volneratus; in: festosque dies de nomine Phoebi (6. 70), where festos is for sacratos; and: adytis cum lubricus anguis ab imis (5. 84), where lubricus is for lapsus. Such participles in -tus or -sus 201 were once active in meaning as well as passive; we have still cretus (2. 74) and titubata (5. 332) to prove this for Latin. So with the adjectives substituted for them; we have caecus usually active, but passive in: vada dura . . . saxis . . . caecis (3. 706), being for occultis; gnarus usually active, but passive in: gnarum id Caesari (Tac. Ann. 1. 5. 4), being for notum; ignarus usually active, but passive in: mare magnum et ignara lingua commercio prohibebant (Sall. Jug. 18. 6); in: ignarum Laurens habet ora Mimanta (Aen. 10. 706) it seems to join both passive and active forces, ‘a stranger—unknowing and unknown’.

But in Aen. 5. 84 lubricus may well be short for lubrice lapsus; certainly in: udo turpia membra fimo (5. 358) turpia is for turpiter foedata. In: facilis iactura sepulcri (2. 646) facilis seems for facile ferenda; in: ager Tusco . . . proximus amni, longus in occasum (11. 317) longus for longe patens; in: minus est gravis Appia tardis (Hor. Sat. 1. 5. 6) tardis for tarde procedentibus; in: quo patre sit natus, num ignota matre inhonestus (Sat. 1. 6. 36) inhonestus for inhoneste natus; mors falsa (Mart. 7. 47. 9) falsa for false nuntiata, and in: hunc ego te, Euryale, adspicio? (Aen. 9. 481) hunc seems for hoc modo laniatum. No doubt on this shortening is propped the use of the adjective for the adverb in: hunc primo levis hasta Themillae strinxerat (Aen. 9. 576), where levis is for leviter; and in: Aurora socios veniente vocari primus in arma iube (10. 242); though in the latter hypallage seems involved, the shortening being for: primo veniente . . . primus iube. We have hunc for huc hunc in:

Dis equidem auspicibus reor et Iunone secunda

Hunc cursum Iliacas vento tenuisse carinas (Aen. 4. 46),

where Servius read huc; but in his note to Aen. 1. 534 he reads hunc. And in:

Hoc precor, hunc illum nobis Aurora nitentem

Luciferum roseis candida portet equis (Tib. 1. 3. 93-4),

we have hunc for huc. So in: hic cursus fuit (Aen. 1. 534) hic is plainly for huc; Servius suggests that it is ‘pro illuc’, which would be the prose equivalent. In: Anthea si quem iactatum vento videat (Aen. 1. 181) quem is for usquam.

We have the opposite construction in: unde genus Longa nostrum dominabitur Alba (Aen. 6. 766), where unde is for unde ortum; and in: per nudam infra (iacentem) glaciem (Liv. 21. 36. 6). We have an adverb taking the place of an adjective in the use of partim for alii; as in: partim galea clipeoque resultant inrita, deflexit partim 202 stringentia corpus alma Venus (Aen. 10. 330). In: ergo nunc Dama sodalis nusquam est? (Hor. Sat. 2. 5. 101) nusquam seems short for nullus usquam; and so in: ceu cetera bella forent nusquam (Aen. 2. 438). In: numquamne ad se nisi filios familiarum venturos? (Tac. Ann. 1. 26. 5) numquam is short for nullos umquam. In: numquam hodie effugies (Buc. 3. 49) numquam seems for nullo pacto umquam; and so in: hodie numquam monstrabo (Ter. Ad. 570) and: numquam omnes hodie moriemur inulti (Aen. 2. 670).

But in: ne qua forent pedibus vestigia (porro) rectis (Aen. 8. 209) it is the adverb and not the participle that is omitted; and so it is with the verb in: quae sibi quisque timebat, unius in miseri exitium conversa (aequo animo) tulere (Aen. 2. 131). As a rule in such a union as sorte ductus, it is the verb that is retained. We read: fixusque (re decreta) manebat (Aen. 2. 650), sic te ut (morte) posita crudelis abessem? (4. 681), grege de intacto (stimulis) (6. 38), dum vastabant Pergama reges debita (fatis) (8. 375), nisi mutatum (in acetum) parcit defundere vinum (Hor. Sat. 2. 2. 58), milites (sorte) ducti (Liv. 21. 37. 2), regibus (vita) defunctis (Tac. Ann. 2. 42. 7). We have an adjective substituted for the participle in: saepius (inter se) discordes sunt (Tac. Ann. 2. 56. 1); and it is the opposite of this syntax that we find in: nequam (utilis) and frugi (utilis), where nequam and frugi have come to be regarded as adjectives, which give us comparatives nequior and frugalior.

But in: saepta armis (Aen. 1. 506), where armis is for viris armatis, we have the union of a participle and a noun shortened to a noun representing an idea related to that presented by the union and used for it by metonymy. So in: ut Chio nota si commixta Falerni est (Hor. Sat. 1. 10. 24), where nota is for amphora notata. From Virgil’s: pervius usus (Aen. 2. 453) Tacitus has formed a noun pervium in: ne pervium illa Germanicis exercitibus foret (Hist. 3. 8. 3). The use of aequus for aequo animo in: quod adest memento componere aequus (Hor. Od. 3. 29. 33) seems closely connected with this syntax. Similar too seem: simul divom templis indicit honorem (Aen. 1. 632), where honorem is for: sacra ad divom templa honoranda; and: meritos aris mactavit honores (3. 118), where honores is for: victimas ad aras divom honorandas. Still further developed seems this syntax in: relliquias Danaum atque inmitis Achilli (1. 30), for: relliquias telis Danaum atque inmitis Achilli superstites.

So close is the relation between the present and past participles of deponent verbs that we need not be surprised to find the present 203 participle also constructed in like syntax. And we have: oleo labente (perfusi) . . . nudati socii (Aen. 3. 281), where labente ‘slipping’ is used for lubrico ‘slippery’, or ‘causing to slip’, much as in: pallida Mors (Od. 1. 4. 13) pallida is used for atra or atrox with the force of ‘causing paleness’, and Horace’s: tonsor inaequalis (Ep. 1. 1. 94) is the barber who cuts the hair uneven. We have: vina liquentia (Aen. 5. 238) for liquida, candentis vaccae (4. 61) for candidae, nigrantes terga iuvencos (5. 97) for nigra habentes, ramum Lethaeo rore madentem (5. 854) for madidam, humentem . . . umbram (3. 589) for humidam. We have the opposite in: fluidum lavit inde cruorem (3. 663) for fluentem, and in: tum pavidae tectis matres ingentibus errant (2. 489), fugam trepidi celerare (3. 666), Palmumque fugacem (10. 697) for usque fugientem. In: stipendia militibus, agros emeritis largientur? (Tac. Ann. 1. 28. 6) militibus is short for militibus militantibus and emeritis for militibus emeritis—a good example of distribution. Of the union of a present participle with a substantive we have the participle omitted in: ter maestum funeris ignem lustravere in equis (sedentes) (Aen. 11. 190), gaudens popularibus auris (faventibus) (6. 817), locum tendunt superare priorem (habentes) (5. 155), per vada (vadentes) (Lucr. 1. 200). We have the opposite in: (bovum) mugientium (Hor. Epod. 2. 11), qua colla (in equo) sedentis lucent (Aen. 11. 692), (cuiquam) quaerenti (8. 212). We have in: haec celerans (1. 656) a shortening for: celeriter ferens.



But it is not merely factus ‘that has become’ that is often omitted; the ellipsis of futurus ‘that will become’ is also frequent. I read in Martial: qui scribit . . . puero liquidas aptantem Daedalon alas (4. 49. 5), where my editor tells me that liquidas is proleptic; when I supply futuras with liquidas I have at once a full expression of the meaning. Very plain is the same ellipsis in: mox Italus Mnestheus (Aen. 5. 117), to which Servius’s note is: subaudis ‘futurus’. Virgil gives us: bello caduci (6. 481) for: bello occisi; but in: si mora praesentis leti tempusque caduco oratur iuveni (10. 622) caduco seems for casuro. So in: Chalcidicaque levis tandem super astitit arce (6. 17) we must supply futura with Chalcidica, as the settlement was later than even Aeneas’s time.

We read in Horace: te triste lignum, te caducum in domini caput inmerentis (Od. 2. 13. 11). When we compare this with Cicero’s: vitis quidem, quae natura caduca est (Sen. 52. 15) or: bacae glandesque caducae (Lucr. 5. 1363) ‘dropping berries and acorns’, we see in all three an ellipsis of futurus, but this futurus is in Horace voluntative, a sense we see constantly associated with the pure future sense in both Greek and Latin futures. To: et gener auxilium Priamo Phrygibusque ferebat (Aen. 2. 344) Servius’s note is: gener dicitur et qui est et qui esse volt, and we see that here with gener we must supply the voluntative futurus, as futuri voluntative is to be supplied in: aegram nulli quondam flexere mariti (4. 35). In: huc periture veni (11. 856) periture is ‘doomed to die’, but in: moriturus et ipse (11. 741) Tarchon does not perish in the battle. In: oriturque miserrima caedes (2. 411) caedes is short for: pugna et caedes; and in: nulli tota morerentur in urbe (2. 439) morerentur is short for: pugnarent et morerentur. These are examples of synecdoche; moriturus is here an expressive metonymy for pugnaturus, ‘bent on battle though it involve death’.

We read: sacra Dionaeae matri divisque ferebam auspicibus (futuris) coeptorum operum (Aen. 3. 20) ‘who were to direct the voyage begun’ and: iuvenumque prodis publica (futura) cura (Hor. 205 Od. 2. 8. 8) ‘your appearance will turn the heads of all our youth’. In: cernimus adstantes nequidquam lumine torvo Aetnaeos fratres (Aen. 3. 677) nequidquam seems short for: nequidquam fratri profuturos. In: homines tantos . . . pedibus qui pontum per vada possent transire (Lucr. 1. 200) ‘men so great that on foot they could cross the sea by wading’ per vada seems short for: per vada futura vadentes ‘wading through what would prove mere shallows to men of their size’. So in: speculantur aquas (futuras) et nubila caeli (Geo. 4. 166).

To: portantur avari Pygmalionis opes pelago (Aen. 1. 363) Servius’s note is: quas Pygmalion iam suas putabat. This effect is achieved by the purposed omission of futuras, which gives us the figure we call prolepsis—a figure closely analogous to the προκατάληψις or πρόληψις by which the rhetor strives to anticipate the arguments of his opponent. So in: hi proprium decus et partum indignantur honorem ni teneant (Aen. 5. 229), though the prize is not yet won, the use of partum interferes with our supplying futurum with decus. We see now how Servius comes to feel that gener is used not only of the actual son-in-law, but of the prospective one as well. In: promissam eripui genero; arma impia sumpsi (Aen. 12. 31) Latinus declares the war an impious one; for though the wedlock is not yet consummated, Aeneas is already received into his family, and is no longer a mere suitor. But in: ardentem generum moritura tenebat (12. 55) and: coniuge praerepta (9. 138) with mariti in Aen. 4. 35, we do well to accept the omitted futurus in its voluntative sense. Other like examples are: ereptae magno inflammatus amore coniugis . . . Orestes (3. 331) and: frondentesque ferunt remos (4. 399).

From: sponsi Penelopae (Hor. Ep. 1. 2. 28) it is plain that we have here a variety of metonymy where sponsi is substituted for proci (= sponsi futuri with a voluntative sense). We have metonymies involving prolepsis in: gravidam imperiis . . . Italiam (Aen. 4. 229), where gravidam implies parituram; nullis ille movetur fletibus, aut voces ullas tractabilis audit (4. 439), where tractabilis implies: fore ut commoveretur; caecique ruunt (12. 279), where caeci is for: securi periculi futuri. In: magnum reginae sed enim miseratus amorem (6. 28) no doubt the average Roman reader, feeling that the wife was held loco filiae, was satisfied to put the king’s daughter in locum coniugis. But the relation is Cretan, not Roman, and Virgil is perhaps thinking of the Corona Ariadnae which Ovid describes as (caelo) specie remanente coronae (Met. 8. 181).



Andromache manesque vocabat

Hectoreum ad tumulum, viridi quem caespite inanem

Et geminas, causam lacrimis, sacraverat aras (Aen. 3. 303-5),

we must understand tumulum Hectoreum futurum, as it will only become Hector’s with the coming of his manes, and causam futuram lacrimis, as she hoped to bewail him there. The second is voluntative as it is in: causam (futuram) discordiae (Tac. Ann. 1. 27. 1). We have the ellipsis of futurus transferred to the object of the verb in: faciles (futuras) venerare Napaeas (Geo. 4. 535) giving us the equivalent of a result clause; as we see in: steriles exurere Sirius agros (Aen. 3. 141), where steriles is for: ut steriles fiant; in: quis indomitas tantus dolor excitat iras? (2. 594), sublimem pedibus rapuit (5. 255), id rebus (perituris) defuit unum (12. 643), where perituris is implied in exscindi, and the clause is for: id defuit unum ut res nostrae pessum irent. So too in: quos Elea domum reducit palma caelestes (futuros) pugilemve equumve (Od. 4. 2. 18) and: non equus impiger curru ducet Achaico victorem (Od. 4. 3. 6) ‘to victory’. But when the futurus thus transferred has the voluntative force, as in: deflexit partim stringentia corpus alma Venus (Aen. 10. 331) for: ut corpus modo stringerent, and: placatam Eurydicen vitula venerabere caesa (Geo. 4. 547) for: ut placetur, it is purpose clauses that result. As we have shown, the use of hortator for hortaturus in: comes additur una hortator scelerum Aeolides (Aen. 6. 529) is different in origin, though it lent support to this syntax.

We read: in . . . notos vocem vertere procellae (11. 798), where with vocem we may supply either: ibi perituram, or: eis auferendam. The ellipsis of the gerundive is more common than that of the future participle. At times the context plainly indicates the verb to be supplied, as in: arma amens capio, nec sat rationis in armis (capiendis) (2. 314) and: nec Drances potius, sive est haec ira deorum (luenda), morte luat, sive est virtus et gloria (tollenda), tollat (Aen. 11. 443-4). In: quae moenia clausis ferrum acuant portis in me exscidiumque meorum (8. 386), at first one thinks of converting: in me exscidiumque meorum to: in me meosque exscindendos. But Venus is a goddess and cannot be destroyed; it is meorum, not me, that depends on exscidium; so we must supply: in me petendam meosque exscindendos. As a rule the gerundive to be supplied belongs to a verb found in frequent union with the noun in question, as in: bello (committendo) dat signum (11. 474), ardet in arma (conserenda) (12. 71), non cassa 207 in vota (perficienda) vocavit (12. 780), aliae victu (parando) invigilant (Geo. 4. 158), custodem in vincla (iniciendum) petivit (Aen. 6. 395), cedat amicitiae Teucrorum (conciliandae) (11. 321), peto requiem spatiumque furori (leniendo) (4. 433), frigida bello (gerendo) dextera (11. 338), iaculo (iaciendo) celerem (9. 178), insidiis (locandis) (9. 237), corpora (reficienda) curamus (3. 511), se . . . gentes aeterna in foedera (servanda) mittant (12. 191), cessas in vota (facienda) (6. 51). We have in Tacitus: faciendis castris (Ann. 2. 21. 4), where we might expect struendis.

Very often the general sense of the context at once suggests the verb to be supplied, as in: paci (petendae) medium se offert (Aen. 7. 536), nec spes iam restat Iuli (recipiendi) (1. 556), arcemque adtollere tectis (defendendis) (3. 134), consurgit Turnus in ensem (adigendum) (12. 729), pronus pendens in verbera (exercenda) (10. 586), plurimus in Iunonis honorem (augendum) aptum dicet equis Argos (Od. 1. 7. 8), clipeos ad tela (depellenda) sinistris protecti obiciunt (Aen. 2. 443), sed fama classis amissae ut Germanos ad spem belli (novandi) . . . erexit (Tac. Ann. 2. 25. 1), sed referendum iam animum ad firmitudinem (reddendam) (ib. 3. 6. 3), veniam ordinis (abdicandi) ob paupertatem petenti (ib. 1. 75. 5). History helps us in: non . . . Achilles talis in hoste fuit Priamo (excipiendo) (Aen. 2. 541) and: Antonique graves in sua fata (perpetranda) manus (Prop. 3. 9. 56).

More involved is: ficto pectore fatur (Aen. 2. 107), where with ficto we may supply: ad mala consilia condenda, and in: quem pellis ahenis (et) in plumam (imitandam compositis) squamis auro conserta tegebat (11. 771). Two gerundives are to be supplied in: nullum memorabile nomen (pariendum) feminea in poena (sumenda) est (2. 583), where the first is suggested by the second; and so in: ut novissimi in culpam (subeundam), ita primi ad paenitentiam (obeundam) sumus (Tac. Ann. 1. 28. 7). Easier is the problem in: ante oculos (videnda) interque manus (tractanda) sunt omnia vestras (Aen. 11. 311). In: plures ad curas vitam produxero (Ann. 3. 24. 4) ad curas seems short for: ad res curandas. In: biiugi . . . ad frena leones (Aen. 10. 253) Servius explains ad frena as for frenati. No doubt the primary force was: ad frena trahenda; but just as puer ad cyathum (replendum) came to mean ‘cupbearer’, like the puer ab cyatho of the inscriptions, and then all feeling for the understood gerundive was lost, so here by its analogy the feeling for the gerundive with ad frena also disappears.

In Dido’s words: si mihi non animo fixum inmotumque sederet 208 (4. 15) we feel that her emphasis is rather on the future than on the past; and that inmotum is short for: inmotum et inmobile. So in: simillima proles indiscreta suis (10. 392) indiscreta suis must mean ‘that their friends had not distinguished and could not distinguish’. In: eripe me his, invicte, malis (6. 365) invicte must mean ‘invincible’, i.e. capable of achieving anything no matter how difficult. The past participle here presents the same meaning which in: nemo nisi victor pace bellum mutavit (Sall. Cat. 58. 15) is given by the perfectum logicum; Aeneas never has been and so never can be defeated. So in: te docilis magistro . . . Amphion (Od. 3. 11. 1) docilis is for docilis et doctus.

In: nobilis et fama multis memoratus in oris (Aen. 7. 564) we have a figure that we will examine more fully later. Each word of the pair nobilis et memoratus itself stands for a pair; so that the pair is short for nobilis notusque et memorabilis memoratusque, the four terms being represented by the first and last, as we saw in: Pollucis . . . Cyllarus (Geo. 3. 89). So too in: indeprensus et inremeabilis error (Aen. 5. 591) and: ardet inexcita Ausonia atque inmobilis ante (7. 623).

In: dives inaccessos ubi Solis filia lucos adsiduo resonat cantu (7. 11) inaccessos seems short for inaccessos et inaccessibiles, and so with inrupta in: quos inrupta tenet copula (Od. 1. 13. 18). But in: genus indocile ac dispersum (Aen. 8. 321) it seems that indocile is a metonymy for indoctum; and in: deus abscidit prudens Oceano dissociabili terras (Od. 1. 3. 22) the context requires for dissociabili the sense of dissociato. So in: saepe trans finem iaculo nobilis expedito (Od. 1. 8. 12) nobilis seems for notus, while in: nobile Pallanteum (Aen. 8. 341) nobile has the double force indicated above. So too in: inobservabilis error (Catull. 64. 115) and: soloque inmobilis haeret (Aen. 7. 250). But in: solus ubi in silvis Italis ignobilis aevum exigeret (7. 776) ignobilis is not ignoble, but is used for ignotus. So sutilis (6. 414) is for sutus, and tortile (7. 351) for tortum; the bark that can be sewed is sewed, and the gold that can be twisted to a necklace is so twisted. We have the opposite in: conspectus (8. 588) for conspicuus, and in: quadrifidam (7. 509), qui findi partes possit in quattuor (Serv.). We have seen how labens can be used for lapsus; so we have volatile ferrum (8. 694) for volans ferrum, and volubile buxum (7. 382) for volvens buxum. In: quod missile libro (10. 773) missile seems for mittendum; and in return: volvenda dies (9. 7) is for dies volvens, or, as Servius tells us, for volubilis.


We read: terruit auster euntes (Aen. 2. 111), where euntes is plainly for ituros. Servius explains it as for: ire cupientes, and compares: cum canerem reges et proelia (Buc. 6. 3), where the subjunctive is voluntative. This solution hardly satisfies him, however; for he adds: et est figura Graeca, ubi statuisse aliquid pro incohatione habetur. We are reminded of Clearchus, who, before he joined Cyrus, was put to death by the Spartan magistrates, but made the march to Cunaxa notwithstanding: ἐθανατώθη (Xen. Anab. 2. 6. 4); but this has rather to do with the aorist. It is only in the Iliad that εἶμι has the future εἴσομαι; in Attic it has no future, but is itself a common future of ἔρχομαι. This use seems very old, going back to a time when the present tense was still used for present and future time for verbs. Of all verbs the verb ‘to go’ seems to have been the slowest to develop a future; we still in English use its present as a future tense; Latin has, it is true, developed ibo, but for its future participle Virgil uses ituras twice (Aen. 6. 680 and 758), and euntes twice, here and in: nec nos via fallit euntes (Aen. 9. 243). Forbiger would add: prosequere . . . euntem (12. 73), where, however, euntem is present in relation to prosequere, and: nec vero Alciden me sum laetatus euntem accepisse (6. 392), where eunteni is for venientem. But in Aen. 9. 243 it seems for ituros; Mackail translates ‘nor shall we miss the way we go’; and it seems for: nec nos ituros fallet via qua eundum erit.

As we have seen, in the Comedy and in Virgil eo is used for venio, as in: vos celsis nunc primum a navibus itis? (2. 375). In: cum primum Iliacas Danai venistis ad oras (2. 117) Servius explains venistis as for venire velletis. It is plainly for venturi fuistis ‘you purposed to come’. So in: externi veniunt generi (7. 98); just as the Greek εἶμι is used for ibo, so here veniunt is for venient, and is so written in most of the best codices; cf. Serv. ad loc. Ovid has: Graia iuvenca venit (Her. 5. 118); and for other verbs of motion we find Virgil using the present for the voluntative future, as in congredior (Aen. 12. 13) ‘I will go’, sequor omina tanta (9. 21) ‘I am resolved to follow such omens’, sequimur te sancte deorum (4. 576). In: terras capere aut captas iam despectare videntur (1. 396) capere seems short for eligere capturi ‘choose for alighting’.

Closely associated in use with ire is comes, a term of the greatest importance for the Romans from the social standpoint, and destined to develop into a tide of nobility for the Romance peoples. When we read in its context: comites Catulli (11. 1) we feel that futuri should 210 be supplied; and so in: arma deosque parant comites (Aen. 2. 181) and: hos cape fatorum comites (2. 294). But the close association of comes with ire makes it probable that there was no such ellipsis here for the Roman; just as euntes could be used for ituri, so comes might imply iturus. Comes was used not merely to express the younger soldier under the guidance of the elder, as Euryalus with Nisus; but it was used of the elder soldier assigned to guide and protect the son of his leader, as Epytides with Iulus. Under the later emperors it is already a title of honour, as in: comes rei privatae (Amm. 22. 3. 7), comites sacri stabuli (Cod. Just. 12. 11. 1), comes Africae (Symm. Ep. 4. 48). We may ask what comes stabuli could mean; the answer is found in: custodem ad sese comitemque impubis Iuli Epytiden vocat (Aen. 5. 546), or: et pueri custos adsiduusque comes (Mart. 11. 39. 2), haec comes, haec custos, haec proxima mater haberi (Claudian Rapt. Pros. 3. 176). In: comes stabuli, comes is short for custos et comes, and as in: Pollucis (Geo. 3. 89) Castoris is really meant, so here comes has the sense of custos rather than of comes. So the Constable of France is really the King’s Magister Equitum.

Just as the ellipsis of futurus is frequent, so we find fore omitted, as in: moneret . . . cum dura proelia (fore) gente (Aen. 11. 48), quo mitius (fore) Romanum imperium speraretur (Ann. 2. 56. 4), quem . . . aequiorem sibi (fore) sperabat (ib. 3. 8. 1), impune (sibi fore) putans (Aen. 12. 728). We find the present infinitive of verbs of motion used for the future in: nunc iuvenem imparibus video concurrere fatis (12. 149) and: has audax sperat sibi cedere virtus (Luc. 9. 302); and a verb denoting the result of coming so used in: victi parere fatentur (Aen. 12. 568) ‘the conquered own they are on hand’, for ‘they will be on hand’; for pareo seems to be ‘I am on hand’, the neuter of paro ‘I get on hand’.

Caesar’s uses of the infinitive in: quae imperarentur facere dixerunt (B. G. 2. 32. 3), qui polliceantur obsides dare (ib. 4. 21), seem extensions of this last use. With spero too we often have the infinitive with posse used as a substitute for the future infinitive, as in: nostrasne evadere . . . sperasti te posse manus? (Aen. 9. 561), mene efferre pedem, genitor, te posse relicto sperasti? (2. 657), dissimulare etiam sperasti, perfide, tantum posse nefas, tacitusque mea decedere terra? (4. 305-6). We have the two constructions co-ordinate in: scilicet id magnum sperans fore munus amanti, et famam exstingui veterum sic posse malorum (6. 526-7). But we have just seen how decedere can be written for decedere posse; and to: totumque moved mutarive putas 211 bellum (10. 627) Servius’s note begins: deest posse. This may well be what has happened in: hoc sperem Italiam contingere caelo (5. 18), desine fata deum flecti sperare precando (6. 376), haec se carminibus promittit solvere mentes quas velit (4. 487).

This seems the natural place to deal with anachronism in the poets, especially as the Latin critics call it prolepsis. Not that it is the same as prolepsis, which is a figure of syntax, while anachronism is an error in chronology. Servius’s note to: passim . . . armenta videbant Romanoque foro et lautis mugire Carinis (8. 361) is: sed et hic prolepsin facit; nam postea sic (i.e. Carinae) dictum est. To: sceptra Palatini sedemque petit Euandri (9. 9) his note is: Palatini prolepsis est. To: ast legio Aeneadum vallis obsessa tenetur (10. 120) he writes: prolepsis; nam legionis nomen Troiani temporis non fuit. But in his note to: ipse Quirinali lituo parvaque sedebat succinctus trabea, laevaque ancile gerebat Picus (7. 187-9) he does not speak of prolepsis; for the name Quirinus is used of Mars as well as of Romulus, and though the ancilia preserved in Rome were said to date from Numa’s time, the form of the shield and its name may have been older. But to these prolepses we may apply the words of Hyginus (Gell. 10. 16. 8): ‘To the poet himself it is granted as a rule to use certain expressions κατὰ πρόληψιν historiae when speaking in his own person, just as Virgil knew of the “Lavinian town” and of the “Chalcidic citadel”’. But in: Laviniaque venit litora (Aen. 1. 2) Servius preferred the reading Lavina, and traced the name to Lavinus, a brother of Latinus; he adds: quamvis quidam superfluo esse prolepsin velint. In: Chalcidica . . . arce (6. 17), though the city of Cumae was founded long after the Trojan war, the temple of Apollo was said to be older; and the prolepsis here is of the name only and is in the words of the poet, as is Carinis.

Of Hippolytus Horace tells us: infernis neque enim tenebris Diana pudicum liberat Hippolytum (Od. 4. 7. 25-6); but Virgil (Aen. 7. 761 ff.) makes Diana recall him to the upper world, where secluded in Italian woods he lives as Virbius. To this we may apply the note Servius wrote to Aen. 6. 618, frequenter enim variant fabulas poetae. So with: Protei columnas (11. 262); the Pillars of Hercules at Gades and those in Pontus are taken as western and eastern bounds of the Mediterranean system of seas; and we have here a southern bound formed on analogy of these, and associated with Proteus. By Proteus Virgil seems to have meant the god whom Tacitus presents as the real Hercules (Ann. 2. 60. 3); but here again 212 we are dealing with myth. This is not so certain of Dido; indeed nothing seems very certain about her. Servius’s account (ad Aen. 1. 267) places the founding of Carthage seventy years before that of Rome, and so two hundred and seventy years after the fall of Troy. But her story had been connected with that of Aeneas by the genial poet Naevius, and Ennius had adopted the legend, which thus became a household tale with the Romans. Probably the nearest approach to truth we have here is what we read in Justin (18. 6. 6), where there is no mention of Aeneas, and Elissa burns herself to avoid the proposals of Iarbas.

In: et socii amissi petierunt aethera pennis (Aen. 11. 272) Virgil makes Diomede relate a metamorphosis of his followers, which on the commonly accepted account took place after Diomede’s death from their grief at the loss of their king. Servius says (ad 11. 271): hoc loco nullus dubitat fabulae huius ordinem a Vergilio esse conversum; but we are still dealing with myths. When in: atque iterum in Teucros Aetolis surgit ab Arpis Tydides (10. 28) the goddess Venus pleading Aeneas’s cause before Jupiter, represents as close at hand an event that was never to happen, the case is more serious. But Venus is pleading a cause; and Virgil seems to have thought it a still more serious violation of the probabilities to attribute unblemished veracity to a causidicus. We have already dealt with the apparent flagrant self-contradiction in Aen. 6. 618.

In Gellius (10. 16. 14) Hyginus is made to cite from Virgil:

Eruet ille Argos Agamemnoniasque Mycenas

Ipsumque Aeaciden, genus armipotentis Achilli,

Ultus avos Troiae, templa intemerata Minervae (Aen. 6. 838-40).

The templa intemerata is interesting; though it is the oldest citation of the verse we have, it finds no place in any of our editions. Hyginus wants to throw v. 839 out of the Aeneid. He thinks that Aeacides is Pyrrhus. But the prophecy of Anchises seems to me quite correct here; Aeacides is King Perseus (cf. Prop. 4. 11. 39-40); and the avenging victor is Aemilius Paullus. Hyginus’s error arose from his failure to see the figure of hendiadys in v. 838, through which Greece is named by Argos and Mycenae, one by two; just as the apparent violations of history in Od. 4. 4. 17-18 and 4. 8. 15-20 are due to a failure to perceive the opposite figure.

Far more serious seemed to Hyginus the prolepsis in: portus . . . require Velinos (Aen. 6. 366). Virgil puts the words in the mouth of 213 Palinurus, and the town of Velia was not founded till about 600 years after his death. How could he already know the name? and how was Aeneas to find a town that was not to come into existence for 600 years? But the victory of Paullus, which Anchises foretells to Aeneas, was won three centuries later, and Palinurus was a shade as well as Anchises, though it is true he had not as yet had Anchises’ time or opportunities for working out the future. But if shades cannot as such foretell the future, what is to become of our faith in necromancy? Servius too speaks of the prolepsis here as vitiosissima, because it is not the statement of the poet himself but of Palinurus; but he adds: quamquam alii ad divinandi scientiam referant, quasi ab umbra dictum. There is no such excuse for Catullus’s Ariadne; she is still alive when the poet puts in her mouth the words: Idomeneosne petam montes? (64. 178). Idomeneus was the grandson of Minos, and was yet unborn when his aunt used his name to designate her native island. The enormity of the prolepsis seems to have had its effect on the spelling of the manuscripts; but Robinson Ellis’s reading seems the only reasonable one; the name of Idomeneus was commonly used in literature to designate Crete.

The reasonable feeling about such anachronisms seems to be that, while they are improper in serious prose, they are not out of place in poetry; and the editors who write Idaeos or Sidonios to avoid the prolepsis just mentioned, fail to see the difference between prose and poetic diction. Tacitus’s prose is highly poetic, and often Virgilian in diction; in: per quae egeritur humus aut exciditur caespes (Ann. 1. 65. 10) he disdains to call a spade a spade. In his account of Augustus’s will (Ann. 1. 8. 3) editors are wrong in inserting the clause: urbanis quingenos. He spurns such painful accuracy, just as did Mr. Mantalini, who, when arrested on a suit of £1,527 4s. 9½d., replied: ‘The halfpenny be demd’.

I read in Horace:

‘Forum putealque Libonis

Mandabo siccis, adimam cantare severis.’

Hoc simul edixi (Ep. 1. 19. 8-10).

Edixi tamquam praetor, adds Orelli; and I am shocked at finding Horace dare to speak as praetor poetarum while Virgil is still alive. It is true that in Od. 1. 32. 1 the vulgate text gives: poscimur; but the best manuscripts and both Scholiasts read: poscimus, which is Bentley’s reading. I find that here too the majority of manuscripts give edixit, as do most of the manuscripts of Porphyrio; Acron has: 214 edixit, al. edixi. Courbaud (Horace, p. 319) is amply justified in condemning the reading edixit as implying a lack of taste and modesty inconceivable in Horace. Who, then, is the speaker? is it Ennius? and does the poet in: puteal Libonis make him foretell the construction of this monument by Scribonius Libo some twenty years after his death? We do not know for certain that it was this Scribonius Libo who constructed this puteal; but even if we did, puteal Libonis is the name in common use for it, and to expect scrupulous chronological accuracy in the words Horace assigns to the jovial Ennius seems a mistake.



To a union consisting of a noun and an adjective we have often a parallel equivalent consisting of a noun and a genitive depending on it. For amore tuo Virgil gives us: amore tui (Aen. 12. 29). Classical prose favours the adjective, but in poetry and in silver Latinity the genitive, which gives greater emphasis to the person, becomes more usual. The opposite seems true of patris and patrius; patris is usual in prose, but patrius is often used for patris in Virgil, as in: patrius amor (Aen. 1. 643) ‘love of his followers’, where it stands for a subjective genitive; in: patriae pietatis (10. 824) ‘love of his father’, where it is for an objective genitive; in: patriae pietatis (9. 294), where both subjective and objective are combined. He does not use paternus so often; but we see it in: arte paterna (8. 226) ‘through his father’s art’, and: regnis paternis (3. 121) ‘from the realms of his fathers’. Paternos (Prop. 3. 9. 37) is ‘of her sons’, and patrio (2. 7. 20) ‘of descendants’. We have fraternus in: fraterna caede (Aen. 4. 21) ‘murder by a brother’s hand’, where it is subjective, and: mortis fraternae (9. 736). In: fraterna morte (Geo. 3. 518) it is extended to a member of the same herd. The use of erilis for eri or erae, so characteristic of the Latin comedy, is rarer in golden Latinity; but we read: gressumque canes comitantur erilem (Aen. 8. 462) and: nisi erile mavis carpere pensum (Od. 3. 27. 63). In: tu si hic sis, aliter sentias (Ter. And. 310) hic is for ego; cf.: hunc hominem velles si tradere (Sat. 1. 9. 47). In: tu Maximus ille es (Aen. 6. 845) the use of ille helps us to understand that of ille for tu in oblique. We see words usually nouns becoming adjectives in: venator canis (12. 751), cognomine terra (6. 383), advena exercitus (7. 38), Tros Aeneas (12. 723). We read: Romula tellus (6. 877), Romulae gentis (Od. 4. 5. 1), pubes Dardana (Aen. 5. 119), Alphea flumina (Geo. 3. 180), Sirius ardor (Aen. 10. 273), manus Ausonia (8. 328), ignes Rutulos (9. 129), Lavinia arva (4. 236), Latium annum (Ov. Fast. 1. 1. 1), Thracius equus (Aen. 9. 49). We have the opposite use of Thraca for Thracia in: gemit ultima pulsu Thraca pedum (12. 335).

Very frequent is the use of an adjective formed from a proper noun instead of its genitive, as in: Herculei sacri (Aen. 8. 270), Herculea 216 umbra (8. 276), laudes Herculeas (8. 288), Circaeum iugum (7. 799), stirpis Achilleae (3. 326), coniugis Hectoreae (3. 488), Priameia virgo (2. 403), Ixionii rota orbis (Geo. 4. 484), Agamemnoniae phalanges (Aen. 6. 489). Turning to common nouns we have: animos agrestes (7. 482) for animos agrestium, the adjective standing for a genitive plural; and so in: multa caede (1. 471), praesentia ora (3. 174), victricia arma (3. 54), semiviro comitatu (4. 215), captivo sanguine (10. 520), sanguine sacro (5. 78), crudelia limina (3. 616), dente invido (Od. 4. 3. 16). We have past participles thus used in: caeso sanguine (Aen. 11. 82), manes sepultos (4. 34), fessos artus (3. 511), sceleratas poenas (2. 576), scelerato ex sanguine (12. 949). While Virgil does use hostilis for hostis or hostium, as in: hostilis facies (3. 407), terram hostilem (10. 489), he finds inimicus a far more forceful and telling adjective for this use, as we see in: castra inimica petunt (9. 315) the enemy’s camp, destined to prove fatal to them; and so in: inimica nomina (11. 84), inimicis ignibus (8. 375). We have an adjective from a different stem from the genitive in: inania regna (6. 269) for umbrarum, infelix vates (3. 246) for malorum, infernas sedes (8. 244) for inferorum, plumoso aucupio (Prop 4. 2. 34) for avium.

But at times the collective meaning of the plural genitive would find expression in a genitive singular; and in: quadrupedante sonitu (Aen. 8. 596) we have the adjective standing for a genitive singular of this kind; and so in: sapientem pascere barbam (Sat. 2. 3. 35) and: feminea in poena (Aen. 2. 584). But in the last example it is a particular woman, Helen, that is really meant; and so in: hostilem ad tumulum (3. 322) Achilles is meant, and in: moribunda ora (Prop. 3. 7. 56) Paetus. But: miser hiatus (ib. 52) is for: os hians miseri iuvenis; as fortia corpora (Aen. 8. 539) is for: corpora fortium virorum, and vicinum funus (Sat. 1. 4. 126) for: mors vicini hominis. So: merentes poenas (Aen. 2. 585) is for: poenas feminae merentis, i.e. Helen, and: vivo amore (1. 721) for: amore vivi viri, i.e. Aeneas. In: ingentem atque ingenti volnere victum (10. 842) ingenti volnere is for: ictu ingentis viri, again Aeneas. So: Gallica ora (Od. 1. 8. 6) is for: ora Gallicorum equorum; strictam aciem (Aen. 6. 291) for: aciem stricti ensis; postera tempestas (Sat. 1. 5. 96) for: posteri diei tempestas; piae terrae (Prop. 3. 7. 9) for: cadaveri pii fili. And so with the demonstratives: ea signa (Aen. 2. 171) is for: eius rei signa; ea cura (Ann. 2. 24. 5) for: eorum cura; has poenas (Aen. 7. 595) for: huius sceleris poenas; cursu illo (7. 383) for: cursu illius buxi; quo gemitu (2. 73) for: cuius gemitu.


The appositive of the possessive adjective is put in the genitive, as in: cum mea nemo scripta legal volgo recitare timentis (Sat. 1. 4. 23). So in: litora fraterna Erycis (Aen. 5. 24), Erycis fines fraterni (5. 630), though here Eryx, and not frater, is treated as the appositive. But in: sola militis commoda (Tac. Ann. 1. 26. 5), which follows this analogy, sola is treated like the possessive and militis is a collective, giving: militum solorum commoda as the real meaning of the phrase. Of the pair thus got in: sola militis, the first term only is expressed in: Areo iudicio (Ann. 2. 55. 2) for: Arei pagi iudicio. We read: sed fratres egregie concordes et proximorum certaminibus inconcussi (Ann. 2. 43. 7), where: fratres . . . inconcussi is shortened for: fraterna amicitia inconcussa, the opposite of the last example. So we have: hoc auro (Aen. 7. 245) for: hac aurea patera; flagrantem pinum (7. 397) for: flagrantem pineam facem; vina (9. 319) for: pocula vini.

Virgil often uses the ablative for the genitive, as in: fuso crateres olivo (Aen. 6. 225), vis alto volnere tardat (10. 857), ore orsa (11. 124). So in: ipse volans tenues se sustulit ales ad auras (5. 861) the adjective ales plainly stands, not for alarum, but for alis. So in: eo me solvat amantem (4. 479) Virgil gives us a poetical variety for: eius me solvat amore. In: ferreus somnus (10. 745) ferreus is not for ferri, but for ferro datus; just as in: conubia nostra (4. 213) nostra is for: nobiscum, in: propinquis nuntiis (Ann. 2. 58. 1) propinquis is for: ex propinquo adlatis; in: vipereum crinem (Aen. 6. 281) vipereum is not ‘such as vipers have’, but viperis constitutum ‘made up of vipers’. The adjective in: rotat ensem fulmineum (9. 442) seems for: fulminis fulgore et velocitate. And so in: vagina . . . eripit ensem fulmineum (4. 580), where the force seems proleptic. In: hanc (silicem) . . . dexter in adversum nitens concussit (8. 237) dexter is for dextra manu. In: aequius huic Turnum fuerat se opponere morti (11. 115) huic seems for: hac manu oblatae; as in: his mecum decuit concurrere telis (11. 117) his is for: hoc in campo coniectis. Just as we had flagrantem pinum for flagrantem pineam facem, here we have in: qua spe Libycis teris otia terris? (4. 271) otia teris for: tempora otio teris.

But we have the accusative also used for the genitive in such constructions as: certus iter (5. 2) with which compare: certus eundi (4. 554), ego pretium ob stultitiam fero (Ter. And. 610), non nihil ad verum conscia terra sapit (Prop. 2. 13. 42), avidum . . . in tempora faenus (Luc. 1. 181), neque pol consili locum habeo neque ad auxilium 218 copiam (Ter. And. 320), gens ferox et ingeni avidi ad pugnam (Liv. 7. 23. 6). So: auras invecta tenebat (Aen. 7. 287) seems for: cursum per auras invecta tenebat; and: frueris posteritate tua (Mart. 7. 47. 10) for: frueris gloria tua apud posteros.

We have seen how the genitive of the owner or possessor often is expressed by an adjective in poetry. Of the remaining Latin genitives most are partitive in meaning, and we find those with the nouns on which they depend substituted for nouns with adjectives in agreement in poetry, and in prose at times, when the adjective expresses number or extent. Virgil gives us the prose form: medio in antro (Aen. 3. 624), but: aulai medio (3. 354) and: castrorum et campi medio (9. 230). These probably follow the analogy of: in praerupti montis extremo (Sall. Jug. 37. 4); and when we compare this with: extremo . . . sub fine laborum (Geo. 4. 116) we see that Sallust’s: in extremo is probably shortened for: in extremo fine. The adjective thus converted to a noun is found most commonly as a neuter plural, as in: angusta viarum (Aen. 2. 332), per opaca locorum (2. 725), ardua terrarum (5. 695), scriptorum quaeque (Sat. 2. 3. 2); and it is often a superlative, as in: pelagi extrema (Aen. 8. 333), in ultimis laudum (Liv. 30. 30. 4), proxima maris (Tac. Ann. 3. 1. 3).

This neuter plural is so much the rule that we find it joined in apposition with the masculine plural in: ductores Danaum . . . prima virorum (Lucr. 1. 86); and in: tibi cuncta tuorum parebunt (Stat. Silv. 3. 3. 197) cuncta tuorum is clearly for omnes tui. Summa passes easily from a neuter plural to a feminine singular; for it is in the final sum that we attain to the highest figures. So it is a stage in this passage when we find Ovid in: summa ducum Atrides (Am. 1. 9. 37) making summa the appositive of a singular noun. But we find the adjective thus converted to a noun in the singular in: sancta dearum (Enn. Ann. 72, M.) and: asperrimo hiemis (Tac. Ann. 3. 5. 2). For superlatives governing a masculine genitive the masculine singular or plural is the more common, as in: haud Ligurum extremus (Aen. 11. 701), extremi . . . hominum Morini (8. 727). In: multos Danaum (2. 398) and: multos illustrium Romanorum (Ann. 3. 6. 1) the sense of degree in the adjective is lost, and we have mere poetic periphrases.

In such constructions the adjective is syntactically a noun; and we see it passing into a noun, when Tacitus writes: provisu periculi (Ann. 1. 27. 2) for: proviso periculo. And we often have nouns implying a high degree of quality taking the place of superlatives in 219 this syntax. Just as in English, so from the beginnings of Latin literature we find flos used for optimum, as in: ea tempestate flos poetarum fuit (Pl. Cas. 18), inde flos salis fiet (Cato R. R. 88. 2), in ipso Graeciae flore (Cic. N. D. 3. 82. 33), flos veterum virtusque virum (Aen. 8. 500). So for pessimum we find scelus, as in: scelus viri (Pl. Truc. 621), abine a me, scelus (feminae)? (Bacch. 1176), ubi illic est scelus qui perdidit me? (Ter. And. 607), artificis scelus (Aen. 11. 407). The metonomy of scelus for homo pessimus is so usual that Terence joins illic and qui with it as though it were a masculine substantive.

We read: Paegnium, deliciae pueri, salve (Pl. Pers. 204), flos delibatus populi Suadaeque medulla (Enn. Ann. 353, M.), and by analogy we find in like syntax nouns chosen to express a high degree of the quality they imply in: non mihi esse P. Lentuli somnum, nec L. Cassi adipes, nec Cethegi furiosam temeritatem pertimescendam (Cic. in Cat. 3. 16. 7) ‘neither need I fear the sleepy Lentulus, nor the fat Cassius, nor that rash madman Cethegus’. So in: P. Clodi furor (Mil. 3. 2), labor domus (Aen. 6. 27), urbis opus (5. 119), specus volneris (9. 700), oris hiatus (11. 680), minae murorum (4. 88), rotarum lapsus (2. 236), velorum alas (3. 520), astrorum ignes (3. 585), stipitis gravidi nodis (7. 507), maris pontus (10. 377), loricae moras (10. 485), regum colla minacium (Od. 2. 12. 12), blanditiae rosae (Prop. 4. 6. 72), militiae flagitia (Ann. 1. 27. 1). To the noun is joined an adjective with a strong comic effect of oxymoron in: satis spissum filum mulieris (Pl. Merc. 755). We have a like syntax in: Teucrum inertia corda (Aen. 9. 55) and: inmania pondera baltei (10. 496). We have this union paralleled with the ordinary prose idiom in: cum tales animos iuvenum et tam certa tulistis pectora (9. 249) ‘such youthful courage and hearts so resolute’.

But at times the noun thus constructed indicates extent just as does the adjective in: in extremo montis. So in: iuga silvarum (6. 256) ‘the wooded heights’, imo barathri gurgite (3. 421), caeli suspectus (6. 579), septem discrimina vocum (6. 646). At times it becomes a periphrasis, serving merely to denote a quality in the object described, as in: ardentes oculorum orbes (12. 670), aequore campi (7. 781), litoris oram (3. 396), auri metallum (8. 445), habitu vestis (8. 723), Ixionii rota orbis (Geo. 4. 484). When we come to uses like: donum virgae (Aen. 6. 409) or: Medorum hostes (Prop. 3. 9. 25) we have a parallel to multos Danaum, giving us a genitive of definition like vox insaniae, or urbem Patavi (Aen. 1. 247). 220 In: triginta magnos volvendis mensibus orbes (1. 269) we have a fine poetic periphrasis for triginta annos, arising from the substitution for annos of magnos volvendis mensibus orbes, where volvendis mensibus is probably a use of the ablative for the genitive. That magnus orbis gave the Roman the same idea as annus is clear from Lucretius’s words: multis solis redeuntibus annis anulus in digito tenuatur (1. 311), where to the ‘circles of the sun’ is opposed the ‘circlet on the finger’. We have the opposite of this syntax in: res animos incognita turbat (Aen. 1. 515) ‘the uncertainty of their state disturbs their minds’, as in: omni arte magistra (8. 442) for omnis artis praeceptis, and in: urbs capta ‘the capture of the city’, missus Hannibal in Hispaniam (Liv. 21. 4. 1) ‘the sending of Hannibal into Spain’.

Writing of: quisquis ingentes oculo inretorto spectat acervos (Od. 2. 2. 24) Professor Tyrrell (Latin Poetry, p. 190) thus censures Horace’s use of acervos: ‘Heaps of what? Of treasures, of course say the commentators. But Horace has not written “heaps of treasures”; he has only written “heaps”’. But if the Romans of Horace’s day habitually shortened auri acervos to acervos, then Horace was merely following usus, quem penes arbitrium est et ius et norma loquendi; and this usus Professor Tyrrell should have known. Horace uses acervus again in: addit acervo (Sat. 1. 1. 33), quid habet pulcri constructus acervus? (1. 1. 44), at suave est ex magno tollere acervo (1. 1. 51), cur . . . carae non aliquid patriae tanto emetiris acervo? (2. 2. 105), ex modico quantum res poscet acervo tollam (Ep. 2. 2. 190). It is true that he writes: aeris acervus et auri (Ep. 1. 2. 47), but where we should expect acervus auri, he has simply acervus. The scholiasts Acron and Porphyrio do not think it necessary to add a word of explanation of this use of acervus, showing how clear and obvious it was to them. Virgil too gives us: magnum alterius frustra spectabis acervum (Geo. 1. 158); Tibullus has: composito securus acervo despiciam dites despiciamque famem (1. 1. 77); Juvenal: e pleno tollatur semper acervo (6. 364); and Ovid: de multis grandis acervus erit (Rem. Am. 424), caeco . . . ademit acervo (Met. 1. 24). True, Juvenal writes: ingens stabat acervus nummorum, Spartana chlamys, conchylia Coa, etc. (8. 100 ff.) giving a formal and detailed list. Cicero writes: qui tantos acervos pecuniae capiat (Leg. Agr. 2. 59. 22), tanti acervi nummorum (Phil. 2. 97. 38). But this is rhetorical prose, which follows a higher standard than the sermo plebis. Speaking generally, we read: acervus frumenti, farris, turis, armorum, lapidum, caesorum, but not auri; then we have 221 simply acervus. And what is our word ‘treasure’ but θησαυρός ‘a store’? Treasures of what? Of gold, of course; but we too habitually omit the word ‘gold’.

How am I to translate: sed ruinae maximae modo iumenta cum oneribus devolvebantur (Liv. 21. 33. 7)? What is: ruinae maximae modo? ‘Ganz wie wenn Gebäude zusammenstürzen’ explains Weissenborn; but he reads maxime, following a second hand. He cites: ruinae modo turbabantur (Liv. 44. 41. 7) of the crumbling of a phalanx, and: deturbati ruinae modo praecipitantur (Tac. Hist. 4. 71. 6) of a hostile array hurled downhill. It is also associated with mountains in: inenarrabilis labor descendentibus cum ruina iumentorum sarcinarumque (Liv. 44. 5. 1). But where shall I find ruina used for ruina domus or ruina aedium? Deiphobi dedit amplam ruinam . . . domus (Aen. 2. 310) points to the opposite, as does: ea lapsa . . . ruinam cum sonitu trahit (2. 465). I turn to Catiline’s words: incendium meum ruina restinguam (Sall. Cat. 31 fin.); ruina here looks at first sight like ruina omnium. Horace gives me: si fractus inlabatur orbis, impavidum ferient ruinae (Od. 3. 3. 8), which looks formidable. When on: suspensa graves aulaea ruinas in patinam fecere (Sat. 2. 8. 54) there follows: nos maius veriti (v. 57), what does Horace mean by maius, of which the fall of the curtains gave warning? What but an earthquake? with which the ruina domus, of which the Comm. Cruq. speaks, would be involved. Interesting here is Virgil’s: caeli ruina (Aen. 1. 129), and especially Servius’s note: id est tonitribus, quorum sonus similis est ruinis. What can Servius’s ruinae be but earthquakes? and in the mountains the avalanche is the phenomenon that corresponds most nearly to the earthquake of the plains. Sir Wm. Smith thinks that: nivis casus (Liv. 21. 35. 6) is an avalanche; but this is a snowfall in the usual sense. Among the Alps avalanches are frequent, not in autumn or winter storm so much as at noonday in the heat of summer sunshine. I think we must translate: but like a mighty avalanche were rolled down beasts of burden with their loads. Ruina is a poetic term primarily for terrae motus.

But what of the syntax ruinis for ruinarum sono so obvious in Servius’s note? One of the most usual ellipses in Latin is that of aedes in: ad Dianae (Ter. Ad. 582), ad Vestae (Sat. 1. 9. 35), ad Castoris (Cic. Mil. 91. 33), prope Cloacinae (Liv. 3. 48. 5). In Livy we read: ubi nunc Vicae Potae est (2. 7. 12), where the use of the genitive as subject of est was so distasteful to Madvig that he supplied aedes. While in Greek this construction subsisted long both for 222 proper and common nouns, in Latin, while it was preserved longer for names of Gods, the genitive of names of men was early changed to suit the syntax of the context. So in: ab Andriast ancilla haec? (Ter. And. 461), if we followed the Greek syntax, we should write: ab Andriae, not: ab Andria, for: from the house of the Andrian woman. We have the like adjustment in: Mysis ab ea egreditur (And. 226), where ab ea is for: ab eius aedibus, in: a me nescio quis exit (Heaut. 510), quisnam hinc ab Thaide exit? (Eun. 545), a fratre quae egressast meo (Phorm. 732). Donatus’s note to And. 461 has led to confusion here; it reads: simpliciter dixit ‘ab Andria est’ pro ‘Andriae est’; nam ex usu sic dicere solemus. But when we look for examples of the use of ab with the ablative for the genitive in Terence’s Latinity, we find none. When we compare with this note Donatus’s note to Phorm. 732: ‘a fratre’ pro ‘a domo fratris’, we see that an ab has fallen out before Andriae; the note should read: ‘ab Andria est’ pro ‘ab Andriae est’.

We have seen how Terence, when it was the shrine of a god, wrote: ad Dianae, but when it was the house of a mortal, ab Andria. But we find Virgil writing not only: proximus ardet Ucalegon (Aen. 2. 312), but: formidatus nautis aperitur Apollo (3. 275), where Apollo, too, is for Apollinis aedes, and: attollit se diva Lacinia contra (3. 552) ‘opposite rises the fane of Juno Lacinia’. The analogy that led to this seems to take the poets of the time further in two directions: (1) it is extended from the shrine or house to the territory or occupation of the person; (2) it is extended to places and lifeless objects generally, becoming rarer as it is so extended.

So from: ab Euandro castris ingressus Etruscis (Aen. 10. 148), for: ab Euandri oppido, we proceed to: tum Cererem corruptam undis Cerealiaque arma expediunt (1. 177), where Cererem is for Cereris fruges, or to: et tandem Turnum experiatur in armis (7. 434), for: vires Turni experiatur bello, or to: extremus galeaque ima subsedit Acestes (5. 498), for sors Acestae; nil praeter Calvum et doctus cantare Catullum (Sat. 1. 10. 19), for carmina Calvi et Catulli; manet sub Iove frigido (Od. 1. 1. 25), for Iovis caelo frigido; magnos aequabunt ista Camillos iudicia (Prop. 3. 9. 31), for magnorum iudicia Camillorum; at memor ille matris Acidaliae paulatim abolere Sychaeum incipit (Aen. 1. 720), for memoriam Sychaei. From the name of a class like the Camilli to a nation the transfer is easy; and we have: Teucrum arma quiescant et Rutuli (12. 78); profectio Hannibalis in Oretanos (Liv. 21. 11. 13), for: in Oretanorum fines; at 223 Cappadoces in formam provinciae redacti (Ann. 2. 56. 4), for: civitas Cappadocum. From this the transition to local names is easy, as in: Hesperium Siculo latus abscidit (Aen. 3. 418), for: Siculo latere; ut Chio nota si commixta Falerni est (Sat. 1. 10. 24), for: ut Chii nota, etc.; ubi non Hymetto mella decedunt viridique certat baca Venafro (Od. 2. 6. 14-16); mare omne in Austrum cessit (Ann. 2. 23. 3), for: in Austri potentiam.

But this idiom passes from proper to common names. We read: quo tempore Vesta arsit (Ov. Fast. 6. 437), but: ubi sedulus hospes paene . . . arsit (Sat. 1. 5. 71) for: hospitis taberna. So we have: ille ducem haud timidis vadentem passibus aequat (Aen. 6. 263), for: ducis passus; nunc aequali tecum pubesceret aevo (3. 491), for: cum tuo aevo; nunc tertia palma Diores (5. 339), for: victor tertiae palmae (i.e. tertio loco); nec bonus Eurytion praelato invidit honori (5. 541), for: invidit ei, qui honore sibi praelatus esset; neque adversus externos militem quaeri (Tac. Ann. 1. 69. 4), for: studia militum excitari. And so for persons in: numquid ego illi imprudens olim faciam simile? (Sat. 1. 4. 136), for: illius facti; longe mea discrepat istis et vox et ratio (1. 6. 92); unde ego mira descripsi praecepta haec (2. 3. 33), for: cuius ex ore; octo aquilae . . . imperatorem advertere (Ann. 2. 17. 2), for: imperatoris animum; nisi quos corpora equorum eodem inlisa toleraverant (2. 24. 2), for: quorum vitam; hunc loquitur grato plurimus ore cliens (Mart. 7. 63. 8), for: huius nomen.

We read in Horace:

Purae rivus aquae silvaque iugerum

Paucorum et segetis certa fides meae

Fulgentem imperio fertilis Africae

Fallit sorte beatior (Od. 3. 16. 29-32).

Bentley was on the right line when he simplified this to: ager meus Sabinus fallit proconsulem Africae sorte beatior; but he found a comparison between ager and proconsul impossible, and changed fulgentem to fulgente. But the English use of Leicester for Robert Dudley might have enlightened him. We noticed that in English we use horse for horseman, but not horseman for horse; while the Roman uses horseman for horse as well as horse for horseman. So the Roman says: Ucalegon proximus ardet, when he means the house of Ucalegon, and he uses ager Horati here, when he means dominus agri, Horace himself. We say: ‘among the nobles of Elizabeth’s day Leicester was the most unhappy’ but we do not say: ‘Robert Dudley was the next town to be laid waste’. If the comparison had 224 been between Horace and Africa, instead of the farm and the proconsul, Bentley might have been more fortunate.

In: pastorem ad baculum possum curare (Prop. 4. 2. 39) what can pastorem curare mean? Most editors emend to: pastor me ad baculum possum curvare, on the principle, apparently, that, if we can’t translate what the ancients wrote, we can at least emend it to something that we can translate. Even this poor satisfaction they have hardly attained here; for: ‘as a herdsman I am able to bend to the staff’ does not seem brilliant. Could the statue of Vertumnus ‘bend to the staff’? When we compare the passage with: quin tu tuam rem cura potius quam Seleuci (Pl. Mil. 951), or: stultitiae videbatur alienam rem periculo suo curare (Sall. Jug. 83. 1), and recall the use of imperatorem for imperatoris animum in Ann. 2. 17. 2, or hunc for huius nomen in Martial 7. 63. 8, we may see that pastorem could be for: rem pastoris ‘the office of the herdsman’. This is not the place to deal with: ad baculum, though pastorem ad baculum is parallel to our ‘shepherd with his crook’. Returning to the manuscript reading, for we have no evidence that Propertius wrote aught but this in this verse, we may translate: ‘I can play the part of the herdsman with his staff’, which fits the context.

We read in Propertius:

et nova flamma

Luxit in obliquam ter sinuata facem (4. 6. 30).

Here: in obliquam facem is short for: in speciem obliquae facis, and we translate: ‘and a strange flame he flashed forth curving thrice into the appearance of a torch held aslant’—a description of the lightning flash in its zig-zag course that has no equal to my knowledge. We have like uses of this idiom in: fit strepitus tectis vocemque per ampla volutant atria (Aen. 1. 725) for: vocis sonum; quem falsa sub proditione Pelasgi . . . demisere neci (2. 83) for: falso sub crimine proditionis; labat ariete crebro ianua (2. 492-3) for: arietis crebro ictu; non laudis amor nec gloria cessit pulsa metu (5. 394) for: non laudis amor nec gloriae amor, a use of three for four, as are many others of the examples we have quoted. In: lucosque sub alta consulit Albunea (7. 82) it is not the name of the shrine, but that of the Sibyl, that is omitted; and lucos is not short for lucos Sibyllae, but is a metonymy for Sibyllam.



We read: extemplo Aeneae solvuntur frigore membra (Aen. 1. 92), to which Servius notes: extemplo ilico, statim. Et est augurum sermo. Templum enim dicitur locus manu designatus in aere, post quem factum ilico captantur auguria. Then to: se tollit ad auras (Aen. 2. 699), verbum augurum, qui visis auguriis surgebant e templo. Unde est extemplo. Again to: extemplo turbati animi (8. 4), extemplo aut subito aut re vera ex ipso templo, id est post divina arma commota (vide v. 3). This seems confirmed by the syntax of extemplo in Naevius’s verse: extemplo illo te ducam, ubi non despuas (apud Gell. 2. 19. 6), and in: quom extemplo ad forum advenero, omnes loquentur (Pl. Capt. 786), is adornat veniens domi extemplo ut maritus fias (Epid. 361), extemplo Libyae magnas it fama per urbes (Aen. 4. 173). We learn then that extemplo means ‘of a sudden’, or ‘at once’; after the Augur has laid out his templum in the sky and has viewed the auspices, the ceremony being thus complete, he dismisses those assembled by a formula of which extemplo was a part, and probably the conclusion. It is natural to think that this formula was: ilicet extemplo ‘you may go from the templum’. Extemplo is very common in Plautus, less usual in Terence; Virgil has it sixteen times, Cicero twice only, Horace and Caesar not at all.

In: credidi esse insanum extemplo ubi te appellavit Tyndarum (Pl. Capt. 559) the union extemplo ubi seems short for extemplo ubi extemplo, where ubi extemplo would be equivalent to ubi primum, and the whole phrase to simul atque (simul). And in: eaque extemplo ubi ego vino has conspersi fores . . . aperit ilico (Curc. 80) it is plain that extemplo ubi is felt as equivalent to simul atque (simul). But in: utque impulit arma, extemplo turbati animi, simul omne tumultu coniurat trepido Latium (Aen. 8. 4) extemplo and simul are both simple in use, being equivalents with the force of statim. We have simul paired with extemplo in like fashion in: extemplo, simul pares esse coeperint, superiores erunt (Liv. 34. 3. 2). In: extemplo teli stridorem (audiit), aurasque sonantes audiit una Arruns, haesitque (una) in corpore ferrum (Aen. 11. 863) we probably have the full 226 form of the sentence. So too in: dixit (extemplo) et extemplo . . . sensit medios delapsus in hostes (2. 376).

But we saw that extemplo ubi is equivalent at times to ubi primum, and so we have extemplo used with the same meaning as primum or primo in: et prudentiam quidem, non vim dictatoris extemplo timuit (Liv. 22. 12. 6), where Hertz changes non vim to novi, and Weissenborn brackets it. But it has this meaning in: nox inlataque fallunt lumina, et extemplo latuit mensura iacentum (Stat. Achill. 2. 88) and in: erubescit: quid respondeat, nescit; quid fingat extemplo non habet (Cic. Ros. Com. 8. 3). So, as is natural, we find cum extemplo for cum primum in: ne tu me ignores, quom extemplo meo ex conspectu abcesseris (Pl. Capt. 434), quom extemplo ad forum advenero, omnes loquentur (ib. 786), et saepe.

Very interesting is quam extemplo, the reading of the manuscripts in: nam qui amat, quod amat quam extemplo saviis sagittatis percussus est, ilico res foras labitur, liquitur (Pl. Trin. 242), quam extemplo hoc erit factum (Mil. 1176). The editors of course change it to quom extemplo. But beside quom in Latin we have the later and stronger form quando, which in archaic and classical Latin has superseded it as the interrogative in direct and oblique. What is this but quam-do ‘up to what time?’ where the -do is the do- in donec, and the de in quamde (Lucr. 1. 640) ‘up to what point?’, ‘how much?’. As quam is ‘to what degree?’, quamde seems ‘up to what degree?’. We have the feminine adjective in alias ‘at another time’; in quam extemplo the quam seems ‘to what time?’ for which quando is ‘up to what time?’ We have also tam for ‘up to that time’ in: tam modo (Pl. Trin. 609), the Praenestine for tantum modo.

To: ilicet obruimur numero (Aen. 2. 424) Servius’s note is: ilicet confestim, mox. Sane apud veteres ‘ilicet’ significabat sine dubio ‘actum est’. Origo autem significationis inde descendit: olim iudex ubi sententiam dixerat, si dare finem agendis rebus volebat, per praeconem dicebat ‘ilicet’ hoc est ‘ire licet’, id est acta et finita res est; and he cites in illustration: em tibi, rescivit omnem rem, id nunc clamat, ilicet (Ter. Ad. 791), actum est, ilicet, periisti (Eun. 54). To: ilicet (Phorm. 208) Donatus’s note is: semper ‘ilicet’ finem rei significat, ut actum est; and he adds: ilicet per syncopam; sic iudices de consilio demittebantur, suprema dicta cum praeco pronuntiasset ‘ilicet’, quod significat ire licet. Hand quotes Scaliger (ad Varron. 6. 2): extemplo verbum est sacrorum, ut ilicet iudiciorum. And yet Servius (ad Aen. 6. 231) tells us that the novissima verba closing the 227 ceremony of burning the dead were ilicet. Legal proceedings with the Romans were an outgrowth of sacral ceremonial; so we may conjecture that of the phrase ilicet extemplo, ilicet was transferred to proceedings in a court of law, and the formula used in closing a case there was: actum est, ilicet.

That Servius and Donatus were right in making it a shortened ire licet seems certain. The strong accent on the first syllable led to the loss of e in the second, and r was assimilated to the following l. But Charisius (p. 200, K.) writes: ilicet nunc pro ‘ilico’, id est statim; antiqui pro ‘eas licet’. Schneider thought that ilicet was for i licet. We have noticed that the infinitive with licet is primarily an infinitive used as imperative, and for this the subjunctive is a common equivalent. The real nature of ilicet seems plain in: ilicet parasiticae arti maximam malam crucem (Pl. Capt. 469), ilicet vadimonium ultro mi hic facit (Epid. 685), ilicet. quid hic conterimus operam frustra? quin abeo? (Ter. Phorm. 208).

It is also used with the sense of ‘all is lost’, as in: ilicet: mandata eri perierunt una et Sosia (Pl. Amph. 338); periit opinor. actum est, ilicet, me infelicem et scelestam (Cist. 684-5). Clearly ilicet with the sense of actum est is a metonymy resulting from the shortening of actum est, ilicet. We have it again in full in: actumst, ilicet, peristi (Ter. Eun. 54), and shortened in: ipsast; ilicet; desine; iam conclamatumst (ib. 347).

Charisius (p. 200, K.) cites from Afranius: an tu eloquens ilicet? and explains ilicet as subito vel extemplo; i.e. of the phrase ilicet extemplo, ilicet is here used with the force of both terms, usually expressed by extemplo. This seems its meaning in classical poetry, as in: fugit ilicet ocior Euro (Aen. 8. 223), fractas utinam tua tela sagittas ilicet (aspiciam) (Tib. 2. 6. 16), ilicet igne Iovis lapsisque citatior astris tristibus exsiluit ripis (Stat. Theb. 1. 92). The word is rare even in Latin comedy; Virgil has it five times, Cicero never.

Ilico belongs to the Comedy and to classical prose, as its quantity, īlĭcō, allows it no place but in iambic or trochaic verse. Servius in his note to ilicet (Aen. 2. 758) says that ilico like ilicet means the same as confestim; sed metri ratione variantur. Many of its uses in the Comedy show the phrase in its primary meaning, which approaches: eo ipso in loco. So in Naevius’s verse: septimum decimum annum ilico sedentes (ap. Non. p. 518, M.), cum quo irent nesciebant, ilico manserunt (Hemina ap. Non. l.c.), heus tu, asta ilico (Pl. Trin. 1059), tandem ilico adesdum (Mil. 1030). Nonius gives us ilico with 228 the sense of illo in: sed quam longe est, cum isti ilico (Turpil. ap. Non. l.c.), where isti ilico must be for istum in locum. This is the use of the locative, so common in Sanskrit, to denote limit of motion, which we see in: adveniens domi (Pl. Epid. 361) and: procumbit humi bos (Aen. 5. 481). For place Tacitus seems to return to its primary force in: castra metari in loco placuit (Ann. 1. 63. 7); but he uses it for time in: reus ilico defendi postulabat (ib. 13. 52. 2).

To: otiose nunc iam ilico hic consiste (Ter. Ad. 156) Donatus has this note: ilico modo locum, non tempus significat. At first sight one feels like taking modo here for hic—it is often for nunc—and taking the note to mean that ilico must here be joined with hic. But to: missast ancilla ilico obstetricem arcessitum ad eam (Ter. And. 514) he says: ilico quod Graeci dicunt αὐτόθεν, nam loci significatio est, etiam brevitatem temporis notans. But ilico here, which he would take as inde, the opposite of Turpilius’s use, is plainly temporal, just as in: percussit ilico animum (And. 125) and in Pacuvius’s verse: repudio auspicium; regrediendumst ilico (ap. Non. l.c.), ipse hinc ilico conscendit navem (Pl. Rud. 62), fugere e conspectu ilico (Ter. Hec. 182). Its transition from place to time is natural and easy, as we see in: te certo heri huc advenientem ilico salutavi (Pl. Amph. 714), where we may translate ilico either ‘on the spot’ or ‘immediately’. We find ilico coupled with extemplo in: quam extemplo hoc erit factum . . . ibi tu ilico facito ut venias . . . huc (Pl. Mil. 1176-7), tristes ilico, quom extemplo a portu ire nos cum auro vident, subducunt lembum (Bacch. 303-5). It is coupled with continuo in: nam postquam audivi ilico ex meo servo illam esse captam, continuo argentum dedi (Epid. 563-4). And like extemplo it takes the meaning of primum in: qui aequom esse censent nos a pueris ilico nasci senes (Ter. Heaut. 214), scivi equidem in principio ilico nullam tibi esse in illo copiam (Pl. Epid. 324).

We have already noticed the union: ilico hic (Ter. Ad. 156); we have it again in: ilico ante ostium hic erimus (Caecil. ap. Non. l.c.). We saw: ibi ilico (Mil. 1176), found again in: quin ibi ilico adsit (Merc. 362). We have also: istic sta ilico (Merc. 910) and: ilico intra limen istic adstate (Most. 1064), ibidem ilico puer abs te cum epistulis (Cic. Att. 2. 12. 2). Here we have four unions equivalent in meaning to: in hoc loco, in eo loco, in isto loco, and: in eodem loco, and pointing clearly to the primary force of ilico.

In enimvero, enim is from the stem eno- ‘that’ (cf. Skt. anena), of which it seems a locative; and the phrase was perhaps primarily shortened from enimvero tu dixisti ‘in that saidst thou truly’. Hand’s 229 painstaking treatment of the phrase is vitiated by his failure to see that enim and vero, standing both at times for enimvero, will have the same meaning. It is just as though in Talleyrand’s maxim: trop de zèle, we failed to recognize that trop is short for pas trop.

We find immo enimvero in: Pa. Incommode hercle. Ch. Immo enimvero infeliciter (Ter. Eun. 329), immo enimvero ego sum, inquam, Orestes (Pacuv. 365, R.), immo enimvero corpus Priamo reddidi (Ace. 667, R.). More usual is immo vero, as in: Pam. Nescis, Parmeno, quantum hodie profueris mihi. . . Par. Immo vero scio, neque hoc imprudens feci (Ter. Hec. 875-7), non igitur patria praestat omnibus officiis? Immo vero; sed ipsi patriae conducit pios cives habere (Cic. Off. 3. 90. 23). But we have immo enim with the same meaning in: Ch. Orandi iam finem face. . . Si. Immo enim nunc cum maxime abs te postulo atque oro (Ter. And. 821-3), Ch. Duras fratris partes praedicas. Pa. Immo enim si scias . . . magis id dicas (Eun. 354-6). We have also immo for immo vero in: An. Ubi? domin? Ch. Immo apud libertum Discum (Eun. 608), Dor. Hae quid ad me? Tox. Immo ad te adtinent et tua refert (Pl. Pers. 497). We have vero with like meaning in: Dixisti enim non auxilium mihi sed me auxilio defuisse. Ego vero fateor me quod viderim mihi auxilium non deesse, idcirco illi auxilio pepercisse (Cic. Planc. 86. 35). We have then immo enim vero, immo enim, immo vero with the same meaning; and for immo vero either immo or vero with this same force—indicating successive uses of one for a pair. When, as in: Dem. Tune es adiutor nunc amanti filio? Lib. Sum vero (Pl. Asin. 58), vero indicates simply assent, its use is probably simple, and not the result of shortening.

Hand (2, p. 405. 2) tells us that enim vero is used for tum vero. We read: tunc enim vero deorum ira admonuit (Liv. 2. 36. 6), enim vero tum Latini gaudere facto (2. 22. 6), and then: tum vero ingentem gemitum dat pectore ab imo (Aen. 1. 485). In: cum gladii abditi ex omnibus locis deverticuli protraherentur, enim vero manifesta res visa iniectaeque Turno catenae (Liv. 1. 51. 8) Duker wished to change enim vero to tum enim vero; and in: enim vero conclamant bonum ut animum haberent (24. 31. 1) Hand thinks enim vero for tum enim vero. These indicate tum enim vero, tum vero, and enim vero with the same meaning.

We read: sed enim vero cum detestabilis altera res et proxima parricidio sit, quid ad deliberationem dubii superesse? (Liv. 45. 19. 14), then: sed enim οἰκονομία si perturbatior est, tibi assignato (Cic. Att. 6. 1. 8), and: sed enim gelidus tardante senecta sanguis hebet (Aen. 5. 230 395). So in: mox Rhescuporis egredi fines . . . et resistenti vim facere, cunctanter sub Augusto . . . enim vero audita mutatione principis inmittere latronum globos (Ann. 2. 64. 6) enim vero is probably for sed by metonymy, as Schwartz assumed it to be, being shortened from sed enim vero.

We read: ceteri tribuni militum nihil contradicere. At enim vero Sergius Verginiusque . . . primo deprecari ignominiam, deinde intercedere (Liv. 5. 9. 3). Then: Ca. Ita faciam. Meg. At enim nimis longo sermone utimur (Pl. Trin. 806), Att. Adsentior, quoniam omnis haec in religione versatur oratio. M. At vero, quod sequitur, quomodo aut tu adsentiare, aut ego reprehendam, sane quaero (Cic. Leg. 2. 34. 14). Here at enim vero, at enim, at vero evidently bear the same meaning. And as we shall see, at is used for at enim.

Just as we had tum enim vero and enim vero tum, so we find enim vero and verum enim, as in: verum enim metuo malum (Ter. Phorm. 555). We have certe enim in: certe enim hic nescio quis loquitur (Pl. Amph. 331). But if verum is thus equivalent to vero, which denotes assent, whence comes the adversative meaning which verum almost always shows? The force of assent is plain in: tum Brutus . . . sed tu orationes nobis veteres explicabis? Vero, inquam, Brute (Cic. Brat. 300. 87). But just as vero shortened from immo vero shows a strong adversative force, so we find verum with this force in: merito maledicas mi, si non id ita factum est. Verum haud mentior, resque uti facta dico (Pl. Amph. 572-3). In a very few cases we find verum with the meaning of assent, as in: Ct. Men quaerit? Sy. Verum (Ter, Ad. 543), So. Facies? Ch. Verum (Heaut. 1013). But we find verum tamen, giving the adversative force usually conveyed by verum, in: consilium capit primo stultum, verum tamen clemens (Cic. Verr. 2. 5. 101. 39), nam quom pugnabant maxime, ego tum fugiebam maxime. Verum quasi adfuerim tamen adsimulabo, atque audita eloquar (Pl. Amph. 200). Most usual, however, is tamen with this meaning, as in: nec satis digna quoi committas primo partu mulierem. Tamen eam adducam? (Ter. And. 231), hic haedos depone; tamen veniemus in urbem (Buc. 9. 62). Tamen consists of the same accusative that we have in tam with the preposition en (= in) suffixed, and means ‘up to that point’, or ‘for all that’. We have the two forms in tametsi and tamen et si, evidently with the same meaning. The pair: verum tamen ‘in truth for all that’ is expressed commonly by tamen, but often by verum, which, standing for tamen by metonymy, will have its adversative force.


Cum maxime in many of its uses has been a puzzle to scholars. In some it presents no difficulty, as in: quom secundae res sunt maxime, tum maxime meditari secum oportet (Ter. Phorm. 241), qui cum maxime fallunt, id agunt ut viri boni esse videantur (Cic. Off. 1. 41. 13), and in Sosia’s words: nam quom pugnabant maxime, ego tum fugiebam maxime (Pl. Amph. 199) ‘when they were in the hottest of the fight, then I was in the hottest of my flight’. Nor in such uses as: immo enim nunc quom maxime abs te postulo (Ter. And. 823) is it difficult to find the meaning. For if we asked Sosia: quando maxime fugiebas? his answer would be: tum quom maxime pugnabant. Sed quando maxime timebas? Non tum, sed nunc quom maxime cum ero rem disputo; qui timeam ne meae inertiae poenas mihi sit dandum. And so in Terence’s: nunc cum maxime postulo ‘now it is when most of all I ask’. In: haec cum maxime loqueretur, sex lictores circumsistunt (Cic. Verr. 2. 5. 142. 54) nunc is implied in haec.

We have seen how extemplo is used for ilicet extemplo, tamen for verum tamen, enim vero for tum enim vero. So for nunc cum maxime we have cum maxime in: atqui quom maxime volo te dare operam ut fiat, verum alia via (Ter. Heaut. 788), where clearly: quom maxime volo means ‘now most of all I wish’. The meaning is still more obvious in: hoc quod futurum dico, cum maxime fit, et pars eius magna iam facta est (Sen. Ep. 120. 18), where Erasmus wanted to write nunc cum maxime. So in: verum tamen antiqua neglegimus: etiamne ea neglegemus quae fiunt cum maxime? quae videmus? (Cic. Har. Resp. 32. 15) ‘which are happening just now before our eyes’. Here Hand feels that cum maxime must be for nunc cum maxime. In: quae passus est reus . . . quae cum maxime patitur (Quint. 6. 1. 23) ‘what he is suffering just now’, Spalding will not admit Ernesti’s change to tum, but holds that the ellipsis is rather of: eo quando quidque agitur tempore; as the phrase is general—in universum proposita. While it is general, it is expressed as an absolute present, and such changes are not formed as Spalding supposes, but unconsciously and in the sermo plebis. Very clear seems the meaning in: quia nemo nostrum novit nisi id tempus, quod cum maxime transit (Sen. de Ben. 3. 3. 3) ‘the time which is just now passing’. The phrase here too is general, but is plainly for nunc cum maxime. For further examples see Cic. Verr. 2. 4. 82. 38; Tac. Hist. 1. 29 and 84; 4. 65. Priscian says that cum maxime is the Greek ἐπεὶ μάλιστα. In: cum maxime haec in senatu agerentur, (tum) Canuleius pro suis 232 legibus . . . ita disseruit (Liv. 4. 3. 1) and: cum maxime haec imperator . . . agerent, (tum) tubae cornuaque ab Romanis cecinerunt (30. 33. 12) tum is to be supplied as indicated.

But in: cum maxime haec dicente Gaio, puer . . . delapsus est (Petron. 54. 1) the principal verb is in a past tense, and cum maxime is for iam cum maxime. So too in: surgentibus cum maxime partibus honesta specie praetenderentur (Tac. Hist. 3. 4. 4) and: litora et lacus Campaniae cum maxime peragrantem (Ann. 3. 59. 4) ‘when he was just now strolling along the shores and lakes of Campania’.

From iam cum maxime the transition is easy to tum cum maxime; but we have already derived this phrase from Plautus’s verse, and though we meet it first in Livy, it is probably older than iam cum maxime. We read: castra amissa esse et tum cum maxime ardere (Liv. 40. 32. 1) ‘the camp was lost and was just then on fire’. See also Liv. 33. 9. 3; 43. 7. 8; Curt. 5. 7. 2. We have cum maxime for tum cum maxime in: fuit . . . vetus illa sapientia, cum maxime nascens, rudis (Sen. Ep. 95. 14) and: coeptantem cum maxime coniurationem disiecit (Tac. Ann. 4. 27. 2). Tum cum maxime does not seem to differ appreciably in meaning from tum maxime, which we read in: ne aut consulem tum maxime res agentem a bello avocarent (Liv. 27. 4. 2). See also Liv. 31. 18. 2; 7. 23. 6; Suet. Tib. 14. 4. We read: eo maxime tempore Abydum oppugnabat (Liv. 31. 14. 4), an extension of tum maxime. In: at vestis tamen illa sanguine madens ita repraesentavit imaginem sceleris, ut non occisus esse Caesar, sed tum maxime occidi videretur (Quint. 6. 1. 31) tum maxime seems to have the full force of tum cum maxime. If so, it would follow the same course of shortening that we see in: immo vero for: immo enim vero.

We read in Terence: amabat ut quom maxime tum Pamphilus (Hec. 115), to which Donatus’s note is: cum maxime pro nimis. But ut cum maxime is the phrase to be explained; and the meaning plainly is: Pamphilus was then in love to that degree in which he was when most in love. So in: domus celebratur ita ut cum maxime (Cic. ad Q. Fr. 2. 4. 6). We have cum maxime short for ut cum maxime in: quem armis oppressa pertulit civitas, ac paret cum maxime mortuo (Off. 2. 23. 7) ‘and obeys when dead as much as ever’. So in: video te, mi Lucili, cum maxime audio (Sen. Ep. 55. 11) ‘I hear you as clearly as when you were here’, intimos affectus meos tibi cum maxime detego (Ep. 96. 2) ‘as clearly as ever’.


We read in Terence: quin etiam insuper scelus . . . vestem omnem miserae discidit (Eun. 645). Quin etiam in its primary meaning ‘why not further now’ in longer clauses took on a force of asseveration ‘nay actually’, which we see here. We shall be content with a survey of its use in Virgil in its full and shortened forms. We have the full form in: quin etiam caeli regionem in cortice signant (Geo. 2. 269), ausus quin etiam voces iactare per umbram (Aen. 2. 768), quin etiam hiberno moliris sidere classem (4. 309), mortua quin etiam iungebat corpora vivis (8. 485). We have quin et in: quin et supremo cum lumine vita reliquit (6. 735) and: quin et avo comitem sese Mavortius addet Romulus (6. 778). We have quin only in: quin aspera Iuno . . . consilia in melius referet mecumque fovebit Romanos (1. 279), quin protinus omnia perlegerent oculis (6. 33), quin . . . idem orans mandata dabat (6. 115). We have etiam only in: dissimulare etiam sperasti, perfide, tantum posse nefas (4. 305), per scelus, ecce, etiam Troianis matribus actis exussit foede puppes (5. 793), ipse etiam Ascanius curvo direxit spicula cornu (7. 496).

We have: dicite, quandoquidem in molli consedimus herba (Buc. 3. 55) and: quandoquidem Ausonios coniungi foedere Teucris haud licitum (Aen. 10. 105). Quando in: fabor enim quando haec te cura remordet (1. 261) and: his se, quando ultima cernunt . . . parant defendere telis (2. 446) presents the same meaning. So with quandoquidem in: deos quaeso ut sit superstes, quandoquidem ipsest ingenio bono (Ter. And. 487), and with quando in: meus fac sis postremo animus, quando ego sum tuos (Eun. 196). This meaning of quando is so far removed from its usual temporal force, as in: Clit. Iam aderunt. Clin. Quando istuc erit? (Heaut. 238), that I am forced to regard in these passages quando as short for quandoquidem. I read in Cicero: quatenus autem sint ridicula tractanda oratori, perquam diligenter videndum est (De Orat. 2. 237. 58), but in Horace: mundus erit qua non offendat sordibus (Sat. 2. 2. 65), which Acron explains: quatenus non erit sordidus.

My attention has been called to saepe numero by my colleague, Prof. De Witt. Walde derives saepe from the root of saepio ‘I hedge’; but he does not discuss the question whether saepe or saepe numero is the older form. Saepe numero seems much more common in prose than saepe, and it is in prose that full forms are usually preserved. Is saepe here an imperative, as Prof. De Witt suggests? and is saepe numero ‘fence off from count’ and so ‘countless’, 234 as Prof. De Witt suggests? Probably saepio, like capio, had an older form saepere for the infinitive. Lucretius’s use of cupiret (1. 71) is significant of the tendency here. Saepe, shortened from saepe numero, was regarded as an adverb and developed by analogy the comparison: saepius—saepissime, and later the rare adjectival forms: saepior and saepis.

Numero is not used for saepe numero, nor is quidem for quandoquidem, nor tenus for quatenus. But the use of numero as an adverb in archaic Latin is interesting. Gesner thinks that in: numero mihi in mentem fuit (Pl. Amph. 180) numero is for: opportune. Ribbeck says the Codex Palatinus has: nunc vero, and will emend to: nunc verbo (Coroll. ad Trag. Fr. XV). But Nonius read numero, and tells us it is for cito (I. p. 571. L. M.). If it is for opportune it is probably shortened from numero modoque: cf. nil extra numerum fecisse modumque (Ep. 1. 18. 59). Nonius cites from Turpilius: numquam nimis numero quemquam vidi facere, quom factost opus (l.c.) and Festus (p. 170, M.) from Afranius: perfalsum et abs te creditum numero nimis, where numero is for cito. Numero seems the shortened form of nimis numero in: neminem vidi qui numero sciret quique scito opust (Naev. Trag. Fr. 61, R.), en umquam numero matri faciemus volup (Nel. Carm. 1, R.), ne istum numero amittas subitum oblatum (Acc. Trag. 144, R.), numero te expugnat timor (ib. 503, R.), numero inepti pertimuistis cassam terriculam adversari (Afran. 270, R), numero ac nequiquam egi gratias (ib. 312, R.), numero huc advenis ad prandium (Pl. Men. 287), Py. Perii. Pe. Haud etiam: numero hoc dicis (Mil. 1400), Pa. Nimium saevis. Sy. Numero dicis (Cas. 647), o Apella, o Zeuxis pictor, cur estis numero mortui (Poen. 1272). Varro (ap. Non. 352, M.) tells us that mothers in premature delivery prayed to Numeria, and to a child so born was given the name Numerius: quod qui cito facturum quid se ostendere volebat, dicebat numero id fore.

But in: neque sat numero mihi videbar currere (Turpil. 151, R.), quid ita numero venit? (Caecil. 2, R.), and perhaps in: ac discedens numero venire ait adulescentem (Varro ap. Non. l.c.) numero is simply for cito; as some think it is in: quae cum causa Musarum esse dicuntur volucres, quod et . . . cymbalis et plausibus numero redducunt in locum unum (Varro R. R. 3. 16. 7); though to me the context seems to point to the use of numero here for numero modoque: cf. verae numerosque modosque ediscere vitae (Ep. 2. 2. 143). This use of numero for cito is probably a shortening from cito numero, 235 for which cito is the usual shortening, just as continuo is the usual shortening for continuo cursu.

For sero ‘too late’ we have nimis sero in: tametsi incidamus oportet media, ne nimis sero ad extrema veniamus (Cic. Phil. 2. 47. 19). But as sero in itself involves the idea of defect, the use of sero for nimis sero came probably very early. That of cito for nimis cito seems to occur first in the third century A.D., as in: qui cito reorum causas audierat (Capitol. Aurel. 24. 2). From sequos ‘the follower’, which we have in heres secus, and pedisequus, we get the adverb secus in the same way as adversus and rursus. This adverb, usually opposed to recto and bene, we find in: etiam si secus acciderit (Cic. Fam. 6. 21. 2). Such a union as we find in: nobis aliter videtur: recte secusne, postea (Cic. Fin. 3. 44. 13) points to recto secus ‘inferior to the right’ as the old form from which it is shortened; but I have not found it. The form sequius or secius or setius points still more clearly to an older secius recto, which seems implied in: sed memet moror, quom hoc ago secius (Pl. Cist. 692) or: ratio talis sequius ceciderit (Afran. 293, R.).

We have the union tantum modo ‘so much by measure’ or ‘fully as much’ in: velis tantum modo, quae tua virtus (Sat. 1. 9. 54), where the ellipsis of: et tantum modo consequaris is evident. The union seems to imply an ellipsis of neque plus, and corresponds to our ‘only’, as we see in: unum hoc tantum modo neque praeterea quidquam notatum est (Suet. Tib. 11. 3), cum tantum modo potestatem gustandi feceris (Cic. Rep. 2. 28), pedites vero tantum modo humeris ac summo pectore exstarent (Caes. B. C. 1. 62). We have it shortened to modo in: potin ut semel modo, Ballio, huc cum lucro respicias? (Pl. Pseud. 264), uni modo gessi morem (Most. 200), hi unum modo quale sit suspicantur (Cic. Orat. 28. 9); and to tantum in: excepit unum tantum, scire se nihil scire; nihil amplius (Cic. Acad. Prior. 2. 74. 23), notus mihi nomine tantum (Sat. 1. 9. 3); and so usually in non modo and non tantum. We have it with si in: si modo est haec ars (Cic. de Orat. 2. 157. 38), si modo ego et vos scimus inurbanum lepido seponere dicto (Hor. A. P. 272), modo si licet ordine ferri (Ov. Trist. 2. 263).

We have it negatived in: modo non monies auri pollicens (Ter. Phorm. 68), which Donatus says is equivalent to μόνον οὔ. So too in: equum . . . modo non vivum (Val. Max. 8. 11. ext. 7) and: modo non loquentibus signis aperte monstrabat (Amm. 21. 14. 1). It has the meaning of ‘almost’, and seems short for tantum modo non sed 236 paene, where we have the first part shortened to modo non, but taking by metonymy the meaning of the omitted second term paene. We have tantum non with the same meaning in: nam cum vineae tantum non iam iniunctae moenibus essent (Liv. 5. 7. 2), tantum non statim a funere (Suet. Tib. 52. 1). In: peccare fuisset ante satis, penitus modo non genus omne perosos femineum (Aen. 9. 141) modo non omne is clearly for paene totum.

We have: tantum quod non in: tantum quod hominem non nominat, causam quidem totam perscribit (Cic. Verr. 2. 1. 116. 45) ‘he all but names the man, the case in sooth he fully delineates’. We have the positive of this in: tantum quod ex Arpinati veneram (Fam. 7. 23. 1) ‘I was just come from Arpinum’, and: tantum quod ultimam imposuerat Pannonico bello Caesar manum (Vell. 2. 117), de navi Alexandrina, quae tantum quod appulerat (Suet. Aug. 98. 2), Iuliam primum Marcello . . . tantum quod pueritiam egresso . . . dedit (ib. 63. 1). Older than tantum quod seems modo quod, which the manuscripts give us in: modo quod accepisti, haud multo post aliquid quod poscas paras (Pl. Asin. 168) ‘you have just got something; right after you get ready for a new request’. Goetz and Loewe have changed the reading to modo quom, as the phrase is isolated, but this method is most unsafe. Probably there existed a: modo quod non, but I have not seen it.

Donatus’s note to: modo dolores, mea tu, occipiunt primulum (Ter. Ad. 289) is: evidenter hic modo temporis praesentis adverbium est. Priscian (18. 168, K.) to: Ge. Modo apud portum . . . An. Meumne? (Ter. Phorm. 198) says: ‘modo’ dixit pro ‘nuper’, and to: modo ait, modo negat (Eun. 714) pro ‘nunc ait, nunc negat’. We have: advenis modo? (Hec. 458) ‘are you just now arriving?’ where modo seems short for nunc modo—a union that would be parallel to: tempori modo (Pl. Men. 1020) ‘just in time’ or ‘semel modo’ in: nam ter sub armis malim vitam cernere quam semel modo parere (Enn. F. 269, M.). Such pairs as we find in: nunc quereretur eundem accusatorem . . . ac iudicem esse, modo vitam sibi eripi (Liv. 8. 32. 9) or: nam modo ducebam . . . pisces, nunc in mole sedens moderabar . . . (Ov. Met. 13. 922) point to the existence of such a union; but they may arise from a syncretism of nunc . . . nunc, and modo . . . modo. True, Ribbeck reads modo nunc for modo non in Aen. 9. 141, following Ed. Ven. I, but against the manuscripts and the scholia of Servius, Charisius, Acron, Porphyrio, and Arusianus; but I have found no sound authority for nunc modo or modo nunc. In: iam 237 modo iam possim contentus vivere parvo (Tib. 1. 1. 25) iam modo iam has been emended to iam modo nunc; but in: iam melior iam (Aen. 12. 179) and: iam puto iam (Ov. Trist. 1. 1. 44) we have parallels to iam modo iam, and there is no manuscript authority for the change.

But Tibullus’s phrase iam modo may be the union that lies behind this use of modo. Interesting here is: ilico hic ante ostium: tam modo, inquit Praenestinus (Pl. Trin. 609), where evidently tam modo is Praenestine for ilico. Evidently the Praenestines used tam modo where Tibullus uses iam modo, which the Romans usually shortened to modo. We must take tam here in a temporal sense, as we took quam in Plautus’s quam extemplo; it will mean ‘up to this point of time’ and tam modo will be ‘just now’. Nunc is so often joined with iam in nunc iam that it is not surprising to find Terence’s: modo ait, modo negat paralleled by Ovid’s: nunc huc, nunc illuc . . . curro (Her. 10. 19) and Virgil’s: iamque hos cursu iam praeterit illos (Aen. 4. 157).

In prose we have commonly the union tum . . . tum in this sense, as in: aestus maritumi, tum accedentes, tum recedentes (Cic. N. D. 2. 132. 53), (qui) tamquam machinatione aliqua tum ad severitatem, tum ad remissionem animi, tum ad tristitiam, tum ad laetitiam est contorquendus (De Orat. 2. 72. 17). So we have a union of modo and tum in this sense in: (sol) modo accedens, tum autem recedens (N. D. 2. 102. 40). Other such unions are: nunc adiutor Decimi Bruti . . . mox eiusdem proditor (Vell. 2. 63), modo hos obsidebat montes, paullo post ad illos transgrediebatur (Val. Max. 7. 4. 5), which help us to understand the future use of modo in: domum modo ibo (Ter. And. 594). Following the use of modo for nuper, as in: qui modo felices inter numerabar amantes (Prop. 1. 18. 7), we have tantum for nuper in: serta procul tantum capiti delapsa iacebant (Buc. 6. 16). The frequent use of modo in this sense has led to the extension of its meaning to considerable lapses of time, e.g. to nearly a century in: modo enim hoc malum (i.e. avaritia) in hanc rem publicam invasit (Off. 2. 75. 21).

We read: nunc misero mihi demum exitium infelix, nunc alte volnus adactum (Aen. 10. 849-50), where we have three for four, nunc demum . . . nunc for nunc demum . . . nunc demum. This may indicate the starting-point for the use of enim for enim vero, of verum for verum tamen, of immo for immo vero. The same is true for pairs like nunc . . . nunc, modo . . . modo, tum . . . tum and the like. 238 In: una omnes fecere pedem, pariterque sinistros, nunc dextros solvere sinus (5. 830-1) it is plain that the pair nunc . . . nunc is expressed by a single nunc. So with modo . . . modo in: interea cognitis insidiis Artabanus tardari metu, modo cupidine vindictae inardescere (Ann. 6. 32. 2). Probably this ellipsis accounts for the construction of dum, as we commonly see it in: dum civitas erit, (dum) iudicia fient (Cic. Rosc. Am. 91. 32); for dum is an accusative of duration ‘for the time’, or ‘for that time’; and we may translate Cicero’s sentence: ‘during the time our state will exist, during that time courts will be held’. We say in English: ‘the moment he comes, I go’, which is short for: ‘the moment he comes, that moment I go’. So: ‘while we wait, he works’ is short for: ‘the while we wait, that while he works’. We have no longer dum expressed in the principal clause excepting as an enclitic, as in exspectadum; but its place is taken by an adverb of duration, as in: ego hic tantisper, dum exis, te opperiar foris (Pl. Most. 683), or by a temporal clause, as in: Tityre, dum redeo, brevis est via, pasce capellas (Buc. 9. 23). But in: tu modo, dum lucet, fructum ne desere vitae (Prop. 2. 15. 49) modo . . . dum seems the usual form of distribution of dum modo . . . dum modo.

We have this union dum modo ‘exactly so long’ expressed once only in: dummodo morata recte veniat, (dummodo) dotatast satis (Pl. Aul. 239) ‘exactly so long as she shall be of right character, (exactly so long) is she dowered enough’. So in: quare sit summa in iure dicundo severitas, dummodo ea ne varietur gratia (Ad Q. Fr. 1. 1. 20. 7) ‘wherefore (exactly so long) let there be the utmost strictness in administering justice; exactly so long let it be perverted by no favour’. It is shortened to modo dum in: mea nil refert dum potiar modo (Ter. Eun. 320) ‘exactly the time I shall be master, exactly so long it will make no difference to me’. So in: nec volgi cura tyranno, dum sua sit modo tuta salus (Val. Flacc. 5. 265) ‘the full while his own health is secure, exactly so long has the ruler no care for the common herd’. Dum is short for dum modo in: dum ne tibi (segnior) videar, non laboro (Cic. Att. 8. 11 B), id faciat saepe, dum ne lassus fiat (Cato R. R. 5. 4), quidvis cupio, dum ne ab hoc me falli comperiar (Ter. And. 902), multa . . . bello passus, dum conderet urbem (Aen. 1. 5), vel patiare licet, dum ne contempta relinquar (Ov. Her. 3. 81). We have modo for dum modo in: manent ingenia senibus, modo permaneat studium et industria (Cic. Sen. 22. 7), Tertia aderit, modo ne Publius rogatus sit (Fam. 16. 22), modo Iuppiter adsit, 239 tertia lux classem Cretaeis sistet in oris (Aen. 3. 116), o valeant fruges, ne sint modo rure puellae (Tib. 2. 3. 67), modo postulat ut secum stemus, modo ne intersimus armis, contentum ait se esse (Liv. 32. 21. 5).

We have dum as an enclitic in the principal clause in: adesdum; paucis te volo (Ter. And. 29). In the second clause its place is taken by paucis, as we saw it replaced by tantisper in Pl. Most. 683. That it is for dum modo is probable, for we shall find modo used in exactly the same way. The meaning is ‘I want you for a few minutes; give me your attention for that full time’. So in: manedum sodes (Hec. 844), tangedum . . . idum, Turbalio, curriculo, adfer celeriter duas clavas (Pl. Rud. 796-8), sed vero sinedum petere (Truc. 628). As is clear from the last two examples, it quickly tends to become the equivalent of our ‘pray’ or ‘please’; and that is its usual meaning, especially in agedum and cedodum. This is because, when it is shortened from the pair dum modo, it may take the meaning of modo ‘fully’ by metonymy, as well as the meaning of dum modo by synecdoche.

We have modo used like the enclitic dum in: cedo modo mihi vidulum istum (Pl. Rud. 1127) ‘give here to me for a time that trunk you have’. So in: mane modo, etiam percontabor alia (Men. 922), sine modo ego abeam (Pseud. 239), tu modo, dum lucet, fructum ne desere vitae (Prop. 2. 15. 49). In the first three examples modo seems to have the meaning of dum ‘for a time’, but it soon passes to a mere particle of entreaty, as in: cave modo, ne gratiis (Pl. Asin. 5), sequere hac modo (Men. 562), quin tu i modo (Trin. 583), vos modo, inquit, parcite (Phaedr. 2. 8. 8), modo fac . . . ne quid aliud cures hoc tempore (Cic. Fam. 16. 11. 1), tu modo posce deos veniam (Aen. 4. 50). Being the second term of the union, it is used far more frequently than dum. Agedum is often a mere interjection, so we need not be surprised at finding dum in: ehodum ad me (Ter. And. 184).

We find dum joined to etiam, evidently with the meaning of modo ‘fully’, showing that it is short for dummodo; as in: dissimulabo, hos quasi non videam, neque esse hic etiamdum sciam (Pl. Mil. 992), neque etiamdum scit pater (Ter. Heaut. 229), where etiamdum means ‘even now fully’. So in: integra etiamdum domo sua (Ann. 1. 3. 1), where all my editors emend to tum. In: primumdum, si falso insimulas Philocomasium, hoc perieris (Pl. Mil. 297) primumdum seems short for primumdum omnium, which we have in Pl. Trin. 98 ‘first of all and to the full’.


Nedum ‘far less’ or ‘far more’ presents a more complex problem. Dum here too seems to have the meaning ‘to the full’. In: satrapa si siet amator, numquam sufferre eius sumptus queat, nedum tu possis (Ter. Heaut. 454) the contrast between the governor and the farmer Menedemus adds the idea of ‘less’ to nedum. Often the negative in the first clause is only implied, as in: mortalia facta peribunt, nedum sermonum stet honor et gratia vivax (A. P. 69), where in peribunt non stabunt is implied. When no negative is expressed or so implied in the first clause, nedum changes its meaning to ‘far more’, as in: quae vel socios, nedum hostes victos terrere possent (Liv. 45. 29. 2), ornamenta etiam legioni, nedum militi satis multa (Val. Max. 3. 2. 24 fin.). When by transposition nedum begins the sentence, as in: nedum hominum humilium, ut nos sumus, sed etiam amplissimorum virorum consilia ex eventu, non ex voluntate, a plerisque probari solent (Cic. Att. 9. 7 A init.) the meaning of nedum will depend on whether the following clause is positive or negative.

Dum seems to have its proper meaning of ‘during the time’ in interdum ‘at intervals in the time’, as in: interdum gremio fovet (Aen. 1. 718). In vixdum necdum, nondum, nihildum, nullusdum, nemodum it has the sense of ‘for the full time’, just as our ‘yet’, which we use to translate it, and the German jetzt (M.H.D. je-zu-o) seem to mean ‘right up to now’. So in: vixdum dimidium dixeram (Ter. Phorm. 594) ‘hardly through all the time I had, had I said the half’, necdum sua forma recessit (Aen. 11. 70), fuga ab nulladum parte erat (Liv. 7. 33. 13).

In: cuius modo rei nomen reperiri poterat, hoc satis esse ad cogendas pecunias videbatur (Caes. B. C. 3. 32. 2) modo seems for cumque. In:

Quid iudicare cogitet livor modo,

Licet dissimulet, pulcre tamen intellego (Phaedr. 4. 21. 1-2)

quid . . . modo seems to be for quidcumque. We have the union modo cumque in: et clamant ‘merito’ qui modo cumque vident (Ov. Am. 2. 14. 40), and from the shortening of this union comes by metonymy the use of modo for cumque.

We have also modo suffixed to temporal adverbs that later become prepositions. We have postmodo ‘some time after’ in: me abs te inmerito esse accusatam postmodo rescisces (Ter. Hec. 208), inmeritis nocituram postmodo . . . natis fraudem (Od. 1. 28. 31), publicum in praesentia dedecus, postmodo periculum (Liv. 2. 43. 8), postmodo nativa conspiciere coma (Ov. Am. 1. 14. 56). Modo here is not the 241 simple modo ‘by measure’ but the union dummodo ‘for a full lapse of time’, and postmodo is ‘after the fulfilment of an interval’. Dum, being the accusative, is used to express interval, as is the accusative in: aliquot ante annos (Suet. Caes. 12), and as modo in postmodo expresses interval, for it too we have the accusative in postmodum, used repeatedly by Livy, as in: saepe ex iniuria postmodum gratiam ortam (1. 9. 15), ne postmodum flecti precibus . . . posset (2. 1. 9). To this we may relate praemodum, which Gellius (6. 7. 12) tells us Livius used in his Odyssey quasi admodum in: parcentes praemodum. He adds: quod significat ‘supra modum’ ‘above measure’; but prae does not govern the accusative, and modum is for modo, so that praemodum will mean ‘right in the front rank’.

We usually find propemodum ‘right near’, as in: propemodum quid illic festinet sentio (Pl. Trin. 615), verum propemodum iam scio, quid siet rei (Men. 764). All codices but the Ambrosian palimpsest give us: pol ego propemodo (Pseud. 276), and three of the best codices give: propemodo quis successisset (Liv. 24. 20. 11). This form was adopted by Ritschl as the reading of the best manuscripts in: Me. Tenes iam? Ca. Propemodo (Trin. 780). Brix has returned to propemodum as in his belief the older form, but the opposite is the truth, and we can understand propemodum only through the older propemodo.

Admodum is not: ad iustum modum, as Hand thought, but means ‘quite’ or ‘just’ being from modo ‘exactly’ and ad, which is still an adverb here, ‘to the extent’. Jordanes for admodum usually writes ammodum, but the Thesaurus tells me he writes also ammodo (= admodo). So amodo is from the adverb a ‘from there’ and modo ‘fully’. It is used in Christian writers and in the Vulgate to translate ἀπ’ ἄρτι in: non me videbitis amodo, donec dicatis: benedictus qui venit in nomine domini (Matt. 23. 39), and to translate ἀπὸ τοῦ νῦν in: amodo videbitis filium hominis sedentem a dextris virtutis dei (ib. 26. 64). The fact that amodum is not found as a second form of amodo, but that admodo is as a second form of admodum, leads plainly to the conclusion that modo is the older form of this suffix, a conclusion which confirms its connexion with the suffix dum, so strongly supported by the meaning of modo in postmodo and propemodo.

So with commodo and commodum, and at times with commode, where commodo and commodum are parallel to admodo and 242 admodum; commode is another story. Com is the same word as quom, but is here used in the meaning in which it passes to the prepositions. Illa mecum exit is evidently a later way of saying: illa tum exit cum ego exeo; and the com in commodo will have the meaning expressed by tum . . . cum in our phrase; which is in English ‘at the time’. Commodo and commodum will mean ‘exactly at the time’. When we compare: et commodo eccum exit (Titin. 64, R.), ecce autem commodum aperitur foris (Pl. Mil. 1198), commode ipse exit Lesbonicus cum servo foras (Trin. 400), where only A has commodum, all other manuscripts commode, we see plainly that all three have the same meaning. Commodo, the oldest, is rare; Plautus has it in: incommoditate abstinere me apud convivas commodo conmemini (Mil. 644), and Macrobius in: commodo adsunt feriae (Sat. 1. 2. 1). Commodum is far more usual; and to: illa sese interea commodum huc advorterat (Ter. Eun. 343) Donatus’s note is: tantum quod vel ipso eodemque tempore; and we can best catch its genuine meaning in: attrahitur Lollius commodum cum Apronius e palaestra rediisset (Cic. Verr. 2. 3. 61. 25) (commodum cum) ‘at the very time when’. We have it also in: ad te hercle ibam commodum (Pl. Cas. 593), commodum ad te dederam litteras (Cic. Att. 10. 16. 1), quos Horus ingredientes commodum consecutus comitabatur (Macrob. Sat. 1. 7. 3). But in: ad aquam praebendam commodum adveni domum (Pl. Amph. 669) commodum is plainly the acc. sing. neut. of commodus, with the meaning ‘opportune’. And in return we find commode, the usual bearer of this meaning, used with the meaning of commodum, ‘just at the time’ in: sed quos perconter commode eccos video astare (Rud. 309), emerseram commode ex Antiati in Appiam (Cic. Att. 2. 12. 2).

We have in:

vel dic quid referat intra

Naturae fines viventi, iugera centum an

Mille aret. ‘At suave est ex magno tollere acervo.’ (Sat. 1. 1. 49-51),

a clear instance of the use of at for the prose at enim or at vero.

In the noteworthy hyperbole:

portis alii bipatentibus adsunt

Milia quot magnis umquam venere Mycenis (Aen. 2. 331),

Servius’s note is: umquam pro quondam, a translation neglected by our editors. Umquam is in itself indefinite; and is used either in 243 a general sense, as in: quod si numquam oritur, ne occidit quidem umquam (Cic. Rep. 6. 25), or of the past, as in: quod nemo umquam homo antehac vidit (Pl. Amph. 566), or of the future, as in: cave posthac . . . umquam istuc verbum ex te audiam (Ter. Heaut. 1031). But the adverbs of time standing in union with umquam are often omitted, as is the way with pairs, e.g. in: mihi si umquam filius erit (ib. 217), where umquam is for umquam post, or in: plusque amat quam te umquam amavit (Pl. Epid. 66), where it is for umquam ante. In Aen. 2. 331 of the union umquam ante by metonymy umquam is used for ante. Hence Servius’s note.

In a note to: verum hodie numquam monstrabo (Ter. Ad. 570) Donatus is in doubt whether hodie is superfluous, or numquam hodie is for nullo tempore huius diei. He compares: numquam omnes hodie moriemur inulti (Aen. 2. 670), to which Servius’s note is: numquam pro non; and he further cites: numquam hodie effugies (Buc. 3. 49). Donatus’s note to: Sy. Nilne in mentemst? Ct. Numquam quicquam (Ter. Ad. 528) is: numquam quicquam, παρέλκον pro numquam; id est non. Here we have the union numquam quicquam (= ne umquam quidquam), which in the passages just cited has been shortened in form to numquam, but in meaning to ne quidquam. In: numquam factumst (Pl. Amph. 700), numquam te patiar perire (Men. 1010), hic quidem me numquam inridebit (Capt. 657), numquam dum adero, hic te tanget (Ter. Ad. 157), vi numquam eo subiri potuit (Liv. 3. 23. 4) numquam seems to have the full meaning of the pair numquam quicquam, of which Donatus says: numquam habet plus negationis quam non (ad And. 384).

To: numquam etiam fui usquam quin me amarent omnes plurimum (Ter. Eun. 1092) Donatus has this note: et numquam usquam pro non usquam, id est nusquam. Hand compares with this: mobilis enim et inquieta homini mens data est. numquam se tenet (Sen. ad Helv. 6. 6), where numquam seems rather of place than of time. When we compare this with Donatus’s note, we have an exact parallel to the shortening of ne umquam quidquam noted above, viz. a shortening of ne umquam usquam to numquam in form, but to nusquam in meaning. We may further compare: nusquam equidem quidquam deliqui (Pl. Men. 780), nusquam abero (Aen. 2. 620), where nusquam seems short for numquam usquam. In: quos numquam quisquam neque vocat neque invocat (Pl. Capt. 76) numquam seems short for numquam quoquam, and nusquam is used for nequoquam in: nusquam abeo (Ter. Ad. 246).


I need hardly dwell on the common use of one term for two in such pairs as tam . . . quam, tum . . . quom, is . . . qui. But it is well to recognize clearly how we express in English the resultant quam, which bears the meaning of the union tam . . . quam. We have this resultant in: concede huc, mea gnata, ab istoc quam potest longissime (Pl. Men. 834) ‘withdraw hither, my daughter, from that man the farthest possible’. We do not translate this quam by a relative, but by a demonstrative. So in: huic mandat, ut exploratis omnibus rebus ad se quam primum revertatur (B. G. 4. 21) ‘he instructed this man after examining everything to return to him the earliest (he could)’. And so with the relative in: quibus auditis liberaliter pollicitus (est) (l.c.) ‘having heard them he made them generous offers’. And now we see that commodo (= quom + modo) means to us ‘just at the time’.

But this expression of two related terms by one is so wide in its scope that I will be content with one or two examples more. We read: castra metari in loco placuit, ut opus et alii proelium inciperent (Ann. 1. 63. 7) for: ut opus alii et alii proelium inciperent. It was a like omission we assumed in: dum civitas erit, (dum) iudicia fient. It sometimes escapes the student, as in: si essent omnia mihi solutissima, tamen in republica non alius essem atque nunc sum (Cic. Fam. 1. 9. 21) for: non alius essem atque non alius nunc sum. It is plainer when we turn to the equivalent: idem sum atque (idem) semper fui. We find Tacitus resorting to other means of avoiding like repetitions, as in: commotis per haec mentibus et inter se suspectis (Ann. 1. 28. 8) for: aliis aliis suspectis; and in: cur abstinuerit spectaculo ipse, varie trahebant (Ann. 1. 76. 6) for: alio alii trahebant.

Not that poetry is necessarily averse to repetitions, as Bentley assumed to be an almost constant rule. It is of the very essence of poetic diction that it follows no constant rule; its canon is the quest of variety. In:

Teque dum procedis Io triumphe!

Non semel dicemus, Io triumphe!

Civitas omnis, dabimusque divis

Tura benignis.

Te decem tauri, totidemque vaccae,

Me tener solvet vitulus (Od. 4. 2. 49-54),

the use of te in two successive stanzas for two different persons, for Augustus the triumphant leader, and for Julus Antonius who with Horace is a spectator of the triumph, has presented difficulty to 245 readers. It seems to Mr. Wickham ‘unlike Horace’s finished workmanship to put the same pronoun in an emphatic place in two consecutive stanzas, when the subjects to which it refers are wholly different’. But we have seen that the full and balanced expression of a pair, such as tu . . . tu referring to Augustus, is just what Latin poetry will usually avoid; and when we meet te . . . te balanced in such fashion, there is reason to suspect what is here the truth, that we have not a repetition. But I am laying down no rule; the threefold repetition of tu . . . tu . . . tu in Od. 3. 21. 13-17, all three referring to Bacchus, would at once refute it.



Of all Latin verbs tollo presents the greatest variety in form and meaning. Its variety in meaning is evident in the famous pasquil on Nero:

Quis negat Aeneae magna de stirpe Neronem?

Sustulit hic matrem, sustulit ille patrem (Suet. Ner. 39),

‘the Trojan chief of old bore away his father, this prince of ours has made away with his mother’. It is interesting to see how the same verb comes to mean ‘to save’ and ‘to destroy’. But this is merely the cumulation of a curious series of meanings.

The verb has a variety all its own in its inflexion too. Naturally we should expect to find as its principal parts: tollo—tollere—tuli—(t)latum. Persius has tolli for the perfect in: sorbitio tollit quem dira cicutae (4. 2), his periphrasis for Socrates. We find tetuli repeatedly, as in: pedem nemo intro tetulit (Pl. Most. 471), tetuli ei auxilium (Rud. 68), ibo, hanc tetulero intra limen (Cist. 650), numquam huc tetulissem pedem (Ter. And. 808), incepi, dum res tetulit (ib. 832), ubi forte ita se tetulerunt semina aquarum (Lucr. 6. 672), ad Idae tetuli nemora pedem (Catull. 63. 52), si reditum tetulisset (id. 66. 35). All examples of tetuli are in meaning perfects of fero, and tuli is almost always a perfect of fero, and not of tollo; it is used, however, as the perfect of tollo in Suetonius, as in: ex Scribonia Iuliam . . . tulit (Aug. 63), qui . . . quattuor liberos tulerat (Tib. 47), ex ea novem liberos tulit (Cal. 7). In: caelo supinas si tuleris manus (Od. 3. 23. 1) tuleris seems the poetic shortening for sustuleris.

Along with this use of tuli or tetuli as the perfect of fero, we notice the use of tollo for fero in: Atlas . . . caeli qui sidera tollit (Aen. 8. 141) and in: adibo contra et tollam gradum (Pl. Bacch. 535), with which compare: fert . . . incomitata gradus (Ov. Met. 7. 184) and: quo tulerit gressum (Lucr. 4. 681). We have fero for tollo et fero in: cum castra moveri ac signa ferri iussisset (B. G. 1. 39 fin.), signa ferri ac sequi iubet armatos (Liv. 10. 5. 1), with which compare: sublatis signis ad Caesarem se contulerunt (Vell. 2. 61. 2). So also in: tuum 247 nomen . . . cantantes sublime ferent ad sidera cycni (Buc. 9. 29) and: Atia . . . somniavit intestina sua ferri ad sidera (Suet. Aug. 94. 4); with which compare: clamores simul horrendos ad sidera tollit (Aen. 2. 222), tollemus in astra nepotes (3. 158), fatis ad sidera tolli (12. 795). In ferre pedem and ferre gressum or gradus, ferre seems short for tollere et ferre ‘to raise and bear onward’, a union at times expressed by tollere, but usually by ferre. It seems that from this early union, no longer found in use, proceeded the union of fero and tuli; and that parallel unions like sum . . . fui in Latin and go . . . went in English, had a similar origin.

In: tollo gradus, tollo means ‘I raise from the ground’; Walde connects it with tolleno a sweep or swing-beam. But when we have it in such unions as tolli ad sidera or ad astra, its meaning is intensified, and it is so common in this use that its perfect is sustuli ‘I have raised on high’. It is now parallel, not with fero, but with effero; and we may compare: ad caelum mehercle tollimus verissimis laudibus (Cic. Fam. 15. 9. 1) with: te summis laudibus ad caelum extulerunt (ib. 9. 14. 1). At times laudibus is omitted, as in: Daphnimque tuum tollemus ad astra, Daphnim ad astra feremus (Buc. 5. 51-2), sua in destruendo eo consilia extulit (Ann. 2. 63. 4), quod valet non solum ad augendum aliquid et tollendum altius dicendo (Cic. de Orat. 3. 104. 26). We may further compare tollo in: ni sapiens sic Nomentanus amicum tolleret (Sat. 2. 8. 61), Sol in currum cum Phaethontem filium sustulit (Cic. N. D. 3. 76. 31), hac victoria sublatus Ambiorix (B. G. 5. 38. 1), ultro animos tollit dictis (Aen. 9. 127), tollitur in caelum clamor (11. 745) with effero in: clamorem utrimque ecferunt (Pl. Amph. 228), hic me magnifice ecfero (Ter. Heaut. 709), quos recenti victoria efferri sciret (B. G. 5. 47. 4), quorum animi altius se extulerunt (Cic. Rep. 3. 4. 3).

But effero also means ‘to bear out for burial’, as in: tum tu idem optumumst loces ecferendum: nam iam, credo, mortuost (Pl. Aul. 568); and to: ecfertur (Ter. And. 117) Donatus’s note is: efferri proprie dicuntur cadavera mortuorum. But Virgil writes: haec ubi deflevit tolli miserabile corpus imperat (Aen. 11. 59), pointing to an old union of tollo and effero in this sense also.

But in: avectaque partim finitimos tollunt in agros urbique remittunt (11. 206) tollunt is rather for transferunt; and we have the union tollo et transfero in: tollitur ab atriis Liciniis . . . et trans Alpes usque transfertur (Cic. Quinct. 12. 3). We have tollo for tollo et transfero in: da dextram misero et tecum me tolle per undas (Aen. 6. 370), fotum 248 gremio dea tollit in altos Idaliae lucos (1. 692), quem tollere in altos optabam primum montes (2. 635). Nonius (I. p. 669, L. M.) takes tulere in: unius in miseri exitium conversa tulere (Aen. 2. 131) as the perfect of transferre; it may be the poetic shortening for transtulere, but it seems rather to bear the sense of aequo animo tulere. We have also extollere used for transferre in: res serias omnes extollo ex hoc die in alium diem (Pl. Poen. 500), fugiam hercle aliquo atque hoc in diem extollam malum (Mil. 861), abi intro atque istaec adfer tamen; hodie extollat nuptias (Caecil. apud Non. 1. 470, L. M.); with which compare: se se in annum proximum transtulit (Cic. Mil. 24. 9), causa haec integra in proximum annum transferetur (Cic. Fam. 8. 9. 2). We find extollo used for tollo in: si nostram causam laudando extollemus (Auct ad Her. 1. 5 fin.), and so can understand this substitution of extollo for tollo in the union tollo et transfero.

In a fragment of Turpilius we read: ubi praeter se neminem vidit esse, tollit aufert (195, R.). We have tollo for tollo et aufero in: at tu quantum vis tolle (Ep. 1. 7. 16), partim vel tolleret omnes (ib. 1. 6. 44), erat quod tollere velles (Sat. 1. 4. 11), dapes iubet et sublata reponi pocula (Aen. 8. 175), tollite cuncta, inquit, coeptosque auferte labores (8. 439). So in: sive est virtus et gloria, tollat (11, 444), where Servius renders tollat by consequatur, Mackail by ‘win’, tollat seems for tollat et secum auferat. Striking is the contrast between: tollentem minas (Geo. 3. 421) ‘raising on high his threats’, where minas seems for dentem, and tolle minas (Aen. 10. 451) ‘away with threats’, tolle querelas (Ep. 1. 12. 3), sublatis dolis (Aen. 12. 26), where tolle seems a metonymy for aufer.

We seem to have an extension of this use in the union tollere et eripere, as in: tolle fuga Turnum atque instantibus eripe fatis (Aen. 10. 624), nonne videntur hunc hominem ex rerum natura sustulisse et eripuisse? (Cic. Rosc. Am. 71. 26), with which we may compare rapere et ferre in: alii rapiunt incensa feruntque Pergama (Aen. 2. 374). Tollere seems to stand for this union in: sed stirpem Teucri nullo discrimine sacrum sustulerant (12. 770), avaritiam si tollere voltis, mater eius est tollenda, luxuries (Cic. de Orat. 2. 171. 40), ut exercitum religio tollat, te auctorem senatus retineat (Fam. 1. 1. 3). We have the parallel union tollere et delere in: cuius omnino rei memoriam omnem tolli funditus et deleri arbitror oportere (Quinct. 70. 21), deleatis ex animo suo suspicionem omnem metumque tollatis (Rosc. Am. 6. 2). We have tollo for this union in: id nomen ex omnibus libris tollatur (Att. 13. 44. 3), solem enim e mundo tollere videntur, qui amicitiam e 249 vita tollunt (Lael. 47. 13), belli commercia Turnus sustulit ista prior iam tum Pallante perempto (Aen. 10. 533), cur non incolumi potius certamina tollo? (12. 39), etiam illuc pervenerint proverbium ut tollant anticum (Varro R. R. 2. 9. 9), probably the proverb: canis caninam non est (L. L. 7. 87). Extension from this is easy to: tollo occidendo; and tollo stands for this in: me truncus illapsus cerebro sustulerat (Od. 2. 17. 28), quia Drusum ferro, Metellum veneno sustulerat (Cic. N. D. 3. 81. 33). It was this use that Cicero really intended in: Caesarem laudandum et tollendum censebat, cum aliud diceret, aliud intellegi vellet (Vell. 2. 62 fin.).

We have the interesting phrase liberos tollere usual through all Latinity. We read in the first edition of Plautus: si quod peperissem, id educarem tollerem, bona sua med habiturum omnia (Truc. 399), a reading that seems best supported by manuscript authority, though Schoell wonders that critics have endured it so long, and which gives us the union tollere educare in poetic inversion, that seems the clue to this meaning of tollere. We have similar unions in: me genitor . . . Argolicum terrorem inter Troiaeque labores sublatum erudiit (Aen. 9. 203), si cui . . . validus male filius in re praeclara sublatus aletur (Sat. 2. 5. 46). For unions like these we find tollo used: of the parents in: quicquid peperisset, decreverunt tollere (Ter. And. 219), of the father in: is puerum tollit (Pl. Men. 33), uxorem duxi, natum sustuli (Quint. 4. 2. 42), of the father’s instructions in: verum quod erit natum tollito (Pl. Amph. 501), and of the mother in Pl. Truc. 399. We have tollere used by metonymy for educare in: tu illos duo olim pro re tollebas tua (Ter. Ad. 809). In: qui ex Fadia sustulerit liberos (Cic. Phil. 13. 23. 10), sublato filio Nerone ex Agrippina (Suet. Ner. 5 fin.) it seems transferred from tollo et educo to gigno, and in: quem serva Licymnia furtim sustulerat (Aen. 9. 547) to pario et tollo, where tollo is short for tollo et educo. In: tollite me, Teucri, quascumque abducite terras (Aen. 3. 601) it is for tollo et aufero.

For the union rapio feroque noticed in Aen. 2. 374, we have fero in: frustra retinacula tendens fertur equis auriga (Geo. 1. 514), omnia fert aetas, animum quoque (Buc. 9. 51), postquam te fata tulerunt (5. 34). We read: postquam res sociorum . . . ferri agique vidit (Liv. 22. 3. 7), ferri agique res suas viderunt (38. 15. 10); we find fero used for this union in: non feret quin vapulet (Pl. Amph. 308), and ago in: edepol ne illic pulcram praedam agat (Aul. 610). We have the union ruo ago in: ceteros ruerem agerem (Ter. Ad. 319); ago used for this in: Demoleos cursu palantes Troas agebat (Aen. 5. 265), and ruo in: 250 et ruit atram ad caelum picea crassus caligine nubem (Geo. 2. 308). We read: ceteris prae se fert et ostentat (Cic. Att. 2. 23. 3), but: laetitiam autem apertissime tulimus omnes (ib. 14. 13. 2), non liberalium modo disciplinarum prae se scientiam tulit (Quint. 12. 11. 21). We read: ita sui periculi rationes ferre ac postulare (Cic. Verr. 2. 5. 105. 40), but: dum licitum est ei dumque aetas tulit (Ter. And. 443), et ut aetas illa fert, sibi tamen non pepercisset (Cic. Cluent. 168. 60); and we find the union distributed in: ne aut gravioribus utar verbis quam natura fert, aut levioribus quam causa postulat (Cic. Quinct. 57. 18). We may conjecture that tulit is short for tulit peperitque in: aetas parentum peior avis tulit nos nequiores (Od. 3. 6. 46), Curium . . . utilem bello tulit et Camillum saeva paupertas (1. 12. 42), nec te conceptam saeva leaena tulit (Tib. 3. 4. 90).

But we come to a different kind of union. We read: inimici famam non ita ut natast ferunt (Pl. Pers. 351), eadem hoc quoque fama ferebat (Ov. Met. 12. 200), and with sermonibus for fama: haec omnibus ferebat sermonibus (Varro) (B. C. 2. 17. 2). But we find this shortened to fero in: quod fers cedo (Ter. Phorm. 857), quando et priores hinc Lamias ferunt denominatos (Od. 3. 17. 2), Ceres fertur fruges . . . mortalibus instituisse (Lucr. 5. 14), quem procul Aspis conspiciens ad se ferentem, pertimescit (Nep. Dat. 4. 5), where with ferentem we must supply aditum or cursum. But a most usual formula here is legem fero, which we have in: cum legem agrariam ferret (Cic. Off. 2. 73. 21), with which ad populum is to be supplied, as we see from: nihil . . . ad populum, nihil ad plebem latum esse dico (Balb. 33. 14). The law proposed, but not yet carried is styled rogatio; and so we read: dixit Sullam illam rogationem de se nolle ferri (Sull. 65. 23). But we have: quod Sulla ipse ita tulit de civitate (Caecin. 102. 35), nihil de iudicio ferebat (Sull. 63. 22), lato, ut solet ad populum, ut equum escendere liceret (Liv. 23. 14. 2). So of the juror giving his verdict sententiam ferre is used in: de quo vos sententiam per tabellam feretis? (Cic. Verr. 2. 4. 104. 47), but ferre of the Emperor’s verdict in: paenitentia patiens tulit absolvi reum criminibus maiestatis (Ann. 1. 74. 7). How much this shortening may affect the meaning of the verb, is best seen in a review of the history of puto ‘I think’. Gellius (7. 5) discussing the relation of putus to purus, in opposition to Varro is inclined to derive putus from puto ‘I prune’, and not puto from putus; he thinks that argentum purum putumque in the Carthaginian treaty is silver pure and unalloyed, emaculatam et aliena materia carens (sec. 9). 251 We have seen how from such unions as purus putus we get one word bearing at times the meaning of both, at times of the other term; so that it is easy to see how putus came to mean the same as purus, as Festus tells us it did (p. 217, M.). We do best to begin with the meaning puto shows in its compound amputo, and in: vineas arboresque mature face incipias putare (Cato R. R. 32. 1) or: vitem . . . fingit putando (Geo. 2. 407). In this sense we have putare joined with lavare in: ne lana inquinetur, quominus vel infici recte possit, vel lavari ac putari (Varro R. R. 2. 2. 18); from which union we have the resultant putari with the meaning of both in: qui non reddet temperi (lanam) putatam recte (Titin. 23, R.).

But its use in the phrase: rationem puto ‘I clear up the reckoning’ becomes the most usual of this class; we have this clearing up with the aid of the other party or parties to the account in: putatur ratio cum argentario (Pl. Aul. 527), rationem cum domino (vilicus) crebro putet (Cato R. R. 5. 3), ut rationes cum publicanis putarent (Cic. Att. 4. 11. 1), or with the subject himself, as in: cum eam mecum rationem puto (Pl. Cas. 555). But the items in the account are substituted for the account itself in: dum haec puto, praeterii imprudens villam (Ter. Eun. 632), conliciares (tegulae) quae erunt pro binis putabuntur (Cato R. R. 14. 4), si denique hoc semper ita putatum est (Cic. Div. 1. 84. 39), multaque dura suo tristi cum corde putabant (Aen. 8. 522). Then the items are omitted, as in: mecum argumentis puta (Pl. Amph. 592), recte putas (Ter. And. 141), quis coegit eos falsum putare? (Cic. Sen. 4. 2), where puto is equivalent to aestimo, I value, or consider, or think.

We read in Varro: disputatio et computatio cum praepositione a putando, quod valet purum facere. . . . Sic is sermo in quo pure disponuntur verba, ne sit confusus atque ut diluceat dicitur disputare (L. L. 6. 63). So we find disputo in: ubi disputatast ratio cum argentario (Pl. Aul. 529) with the same meaning as putatur in v. 527. Nonius (I. 602, M.) quoting from Ennius’s Thyestes: ibi quid agat secum cogitat, curat, putat,—and from Caecilius’s Exsul: non haec putas, non haec in corde versantur tibi, tells us that puto is here for disputo. So too in: in meo corde . . . eam rem volutavi et diu disputavi (Pl. Most. 88), quae sunt a me in secundo libra de oratore disputata de ridiculis (Cic. Fam. 7, 32. 2), quibus ex rebus breviter disputatis intellegi potest (Off. 1. 161. 45) the primary force of disputare is evident. Only in such connexions as: disputandumque de omni re in contrarias partes (De Orat. 1. 158. 34), opponuntur ab his qui 252 contra disputant (Rep. 1. 4. 3), paucis cum esset in utramque partem verbis disputatum (B. C. 1. 86. 3) does disputare take the meaning of our ‘dispute’, and the omission of contra or its equivalent is rare even in mediaeval Latin.

We find computo rationem also with the same meaning as puto rationem in: dextera digitis rationem computat (Pl. Mil. 204); and it has the same meaning with rationem omitted in: compellarat (Deiotarum) hospitem praesens, computarat, pecuniam imperarat (Cic. Phil. 2. 94. 37), omnes opertis oculis bona sua computant (Petron. 44), nec dierum numerum ut nos, sed noctium computant (Tac. Germ. 11. 2), (Nestor) suos iam dextra computat annos (Juv. 10. 249). But in: si computes annos, exiguum tempus, si vices rerum, aevum putes (Plin. Ep. 4. 24. 5) we feel how far puto in its later meaning has departed from the older sense still retained in computo. Reputo, deputo, and imputo show no example of the earlier use with rationem.

We read: ut ratio redditur (Pl. Men. 206) and: reddunda in ratione (Lucr. 1. 59), but: possint tamen omnia reddi (id. 1. 566) for: possit omnium ratio reddi, and in: utrumque quid a vero iam distet habebis (id. 1. 758) habebis is evidently for: notum habebis. In: simul ultima signant (Aen. 5. 317) signant seems for: signant oculis; Servius’s note is: deest visu, ut Cicero: notat et designat oculis (Cat. 1. 2. 1). He cites the same phrase in his note to: sidera cuncta notat tacito labentia caelo (Aen. 3. 515), but we do better here to compare: haec ab hominibus callidis et peritis animadversa ac notata (Cic. de Orat. 1. 109. 23), and to make notat for: animadvertit et notat. And in: ora sono discordia signant (Aen. 2. 423), where Servius’s note is: signant designant, we may take signant as short for notant et designant. We may compare: tacitus vestigia lustrat (11. 763) and: quae sit me circum copia lustro (2. 564) with: vestigia . . . lumine lustro (2. 754) and: totum lustrabat lumine corpus (8. 153). We may further compare: quo tutior hospita lustres aequora (3. 377), mixtis lustrabo Maenala nymphis (Buc. 10. 55), arvaque . . . lustrabat (Geo. 4. 519), vacua atria lustrat (Aen. 2. 528) with: lustrandum navibus aequor (3. 385), te lustrare choro (7. 391), lustrat equo muros (9. 58).

Many unions of this kind are so usual that the verb readily suggests the omitted object to the reader. Such we find when we compare: Perseus bellum iam vivo patre cogitatum in animo volvens (Liv. 42. 5. 1), or: multa cum animo suo volvebat (Sall. Jug. 6. 2), or: Fauni volvit sub pectore sortem (Aen. 7. 254) with: multa ipse secum volvens (Sall. Cat. 32. 1), per noctem plurima volvens (Aen. 1. 305), futura 253 volvens non aliud repperit (Ann. 1. 64. 7). So too in comparing: quaeso animum advorte (Pl. Pseud. 277) with: paucis, adverte, docebo (Aen. 8. 50), octo aquilae . . . imperatorem advertere (Ann. 2. 17. 2); in comparing: quo tenderent cursum (Liv. 23. 34. 5) or: cursuque amens ad limina tendit (Aen. 2. 321) with: tendit gramineum in campum (5. 286), tendit quotiens in altos nubium tractus (Od. 4. 2. 26); or: transmittunt cursu campos (Aen. 4. 154) with: tramisit Lesbum (Ann. 2. 54. i); or: pauci tentoria ponunt (Ov. Fast. 3. 527) with: hic saevus tendebat Achilles (Aen. 2. 29), apud vexillum tendentes (Ann. 1. 17. 4).

So we may compare: ut cum uno aetatem degeret (Ter. Phorm. 417) with: potens sui laetusque deget (Od. 3. 29. 42); ut supremam falsa inter gaudia noctem egerimus (Aen. 6. 514) with: haud minus inquies Germanus spe . . . agebat (Ann. 1. 68. 1); incumbe in eam curam et cogitationem (Cic. Fam. 10. 3. 3) with: tum vero Teucri incumbunt (Aen. 4. 397); classem velis aptare iubebat (3. 472) with: classem aptent taciti (4. 289); colonias ab eis decemviris deduci iubet (Cic. Leg. Agr. 2. 73. 27) with: in quascumque velim pelago deducere terras (Aen. 2. 800); votisque deos venerabere seris (7. 597) with: nymphas venerabar agrestes (3. 34). At times of the pair it is the verb that is omitted, as in: quae tenuem aciem (instrui) pateretur (Ann. 1. 64. 7), quinquennio maturius quam per leges (liceret) quaesturam peteret (Ann. 3. 29. 1), vagum ac lascivientem per agros (palari) militem sineret (Ann. 2. 55. 4), statimque (sibi solverentur legata) flagitavit (Ann. 1. 37. 1).

Emo means ‘I buy’; but in its compounds adimo, demo, eximo, promo, sumo, it means ‘I take’; and its root em- or nem- (cf. Germ. nehmen) points to this as its primary meaning. We see the reason for the meaning ‘to buy’, when we compare: ego spem pretio non emo (Ter. Ad. 219) with: quanti eam emit (Pl. Epid. 51) or: emit hosce de praeda ambos de quaestoribus (Capt. 34). It is plain that a first step in the omission of the noun is the substitution for it of a word of like meaning, e.g. quanti for pretio, and that in such substitution the verb retains its secondary meaning got from its association with pretio. We read: militibus ac sagittariis in terram expositis (B. C. 3. 23. 2), expositis in terram multibus (Liv. 24. 40. 9); then: armatis in littora expositis (37. 28. 8), cum . . . mancipia in insulam . . . exponerent (Suet. Claud. 25. 2), where expono assumes the meaning of expono in terram, as it does fully in: socios de puppibus . . . exponit (Aen. 10. 288), achieving the unexpected with many of them by landing them in the water. In: quibus ex navibus cum 254 essent expositi milites (B. G. 4. 37. 1) the substitution of ex navibus for in terram has a like result for expono, giving the force of: expono in terram. Latin, like Sanskrit, frequently substitutes the locative for the accusative in such phrases; and we read: legiones expositae in terra (Vell. 2. 79. 4), expositus in littore (Suet. Caes. 4. 2). Virgil has this in: informi limo glaucaque exponit in ulva (Aen. 6. 416), and with its aid gives us a fine oxymoron in: viros mediis exponit in undis (10. 305) ‘lands the men in the midst of the waves’.

In many cases the meaning of the context indicates the word to be supplied, as in: accipiens (auribus) sonitum saxi de vertice pastor (Aen. 2. 308), gemitumque cadentum accipio (10. 675), nostram nunc accipe mentem (1. 676), with which compare: accipite ergo animis atque haec mea figite dicta (3. 250). So in: ingeminant (laudes) plausu Tyrii (1. 747) or: securim altior exsurgens oranti (ictu) . . . congeminat (11. 698). So we may compare: in nullius umquam suorum necem duravit (Ann. 1. 6. 3) with: durat mentem senatumque rursum ingreditur (Ann. 3. 15. 4). In: sed iura fidemque supplicis (violare) erubuit (Aen. 2. 542), proceres plebemque iuxta (sibi) devinxerat (Ann. 2. 56. 2), nam suam aetatem vergere (ad occasum) (Ann. 2. 43. 1), eo promptior Caesar pergit introrsus (procedere) (Ann. 2. 25. 4) the ellipses indicated seem clearly required by the context. So in: Percennius et Vibulenus stipendia multibus (etiam militantibus), agros emeritis largientur (Ann. 1. 28. 6), where the relation of the noun to the verb omitted makes the ellipsis easier; and so in: noxque una Hannibali sine equitibus atque impedimentis (agenda erat et) acta est (Liv. 21. 34. 9). Very striking in: mandata Clementi centurioni quae perferret (Ann. 1. 26. 1) is the use of mandata for: mandata sunt mandata; as is in: motum ex Metello consule civicum (Od. 2. 1. 1) the use of motum civicum for: motum civicum motum.

Very easy too seems the ellipsis in: armatumque auro circumspicit (et videt) Oriona (Aen. 3. 517), where the first verb we meet in the following verse is: videt. Easy to supply are the ellipses in: in medium quaesita (et parta) reponunt (Geo. 4. 157), quocumque lectum (et pressum) nomine Massicum (Od. 3. 21. 5), sume superbiam quaesitam (et partam) meritis (3. 30. 15), tuos labores impune carpere (et edere) lividas obliviones (4. 9. 34), stantia non (sed ruentia) poterant tecta probare deos (Mart. 1. 12. 12), scrinia da magnis, me manus una capit (et habet) (1. 2. 4), teritur (et legitur) noster ubique liber (8. 3. 4), vix implet cocleam peracta messis (frugibus ablatis) (11. 18. 23). 255 Not very difficult are those in: quae manent (et exspectant) culpas etiam sub Orco (Od. 3. 11. 29), ego illis mollior nec te feriam (et occidam) (ib. 43), quos . . . iam flammae tulerint inimicus et (percusserit et) hauserit ensis (Aen. 2. 600). In: ergo instauramus Polydoro funus (3. 62) behind instauramus, which is here no longer ‘to establish’, but ‘to renew’, lies the pair: renovavit et instauravit (Cic. Verr. 1. 11. 4). Behind the adserto in: annum . . . adserto qui sacer orbe fuit (Mart. 7. 63. 10) lies the union: vindicare et adserere (Manil. 2. 815). More difficult is the syntax of euhoe in: qui tum alacres . . . furebant euhoe bacchantes, euhoe capita inflectentes (Catull. 64. 255), where probably furebant is short for furentes clamabant, and euhoe is the cognate object of clamabant thus implied. We may compare euantes orgia (Aen. 6. 517) which is for: celebrantes orgia et euhoe clamantes.

For it is not uncommon to find a participle coalescing with its verb to form a new verb bearing the meaning of both. So we have: aequora tuta silent (1. 164) for: silentia iacent; ture calent arae (1. 417) for: fragrantes ardent; iam multi crudele canebant artificis scelus (2. 124) for: canentes praedicebant; furit aestus ad auras (2. 759) for: fertur furens; illa (arbor) usque minatur (2. 628) for: minans movetur; praedam pedibus circumvolat uncis (3. 233) for: circumvolans rapit; haec ubi deflevit (11. 59) for: deflens locutus est; pecuniam heredi properet (Od. 3. 24. 62) for: properans cogit; stupet Albius aere (Sat. 1. 4. 28) for: stupens tuetur; milia tum pransi tria repimus (1. 5. 25) for: repentes conscendimus.

We have not participles, but adjectives omitted in: fervet opus (Aen. 1. 436) for: fervidum agitur opus; flammasque ministrant (1. 213) for: ministri praebent; and gerundives by a like easy transition in: nunc etiam superare necessest corpora rebus (Lucr. 1. 579) for superando superesse; causas penitus tentare latentes (Aen. 3. 32) for: tentando exquirere; excudent alii . . . aera (6. 847) for: excudendo efficient; siccis oscula falle genis (Prop. 4. 11. 80) for: da fallendo; inludo cartis (Sat. 1. 4. 139) for: inludendo do versus cartis.

Thence we pass to adverbs or equivalent ablatives, as in: rapuitque in fomite flammam (Aen. 1. 176) for: raptim abstulit; campum eripi iubet (Ann. 1. 63. 1) for: raptim occupari; sumpsisse merentes laudabor poenas (Aen. 2. 586) for: cum laude dicar; aedes ululant (2. 488) for: ululatu resonant; tempora navali fulgent rostrata corona (8. 684) for: cincta rostris navalis coronae; multa cum libertate notabant (Sat. 1. 4. 5) for: nota adficiebant. In: maturate fugam (Aen. 1. 137) for: mature 256 fugite, and: celerare fugam (1. 357) for: celeriter fugere, we seem to have the inverted form that leads to this shortening, which probably involves the same figure that we shall presently examine in the expression of four terms by two.

We pass on to the expression of the verb and its participial object by a single verb. We seem to have the inverted form leading to this in: inventaque flumina monstrat (Aen. 6. 8) for: invenit et monstrat, servetis revocatum a morte Dareta (5. 476), exceptum comiter iuvenem sueta . . . liberalitate auget (Ann. 3. 8. 2), multos Angrivarii . . . redemptos ab interioribus reddidere (Ann. 2. 24. 5), where the union of two verbs is converted into that of a verb with its object. We have a single verb used for this union in: terram inter fluctus aperit (Aen. 1. 107) for: apertam ostendit; tres (naves) in saxa latentia torquet (108) for: tortas inmittit; aperit ramum (6. 406) for: apertum ostendit; noctes quas de me fatiges (Prop. 4. 11. 81) for: fatigatas degas; omnes composui (Sat. 1. 9. 28) for: conlectos posui.

We must supply substantives instead of participles in: omnium egenos urbe domo socias (Aen. 1. 600) for: socios accipis; faciem illius falle dolo (1. 684) for: sume falso dolo; et casum insontis mecum indignabar amici (2. 93) for: indignum volvebam animo; fugam . . . moliri (2. 109) for: molestam perficere; rite secundarent visus (3. 36) for: secundos darent; haud dubitanda (3. 170) for: in dubium vocanda; ipsi transtra novant (5. 752) for: nova faciunt; hastilia densat (11. 650) for: densa iacit; ne facundiam violentia praecipitaret (Ann. 3. 19. 1) for: praecipitem daret; qui scis eos nunc discordare inter se? (Ter. And. 575) for: se discordes praebere.

We supply gerundives in: artes quas doceat quivis eques atque senator (Sat. 1. 6. 77) for: docendas curet; corpora frangeret ad saxum (Aen. 3. 625) for: pelleret frangenda; et silvis aptare trabes et stringere remos (1. 552) for: aptandas . . . et stringendos detrahere; and infinitives in: nec saxa nec ullum telorum interea (inmitti) cessat genus (Aen. 2. 468), probare deos (esse) (Mart. 1. 12. 12), aurum Fabricius te tribuente volet (accipere) (11. 5. 8), se ortum . . . (probare) volebat (Aen. 1. 626). Asconius tells us at the end of his preface to the Verrine orations, that it was usual for the orators of old to end their speeches with the word dixi (cf. Ter. Phorm. 437 and 439); and so: dixerat (Aen. 2. 621) is probably short for: ‘dixi’ dixit, or: ‘dixi’ dicebat.



Thus far we have been busy with what Donatus (ad Ter. And. 230) calls the figure δύο δι’ ἑνός; we turn now to its opposite, the figure ἓν διὰ δυσίν, or, as we write it, hendiadys. Servius (ad Aen. 1. 61) thus defines it: est figura ut una res in duas dividatur, metri causa interposita coniunctione ut alio loco; pateris libamus et auro (Geo. 2. 192), id est pateris aureis. The phrase he is explaining is: molem et montes, which he takes for molem montium, and he evidently agrees with our later grammarians, who confine the figure to the forms of it found in: pateris et auro for pateris aureis, and molem et montes for molem montium; indeed he goes so far as to think of the et as having no meaning and being inserted merely for ihe metre. The manuscripts here make him name the figure ‘endiadis’; but Thilo has not admitted this term into his text of Servius, feeling that it proceeds from a later reviser. And Servius in his note to: sternere nec iacta caecum dare cuspide volnus (Aen. 10. 733), which is: unum sensum per duos extulit, though he does not use the term endiadis, seems to extend the figure to verbs as well as nouns. A rational account of the nature and development of hendiadys, as the opposite of δύο δι’ ἑνός, which it seems to be, will extend the figure to verbs as well as nouns, and will find in the varieties of it given by the grammarians merely its crowning developments.

The most obvious use of two for one in the poets is in repetitions for poetic emphasis; such as: et magis magis in dies et horas (Catull. 38. 3), Rhaebe, diu, res si qua diu mortalibus ulla est, viximus (Aen. 10. 861), dulce ridentem Lalagen amabo, dulce loquentem (Od. 1. 22. 23), inde domum, si forte pedem, si forte tulisset, me refero (Aen. 2. 756), eheu fugaces, Postume, Postume, labuntur anni (Od. 2. 14. 1), ibimus, ibimus, utcumque praecedes (2. 17. 10), heu fuge crudeles terras, fuge litus avarum (Aen. 3. 44), iam parce sepulto, parce pias scelerare manus (3. 41-2). There is no syntactical irregularity here involved; but when we compare Horace’s: geminus Pollux with Catullus’s: gemelle Castor et gemelle Castoris, 258 we feel that it is emphasis that lies at the basis of this figure. The same seems true of the poet’s use of non sine for cum (Catull. 13. 4), of non secus for velut (Od. 2. 3. 2), of post deinde (Ter. And. 483) mox deinde (Tib. 1. 5. 73), verum hercle vero (Pl. Curc. 375). We approach the figure in: a prima . . . origine (Aen. 1. 753), littoris oram (3. 396), dictus sacer (6. 138) for consecratus, caeli sidera (8. 141) for caelum, spiramenta animae (9. 580) for pulmones, dives avis (10. 201) for generosa, pacis ramos (11. 332) for olivas, pedibus uncis (3. 233) for ungulis, nescios fari (Od. 4. 6. 18) for infantes, prominentes oras (Ann. 2. 24. 3) for promuntoria, biiugis equis (Mart. 1. 12. 8) for bigis.

We have already noticed the emphatic force of the neuter plural in phrases like: prima virorum (Lucr. 1. 86), summa ducum Atrides (Ov. Am. 1. 9. 37). We may further note: angusta viarum (Aen. 2. 332), opaca locorum (2. 725), summa navium (Tac. Hist. 3. 47. 4), scriptorum quaeque (Sat. 2. 3. 2), multos Danaum (Aen. 2. 398), multos illustrium Romanorum (Ann. 3. 6. 1), prominentia montium (Ann. 2. 16. 2), silvarum ac montium profunda (Agric. 25. 1). But we find a noun substituted for the adjective in: minae murorum (Aen. 4. 88) for: muri minantes, rotarum lapsus (2. 236) for: rotas labentes, blanditiae rosae (Prop. 4. 6. 72) for: rosae blandae, hostes Medorum (3. 9. 25) for: hostiles Medos, summa . . . opum vi (Aen. 12. 552) for: summis viribus, odora canum vis (4. 132) for: canes multi et sagaces.

Very old is the use of corpora we see in: delecta virum . . . corpora (Aen. 2. 18) for delecti viri, corpora natorum (6. 22), corpora . . . magnanimum heroum (6. 306), multa virum . . . corpora (10. 662); for we read in Ennius: ter quattuor corpora sancta avium (Ann. 90, M.). On the analogy of this we have: formae magnorum luporum (Aen. 7. 18), forma tricorporis umbrae (6. 289), regum colla minacium (Od. 2. 12. 12), variarum monstra ferarum (Aen. 6. 285), summa . . . fastigia rerum (1. 342), ignoti nova forma viri (3. 591), venerabile donum fatalis virgae (6. 409), impendentium montium altitudines inmensitatesque camporum (Cic. N. D. 2. 98. 39). Easy is the transition from multos illustrium virorum to multos et illustres viros; but when we try to convert formae luporum magnorum in like fashion we have at once a hendiadys in: formae et lupi magni. The same is the case when we turn monstra variarum ferarum.

Parallel to the two pairs of words connected by que in Cic. N. D. 2. 98 quoted above we have like pairs connected by et and designating a single object in: ingens argentum Dodonaeosque lebetas (Aen. 3. 466), 259 fortuna omnipotens et ineluctabile fatum (8. 334), arma artis opisque tuae (Aen. 8. 377), secutus Ledaeam Hermionem Lacedaemoniosque hymenaeos (3. 328), caestus ipsius et Herculis arma (5. 410), Danaum insidias suspectaque dona (2. 36), nec dulces natos Veneris nec praemia noris (4. 33), stridor ferri tractaeque catenae (6. 558), tumor omnis et irae concessere deum (8. 40), aureus arcus et arma Dianae (11. 652). You will notice that of the ten examples cited, examples of true hendiadys, the last five consist each of two pairs, one of which is a noun with its adjective, the other a noun with a dependent genitive, and that the remaining four were made up of unions of one or other of these forms. We showed how out of the noun with its adjective was evolved the noun with its dependent genitive; and how out of the second could easily be developed a common form of hendiadys.

But we have not a like balancing of two and two in the following: where the second term gives a fuller description of the object than the first: et pondus et ipsa vinclorum inmensa volumina (Aen. 5. 407), at specus et Caci detecta apparuit ingens regia (8. 241), pellem horrentisque leonis exuvias (9. 306), infelix avis et Cecropiae domus aeternum opprobrium (Od. 4. 12. 6). We have the opposite in: ut prima novercae monstra manu geminosque premens eliserit angues (Aen. 8. 289). In: voces vagitus et ingens infantumque animae flentes (6. 426) we have a hendiadys coupled with a union that is a combination of the adjectival and genitive unions that made up the unbalanced unions also that we have just quoted. We can feel throughout the development of this figure the influence of these unions on the figure we derived from the second of them.

Easier to understand and probably older are the unions that express two elements which unite to constitute the object represented. We have: farre pio et saliente mica (Od. 3. 23. 20) for the mola salsa, rore levi et ramo felicis olivae (Aen. 6. 230) for the lustratio, pacem aeternam pactosque Hymenaeos exercemus (4. 99), arbuteis texunt virgis et vimine querno (11. 65). Most common are pairs corresponding to our ‘purple and gold’, as in: mores sibi emit auro et purpura (Pl. Most. 286), auro ductores longe effulgent ostroque decori (Aen. 5. 132), ostroque insignis et auro (4. 134), purpurei cristis . . . auroque corusci (9. 163), vestes auroque ostroque rigentes (11. 72), ductores auro volitant ostroque superbi (12. 126), regali conspectus in auro nuper et ostro (A. P. 228).

That this form of hendiadys is the more primitive is confirmed by the number of simple pairs I have met with in it, as in: in saxis ac 260 speluncis (Lucr. 1. 348), thalami taedaeque (Aen. 4. 18), velatum auro vittisque iuvencum (5. 366), Paridis direxti tela manusque (6. 57), voltum lacrimis et ora rigabat (9. 251), licet arma mihi mortemque minetur (11. 348), castra Aeneas aciemque movebat (11. 446), lances donaque saepe dedit (Mart. Ep. Lib. 29. 6). In the following we have a single term followed by a pair: cinerem et sopitos suscitat ignes (Aen. 5. 743), gemitus iraeque leonum (7. 15) where irae seems a metonymy for fremitus, tergo stratisque . . . velleribus (7. 94), crates et molle feretrum (11. 64), currum rotasque volucres instabant (8. 433), Argos Agamemnoniasque Mycenas (6. 838) for totam Graeciam. More extended than the two balanced pairs, and leaning to the second are: amissis remis atque ordine debilis uno (5. 271), calidos latices et ahena undantia flammis (6. 218), angit inhaerens elisos oculos et siccum sanguine guttur (8. 261); while leaning to the first we have: armentalis equae mammis et lacte ferino (11. 571), conum insignis galeae cristasque comantes (3. 468). We have four balanced by three in: matrisque . . . tremenda Carmentis nymphae monita et deus auctor Apollo (8. 335-6). That the nexus through the genitive is a later development in such unions than the adjectival union, is indicated by the marked preponderance of the latter in this class of hendiadys.

Of the type represented by molem et montes I have noted the following examples consisting like it of a single pair: vocamus in partem praedamque Iovem (Aen. 3. 223), en dextra fidesque (4. 597), hic membris et mole valens (5. 431), quae forma viros fortunave mersit (6. 615), haud vinclo nec legibus aequam (7. 203), cenae sine aulaeis et ostro (Od. 3. 29. 15), spiritus et vita redit bonis (4. 8. 14), saecula posterique possint (Mart. 10. 20. 16). Of a single term followed by one doubled by an adjectival or genitive adjunct we have: floribus et dulci . . . complectitur umbra (Aen. 1. 694), telis et luce coruscus ahena (2. 470), ferro accisam crebrisque bipennibus (2. 627), arboribus clausa circum atque horrentibus umbris (3. 230), miratus . . . adventum sociasque rates (5. 36), fama . . . et clari nomen Acestae (5. 106), nodos et vincula linea rupit (5. 510), clausae tenebris et carcere caeco (6. 734), fronde super galeam et felici comptus oliva (7. 751), truncis et duro robore nata (8. 315), tolerare colo vitam tenuique Minerva (8. 409), socios inhumataque corpora terrae mandemus (11. 22), in pastus armentaque tendit equarum (11. 494), libro et silvestri subere clausam (11. 554), invadunt Martem clipeis atque aere sonoro (12. 712), ne metuas fastus limenque superbum (Mart. 1. 70. 13). Of a term thus doubled followed by a single term we have: barbarico postes 261 auro spoliisque superbi (Aen. 2. 504), ire ad conspectum cari genitoris et ora (6. 108), repertorem medicinae talis et artis (7. 772), saevo gelu duramus et undis (9. 604), peregrina ferrugine clarus et ostro (11. 772), ut Dirae stridorem agnovit et alas (12. 869), liberas fruges et Cererem ferunt (Od. 3. 24. 13). Of two terms each thus doubled we have: obscuros colles humilemque videmus Italiam (Aen. 3. 522), securos latices et longa oblivia potant (6. 715), litoreas agitabat aves turbamque sonantem (12. 248), impios Titanas inmanemque turbam sustulerit (Od. 3. 4. 42).

Of the type in: pateris et auro I have noted the following: pallam signis auroque rigentem (Aen. 1. 648), lacrimis et mente morata (4. 649), radiisque ardentem lucis et auro (7. 142), Aeneae sedem et secreta petebat (8. 463), odiis et crimine . . . infensus (11. 122), da . . . fortunam atque viam (10. 422), pugnae nodumque moramque (10. 428), primus . . . fidei et constantiae dies (Ann. 1. 58. 1), ira et dissimulatio (Ann. 2. 57. 4), auxilium adventumque dei (Aen. 8. 201), primi sub limina solis et ortus (6. 255), squamis serpentum auroque polibant (8. 436), duplici squama lorica fidelis et auro (9. 707), loricam consertam hamis auroque trilicem (3. 467). When we turn from the two elements constituting the object, as in: ostro insignis et auro to one set in this form, as in: signis auroque rigentem, or from: effulgens in auro nuper et ostro to: radiis ardens lucis et auro, we have at once this form of hendiadys springing from the representation of one element shaped after the analogy of the two. Probably then hendiadys is the extension by analogy to a single constituent element, of a syntax that is natural and proper to two constituent elements. Further extension by analogy has no doubt obscured it totally in most examples.

When we turn to verbs, we notice in poetry the use of many pairs for single verbs. Most common here is the union of dare with a noun or adjective, as in: dicta dedit (Aen. 2. 790), cursum dedit (10. 870), fugam dant (12. 367), dedit quietem (8. 30), partu dabit (1. 274), dat inermum (10. 425), placata dant (3. 70), vasta dabo (9. 323). So with dicere in: dicere carmen (Hor. C. S. 8), nomine dicunt (Aen. 6. 441), cognomine dixit (3. 335), and with others in: edidit ore (7. 194), fecere profanos (12. 779), tulit gressum (6. 677), rotam volvere (6. 748), loco statuit (12. 506), phrases which at once suggest the single verb for which they stand. Less obvious is: servitio premet (1. 285) for comprimet, manus dedisset (11. 568) for se dedidisset, vomere exercent (11. 318) for arant, vestigia figit (6. 159) for constitit, gelidus coit (3. 30) for congelat, aequa solo ponam (12. 569) for solo 262 aequabo. More difficult is: volvens arcana movebo (1. 262) for obscura aperiam. We have a fourfold union: finem dedit ore loquendi (6. 76) for dixerat.

The usual way of shortening two verbs to one in poetry is by giving one of them the form of a participle, as in: discite . . . moniti (Aen. 6. 620), turbata arripe castra (9. 13), patribus dat iura vocatis (5. 758), exceptum inmerserat (6. 174), pressoque obmutuit ore (6. 155), emissa manu contorsit spicula (11. 676). In the last example we have a poetical inversion of: contorta manu emisit spicula; we have a like inversion in the order of the verbs in: progressi subeunt luco fluviumque relinquunt (8. 125) for: fluvio relicto progressi luco subeunt. So in: vix pauca furenti subicio et raris turbatus vocibus hisco (3. 314) for: vix raris et turbatis vocibus hiscens furenti pauca subicio, and in: aulai medio libabant pocula Bacchi impositis auro dapibus paterasque tenebant (3. 354-5) for: pateras tenentes . . . libabant; in: sentiat et tandem experiatur Turnum in armis (7. 434) for: expertus Turnum in armis sentiat (quantum possit); in: di cuius iurare timent et fallere numen (6. 324) for: iuratum fallere numen; in: cui datus haerebam custos cursusque regebam (6. 350) for: haerens regebam; in: quod saepe malae legere novercae miscueruntque herbas (Geo. 3. 282) for: cum quo lecto miscuerunt herbas; in: quales Threiciae cum flumina Thermodontis pulsant et pictis bellantur Amazones armis (Aen. 11. 659-60) for: pulsant bellantes; and in: quasque Aniena sacras Tiburs per flumina sortes portarit sicco perlueritque sinu (Tib. 2. 5. 69-70) for: lautas (i.e. puras) perportarit. In: sequitur sic deinde Latinus suspiciens caelum, tenditque ad sidera dextram (Aen. 12. 195-6) the union sequitur—suspiciens—tenditque seems for sequitur—suspiciens—tendensque.

Besides these uses of two verbs for a verb and a participle, to which we might add: ad templum Palladis ibant . . . peplumque ferebant (1. 480), we have verbs for participles in adverbial relations, as in: numina magna vocat meritosque indicit honores (3. 264) for: numinibus vocatis indicit; in: litore ahena locant alii flammasque ministrant (1. 213) for: ahenis in litore localis flammas alii ministrant; in: bellum ingens geret Italia populosque feroces contundet (1. 263) for: bello ingente cum Italis gesto . . . contundet; in: hic ego tuas sortes . . . ponam lectosque sacrabo viros (6. 73) for: sortes ponam sacratis viris lectis. Verbs are also used for participles in the object relation in: accipite ergo animis atque haec mea figite dicta (3. 250) for: dicta accepta in animis figite; in: effigies Pisonis traxerant in Gemonias ac 263 divellebant (Ann. 3. 14. 6) for: effigies tractas in Gemonias divellebant. In: abdiderat sese atque aris invisa sedebat (Aen. 2. 574) for: abdita in aede atque aris invisa sedebat, invisa has a double force, and aris may be taken either as instrumental or dative. In: nec veterum memini laetorve malorum (11. 280) laetorve seems for laetus.

More difficult to convert is the hendiadys in: atque idem fugientem haud est dignatus Oroden sternere nec iacta caecum dare cuspide volnus (10. 732-3), where Servius tells us that by sternere and caecum volnus dare Virgil conveys a single idea. So with: verso tenuis cum cardine ventus impulit et teneras turbavit ianua frondes (3. 448-9), where we might change the syntax to: tenui impulsae vento per ianuam apertam frondes tenerae turbatae sunt; and in the previous example perhaps sternere hasta clam coniecta may give the prose construction. But in: spirat adhuc amor, vivuntque commissi calores Aeoliae fidibus puellae (Od. 4. 9. 10-12), while the two verbs express one idea, I have no thought of reducing them to one expression. I have ventured to give these few examples of hendiadys of the verb in the hope that I may call attention to this use of the figure, not because I think I am giving it adequate treatment.



In poetry an adjective belonging to each of a pair of substantives is often expressed only with the second, when to it there is subjoined the enclitic que which connects the substantives, and indicates that the adjective to which it is subjoined must be understood with the first as well. Troes trepidique Latini (Aen. 12. 730) is for: Troes trepidi trepidique Latini; the same is true of: Rutuli veteresque Sicani (7. 795), Mnestheus acerque Serestus (9. 171), Catillusque acerque Coras (7. 672), Chii veterisque Falerni (Sat. 2. 3. 115). So with common names: gemitu miseroque tumultu (Aen. 2. 486), ludo fatigatumque somno (Od. 3. 4. 11), per titulos memoresque fastos (4. 14. 4), tigres comitesque silvas (3. 11. 13), moribus hic meliorque fama (3. 1. 12), and with verbs in: fervet inmensusque ruit (4. 2. 7), metues doctusque cavebis (Sat. 2. 7. 68). Slightly more involved seems this figure in the following: at Messapus erit felixque Tolumnius (Aen. 11. 429), adiectis Britannis imperio gravibusque Persis (Od. 3. 5. 4), illa noto citius volucrique sagitta ad terram fugit (Aen. 5. 242), adsis o, placidusque iuves (4. 578), excutitur pronusque magister volvitur in caput (1. 115). In all of these the word to be supplied with noun or verb seems the same in form as that to which que is subjoined.

But often the word to be supplied is in a form changed to agree with the preceding noun, as in: moenia surgentemque arcem (Aen. 1. 366), pueri innuptaeque puellae (2. 238), insidias suspectaque dona (2. 36), clipeos mentitaque tela (2. 422), Spartam patriasque Mycenas (2. 577) crines incanaque menta (6. 809), fidem mutatosque deos (Od. 1. 5. 6), pro curia inversique mores (3. 5. 7), hortos egregiasque domos (Sat. 2. 3. 24). Slightly more involved are: Lycios fidumque vehebat Oronten (Aen. 1. 113), limen erat caecaeque fores (2. 453), scopulos avolsaque viscera montis (3. 575), craterasque simul pulcrosque tapetas (9. 358), iaculo celerem levibusque sagittis (9. 178), iaculo incedit melior levibusque sagittis (5. 68). Considerably more involved seem: regna Neoptolemi referam versosque penates Idomenei? 265 (11. 264), pila manu saevosque gerunt in bella dolones (7. 664), succedam castris Tyrrhenaque regna capessam (8. 507), inmisitque fugam Teucris atrumque timorem (9. 719), totque maris vastaeque exhausta pericula terrae (10. 57), disce, puer, virtutem ex me verumque laborem (12. 435), spargite me in fluctus vastoque inmergite ponto (3. 605), sanguine cernis adhuc sparsoque infecta cerebro (5. 413). You will tell me that I am heaping up examples of this simple figure beyond all need; my answer is that in all eight of the more involved examples just cited Mackail has failed to translate it. Still more involved is it in: frigus quo duramque famem propellere possit (Sat. 1. 2. 6), culpantur frustra calami, inmeritusque laborat iratis natus paries dis atque poetis (2. 3. 7), in both of which Lonsdale and Lee’s translation misses the figure; Bryce sees it in the first, but misses it in the second. It is not common in the best prose; I cite from Livy: quos (clamores) nemora etiam repercussaeque valles augebant (21. 33. 6).

We shall realize better the need of giving careful attention to this ellipsis when we examine the example of it in:

duri magno sed amore dolores

Pollute, notumque furens quid femina possit

Triste per augurium Teucrorum pectora ducunt (Aen. 5. 5-7).

Mackail translates: ‘But the bitter pain of a great love trampled, and the knowledge of what a woman can do in madness, draw the Teucrians’ hearts to gloomy guesses’. He misses the figure; as does Servius, who paraphrases: nam duri dolores magno animo polluto, id est laeso, et notus feminarum furor ducebat Troianos per triste augurium, scilicet ut crederent se interemisse Didonem. Forbiger’s note: notum substantive est accipiendum, is repeated by Conington, Ladewig, Papillon, Page, and Sidgwick; but Sidgwick deserves credit for perceiving that this does not give perfect sense. He adds: ‘The expression is tolerably clear, though not quite accurate; the grief is Dido’s grief, and it is the thought of this that makes the Trojans anxious’. It is unfortunate that he should blame this lack of clearness and accuracy on Virgil’s Latin, when it is his own lack of skill to translate Virgil’s Latin that is at fault; he has missed the force of que in notumque, that should lead him to supply noti with dolores in the preceding clause. Virgil’s expression is quite clear and accurate, and is much more adequate than what Sidgwick suggests; it is not the thought, but the knowledge of Dido’s grief that makes the Trojans anxious. We should no more think of translating notumque 266 here as if it were merely notum and que ‘and’ than we should translate undique as if it were unde and que ‘and’.

That this figure holds for mixtusque, though our English does not call for the repetition, is plain from: Teucri mixtique Sicani (Aen. 5. 293), Teucri mixtique Latini (11. 134), laetitia mixtoque metu (11. 807). But in: avolsaque saxis saxa vides mixtoque undantem pulvere fumum (2. 609) and: aestuat ingens uno in corde pudor mixtoque insania luctu (10. 871) it seems best to resolve: mixtoque insania luctu into: et insania mixta mixto luctu, and: mixtoque undantem, etc., into: et undantem fumum mixtum mixto pulvere.

The figure is extended to three in: Mnesthea Sergestumque vocat fortemque Serestum (4. 288), Talon Tanaimque neci fortemque Cethegum . . . mittit (12. 513). It is used with pronouns in: Demetri teque Tigelli (Sat. 1. 10. 90), sapiens vitatu quidque petitu sit melius causas reddet (1. 4. 115), nexantem nodis seque in sua membra plicantem (Aen. 5. 279) where it shows that nexantem cited by Priscian and Eutyches is correct, and not nixantem, the reading of many good manuscripts, and adopted by Ribbeck. We have it with adverbs in: licuit semperque licebit (A. P. 58), sedet aeternumque sedebit (Aen. 6. 617), forte sacer Cybelae Chloreus olimque sacerdos (11. 768), fama dediti benigneque excepti Segestis (Ann. 1. 59. 1), hic aliud maius miseris multoque tremendum obicitur magis (Aen. 2. 199). But: defleo equidem filium meum semperque deflebo (Ann. 3. 12. 8) is not an example owing to the presence of equidem in the first clause balancing semper. We have it with a numeral in: Troiam nec fata vetabant . . . stare decemque alios Priamum superesse per annos (Aen. 8. 399), and with nouns alone or with appositives, adjectives, or genitives, as in: sed variat faciemque novat (Ov. Met. 15. 255), nox fabulaeque manes (Od. 1. 4. 16), suspirans imoque trahens a pectore vocem (Aen. 1. 371), tristitiam vitaeque labores (Od. 1. 7. 18).

We have it with the verb in: sic potenti iustitiae placitumque Parcis (Od. 2. 17. 16), recedentis trilingui ore pedes tetigitque crura (2. 19. 32), insanum te omnes pueri clamentque puellae (Sat. 2. 3. 130), fractas utinam tua tela sagittas si licet extinctas aspiciamque faces (Tib. 2. 6. 16). We have it twice in: at hic si plaustra ducenta concurrantque foro tria funera, magna sonabit cornua quod vincatque tubas (Sat. 1. 6. 42-4). The figure is extended from two to three in: quid refert morbo an furtis pereamque rapinis? (Sat. 2. 3. 157), in cicere atque faba bona tu perdasque lupinis (2. 3. 182), and in: fervidus tecum puer et solutis Gratiae zonis properentque Nymphis (Od. 1. 267 30. 6). The examples I have noted, where the word to be understood is either formally or substantially the same as that expressed, give a fair idea of its ordinary range.

It is easy to see that ve, so often confused with que, often takes its place in this figure. We have: aut super Pindo gelidove in Haemo (Od. 1. 12. 6), ne noster honos infractave cedat fama loco (Aen. 7. 332), non Seres infidive Persae (Od. 4. 15. 23), lateris (miseri) miseri capitisve (Sat. 2. 3. 29), non me Lucrina iuverint conchylia magisve rhombus aut scari (Epod. 2. 50), illa tamen se non habitu mutatve loco (Sat. 2. 7. 64), uter aedilis fueritve vestrum praetor (2. 3. 180), quis udo deproperare apio coronas curatve myrto? (Od. 2. 7. 25), non Pyladen ferro violare aususve sororem Electram (Sat. 2. 3. 139), qui . . . si illud idem in rapidum flumen iaceretve cloacam (2. 3. 242). It is natural to expect that vel will be used like ve; and we read: partem vel tolleret omnes (Ep. 1. 6. 43), quem virum aut heroa lyra (dulci) vel acri tibia? (Od. 1. 12. 1), vel Baccho Thebas vel Apolline Delphos insignes (1. 7. 3).

Que is the oldest Latin copulative, and as it is the most usual in poetry, it is natural to find it the conjunction most common in this poetic figure. But it would be strange if et did not follow its analogy here. We find et subjoined like que to the word to be supplied in the first phrase in: (quos) doctos ego quos et amicos prudens praetereo (Sat. 1. 10. 87), (vivunt) campestres melius Scythae . . . vivunt et rigidi Getae (Od. 3. 24. 9-11). But et is usually placed just before the word to be thus supplied. We have it with adjectives in: mittimur Elysium et pauci laeta arva tenemus (Aen. 6. 744), ille te mecum locus et beatae postulant arces (Od. 2. 6. 21), Typhoeus et validus Mimas (3. 4. 53), divitum mensis et amica templis (3. 11. 6), laudem et optatum . . . decus (4. 14. 39), Brontesque Steropesque et nudus membra Pyracmon (Aen. 8. 425), messes et bona vina date (Tib. 1. 1. 24), ut opus et alii proelium inciperent (Ann. 1. 63. 7), and with an adverb in: eam rem volutavi et diu disputavi (Pl. Most. 87). We have it with a noun in: otium et oppidi laudat rura sui (Od. 1. 1. 16), and with verbs in: et tunicae manicas et habent redimicula mitras (Aen. 9. 616), horrendum et dictu mirabile (3. 26), purus et insons . . . si et vivo carus amicis (Sat. 1. 6. 70), at bene si quis et vivat puris manibus (1. 4. 68), audire et videor pios errare per lucos (Od. 3. 4. 7), Plotius et Varius, Maecenas Vergiliusque, Valgius et probet haec Octavius (Sat. 1. 10. 82). But we have examples of et not placed either immediately before or immediately after the word to be supplied, as 268 in: iactes et genus et nomen inutile (Od. 1. 14. 13), et corde et genibus tremit (1. 23. 8), Delius et Patareus Apollo (Od. 3. 4. 64); but when we change et corde et tremit genibus to et corde et genibus tremit, we seem to pass from poetry to prose.

Of other conjunctions I have noted the following examples: sive in: sive deae seu sint dirae (Aen. 3. 262), sive flamma sive mari libet Hadriano (Od. 1. 16. 4), tollere seu ponere volt freta (1. 3. 16), vacui sive quid urimur (1. 6. 19), ficta seu vera promeret (Ann. 1. 6. 6), Matutine pater seu Iane libentius audis (Sat. 2. 6. 20), turdus sive aliud privum dabitur tibi (2. 5. 11). We have aut in: et peccare nefas aut pretium est mori (Od. 3. 24. 24), quae nemora aut quos agor in specus (3. 25. 2); and atque in: iam satis terris nivis atque dirae grandinis misit Pater (1. 2. 1), parce, frugaliter atque viverem uti contentus eo quod mi ipse parasset (Sat. 1. 4. 107). Other examples are: nec cupressi nec veteres agitantur orni (Od. 1. 9. 12), furorne caecus an rapit vis acrior (Epod. 7. 13), incertus scamnum faceretne Priapum (Sat. 1. 8. 2), quae (ludit) velut latis equa trima campis ludit exsultim (Od. 3. 11. 9). We have seen that cum was a conjunction before it was a preposition; and we have: cur pendet tacita fistula cum lyra? (Od. 3. 19. 20) and: induit albos cum vitta crines (Aen. 7. 418), to which Servius notes: id est etiam vittas albas.

The repetition of the same word or of closely related forms at the beginning of consecutive clauses joins them as if by conjunctions; and it was probably through such repetition that conjunctions were first developed. Examples of our figure dependent on such repetitions we have in: nunc hos nunc accipit illos (Aen. 6. 315), nunc hos nunc illos aditus omnemque pererrat arte locum (5. 441), hic illius arma, hic currus fuit (1. 16), hunc equis, illum superare pugnis nobilem (Od. 1. 12. 26), huic mater quamvis atque huic pater adsit (Buc. 4. 56), quem mihi, quem tibi finem di dederint (Od. 1. 11. 1), pariterque (nunc) sinistros, nunc dextros solvere sinus (Aen. 5. 831), quae me fuga, quemve reducit? (10. 670), quantus equis, quantus adest viris sudor (Od. 1. 15. 9), genua amplexus, genibusque volutans haerebat (Aen. 3. 607), Nymphae, Laurentes Nymphae (8. 71).

It seems probable that: Troes trepidique Latini is short for: Troes trepidi trepidique Latini; for I read in the poets: indignanti similem similemque minanti (Aen. 8. 649), quam Turno regi aut regi adparere Latino (8. 17), caedebant pariter pariterque ruebant victores victique (10. 756), salve aeternum aeternumque vale (11. 97), graves nimium nimiumque severi (Mart. 8. 3. 17). But it may be for: trepidi Troes trepidique 269 Latini, for I also find: fortemque Gyan fortemque Cloanthum (Aen. 1. 222), nigris oculis nigroque crine decorum (Od. 1. 32. 11), inter tot curas totque labores (Ep. 2. 2. 66), despiciam dites despiciamque famem (Tib. 1. 1. 78). We have a third form of this fourfold union in: Teucro duce et auspice Teucro (Od. 1. 7. 27), omne caelum et mare omne (Ann. 2. 23. 3). Virgil joins the second of these with the elliptical form in: inde alios ineunt cursus aliosque recursus adversi spatiis, alternosque orbibus orbes impediunt (Aen. 5. 583-4).

The two terms of this fourfold union thus reduced to one by the ellipsis may be different. Variations in inflexion we have already noticed, such as we see in: qui dapibus mensas onerent et pocula (mensis im)ponant (Aen. 1. 706) or: vis rapuit rapietque gentes (Od. 2. 13. 20). We have different words approximating in sense in: gracili sic tamque pusillo (Sat. 1. 5. 69), hoc spatium tantumque morae fuit Ilo (Aen. 10. 400), non Liber aeque, non acuta sic geminant Corybantes aera (Od. 1. 16. 7-8), ab omni corpore seiunctum secretumque esse ab inani (Lucr. 1. 430), dirae facies inimicaque . . . numina (Aen. 2. 622), mitis ut in morem stagni placidaeque paludis (8. 88), scalae improviso subitusque apparuit ignis (12. 576). We may note the following parallel ellipses: iamque dies (unus) alterque dies processit (3. 356), (valde) salso multoque fluenti arbusto (Sat. 1. 7. 28), neque pugno (inerti) neque segni pede victus (Od. 3. 12. 9), Caecubum (vinum) et prelo domitam Caleno tu bibes uvam (1. 20. 9), eripite o socii (remos omnes), pariterque insurgite remis (Aen. 3. 560), instruimus mensas (et aras) arisque reponimus ignem (3. 231), diripiunt dapes (manibus foedis) contactuque omnia foedant inmundo (3. 227), et si fata deum (non inimica fuissent), si mens non laeva fuisset (2. 54). The figure passes into zeugma in: fundi Germanos acie et iustis locis (Ann. 2. 5. 3) for: acie iusta et aequis locis.

But the two terms thus reduced to one may primarily be pairs of correlated and even opposed ideas. We have such fully expressed in: ut ridentibus adrident, ita flentibus adsunt humani voltus (A. P. 101), concidunt venti fugiuntque nubes (Od. 1. 12. 30), informi limo glaucaque exponit in ulva (Aen. 6. 416). We have the following ellipses that seem parallel: lustramurque (sacris) Iovi votisque incendimus aras (Aen. 3. 279), fere spatio extremo (confecto) fessique sub ipsam finem adventabant (5. 327), his magnum Alciden contra stetit, his ego suetus (contendere) (5. 414), sed iaculis (eminus emissis) tutisque clamoribus instant (10. 713), neque nix . . . cana cadens violat (umquam), semper innubilus aether integit (Lucr. 3. 21), 270 discite iustitiam (colere) moniti et non temnere divos (Aen. 6. 620), o et praesidium (firmum) et dulce decus meum (Od. 1. 1. 2), tentabundus (pedibus) manibusque retinens virgulta (Liv. 21. 36. 1), ut credatur novissimum et sine terris (contra sitis) mare (Ann. 2. 24. 1).

We have them opposed in: hinc omne principium (defer), huc refer (omnem) exitum (Od. 3. 6. 6), nec curat Orion leones (feros) aut timidos agitare lyncas (2. 13. 39), vel quo discrimine ripas hae linquunt (repulsae) illae (acceptae) remis vada livida verrunt? (Aen. 6. 320), centum errant annos (exclusi) volitantque haec littora circum: tum demum admissi stagna exoptata revisunt (6. 329-30). In 6. 320 the fourfold figure has been reduced to two.

As a rule the position of the conjunction shows that it is the second of the pair of terms that is expressed, as we see in: gemitu miseroque tumultu or: Typhoeus aut validus Mimas. We saw that, while Cicero uses Castor for Castor et Pollux, Horace and Virgil use Pollux. So the poets in this figure tend to take the second of the two like terms; and when in a phrase like: abietibus patriis et (patriis) montibus aequos (Aen. 9. 674) the first is taken, the figure assumes the form usual in prose. Such examples of this figure, a form so obvious and common as not to merit the name of figure, we note in: claram Rhodon aut Mytilenen (Od. 1. 7. 1), et profestis lucibus et sacris (4. 15. 25), neglectum genus et nepotes (1. 2. 35), patiens pulveris atque solis (1. 8. 4), superis deorum gratus et imis (1. 10. 20), ignes per medios fluviosque (Sat. 2. 3. 57), argenti positi intus et auri (2. 3. 142), neque his fuga nota neque illis (Aen. 10. 757), incenditque animum dictis atque aggerat iras (4. 197).

This tendency in poetry often inverts prose order, giving a second place to the word that gets its name from its prior position in ordinary speech, the preposition or πρόθεσις; as in: nihil astra praeter vidit (Od. 3. 27. 31) or: spemque metumque inter dubii (Aen. 1. 218), or even removing it to a second clause, as in: chlamyde et pictis conspectus in armis (8. 588) or: quae nemora aut quos agor in specus? (Od. 3. 25. 2). Probably to this is also due the inversion of the usual order of names shown in poetry or later prose, as in: Maxime Lolli (Ep. 1. 2. 1), Musa . . . Antonius (1. 15. 3), Postumi Agrippae (Ann. 1. 6. 1), Gallus Asinius and Messala Valerius (1. 8. 4 and 5). We note it in commands, as in: sperne puer neque tu choreas (Od. 1. 9. 16), and in appeals, as in: adsis o placidusque iuves (Aen. 4. 578), and: quare agite o tectis, iuvenes, succedite nostris (1. 627).

Often the logical order of action is thus inverted, giving the figure 271 called Hysteron Proteron, as in: discere et audire (Ep. 1. 1. 48), condo et compono quae mox depromere possim (1. 1. 12), where the order of importance seems to prevail over that of time. So when Quintilian uses the order: Aristarchus atque Aristophanes (10. 1. 54) it is probably not from ignorance of chronology, as I have heard suggested, but from this poetic tendency which affects later prose more and more. It leads to such extended inversions as we have in:

Vidimus, o cives, Diomedem Argivaque castra,

Atque iter emensi casus superavimus omnes (Aen. 11. 243-4),


Postera Phoebea lustrabat lampade terras

Umentemque Aurora polo dimoverat umbram (4. 6-7).

So in: signoque repente corripiunt spatia audito limenque relinquunt (5. 316), aut tu mihi terram inice, namque poles, portusque require Velinos (6. 366), coniunx arma omnia tectis emovet, et fidum capiti subduxerat ensem (6. 524), nunc pateras libate Iovi precibusque vocate Anchisen genitorem et vina reponite mensis (7. 134), sus, quam pius Aeneas tibi . . . mactat sacra ferens et cum grege sistit ad aram (8. 84-5). The inversion is threefold in: castigatque auditque dolos subigitque fateri (6. 567).

Probably it is this same tendency that leads in poetry to the frequent removal of words from a principal to a subsequent relative clause, as in:

multa . . . fieri . . . tuentur

Quorum operum causas nulla ratione videre

Possunt (Lucr. 1. 152-4),

and: alii quorum Comoedia Prisca virorumst (Sat. 1. 4. 2) and: malarum quas amor curas habet . . . obliviscitur (Epod. 2. 37).



Prof. Clement Smith (Odes & Epodes of Horace, p. LXXI, 120) tells how close the connexion is between the distribution peculiar to poetic diction and the ellipsis with que; but he has not seen that both are based on the fourfold union of terms to which we traced that ellipsis. Take the last example he cites: quo beatus volnere, qua pereat sagitta (Od. 1. 27. 11); this question falls into two parts, and the phrase: beatus pereat, belonging to both, is distributed between them. When we restore it to both we have the two fourfold unions by the synchysis of which we get the Horatian form, quo beatus pereat volnere; qua beatus pereat sagitta. So in: nunc tempus equos, nunc poscere currus (Aen. 9. 12) the phrase distributed is: tempus poscere, and when we restore this to each we have the threefold unions restored each to fourfold form. So with: ense pedes nudo, puraque interrita parma (11. 711) pedes interrita is here the distributed phrase, and in: nec Fortunati spernit, nec balnea Fausti (Mart. 2. 14. 11) spernit balnea is distributed. So in: nil intentatum Selius, nil linquit inausum (2. 14. 1) Selius linquit is distributed, and in: pascat et Hybla meas, pascat Hymettus apes (7. 88. 7) meas apes. But it is not always so regular; in: discite iustitiam moniti, et non temnere divos (Aen. 6. 620) it is involved with the ellipsis with que, here for et, giving: discite iustitiam colere moniti, et non temnere divos, where we must supply discite moniti with the second part. In: proles indiscreta suis, gratusque parentibus error (10. 392), though the distribution of suis parentibus is simple, it is distributed to balanced pairs, proles indiscreta and gratus error.

But in reckoning the terms for my fourfold unions I have not taken into account et and que. Words for Aristotle fell into two classes, λόγοι and λέξεις. Λόγοι are those which retain their full meaning and their accent; λέξεις those which, like ἐστίν or τὶς, have lost their primary meaning and accent, and into this class fall prepositions and conjunctions as a rule, many pronouns and the copula ἐστίν. Further in counting terms, if ferocior is a single term, so will 273 be magis fidus; if annum, so centum annos; if minantur, so metum intendunt, or spem offerunt. So in: Anxuris ense sinistram, et totum clipei ferro deiecerat orbem (Aen. 10. 546) the words distributed are Anxuris deiecerat, but balanced with sinistram is: totum clipei orbem. In: quid memorem infandas caedes, quid facta tyranni effera? (8. 483) memorem tyranni are distributed, but infandas caedes and facta effera are balanced as single terms. And in: horror ubique animos, simul ipsa silentia terrent (2. 755) the distribution of animos terrent is involved with the use of simul for simul . . . simul. The puzzle in: te ante quam me amare rebar, ei rei firmasti fidem (Ter. Hec. 581) was rightly solved by Ursinus, though not according to our form; te me amare—quam rem—ante—rebar is the first fourfold union; the second is ei rei—(te me amare)—(nunc)—firmasti fidem, where we have the form we noticed in: discite iustitiam, etc., and supply two to the second from the first, one of which, nunc, is an equivalent of ante.

But in: has equidem memorare tibi atque ostendere coram iampridem, hanc prolem cupio enumerare meorum (Aen. 6. 716-17) we have evidently a triple union of such fourfold figures: has—memorare—tibi—iampridem cupio | has—ostendere—coram te—iampridem cupio | hanc prolem meorum—enumerare—coram te—iampridem cupio. So in: nam barbaris, quanto quis audacia promptus, tanto magis fidus rebusque motis potior habetur (Ann. 1. 57. 1) we have three fourfold unions: barbaris—quanto—magis promptus audacia—quis habetur | barbaris—tanto—magis fidus rebus motis—quis habetur | barbaris—tanto—potior rebus motis—quis habetur.

As if to convince us that such threefold unions arise out of fourfold ones, the Latin poets give us many examples of a fourfold union, to which is appended a threefold one thus arising. Some examples are: nox tibi longa venit, nec reditura dies (Prop. 2. 15. 24), dum nos fata sinunt, (dum) oculos satiemus amore (v. 23), nunc me fluctus habent, versantque in litore venti (Aen. 6. 362), his magnum Alciden contra stetit, his ego suetus (contendere) (5. 414), quem gravis ictu seminecem liquit, saxo lacerumque viator (5. 275), e quibus unus amet quavis adspergere cunctos praeter eum qui praebet aquam; post hunc quoque potus (Sat. 1. 4. 87-8). In: non ego avarum cum veto te fieri vappam iubeo ac nebulonem (Sat. 1. 1. 104) Horace has arranged his terms with such skill that it is hard to decide whether te fieri is to be constructed with veto or with iubeo; it is a fine example of the figure ἀπὸ κοινοῦ. In the following examples 274 we have the opposite order: si patriae volumus, si nobis vivere cari (Ep. 1. 3. 29), tum pueri nautis, pueris convicia nautae ingerere (Sat. 1. 5. 11). In: non, ita me divi, vera gemunt, iuverint (Catull. 66. 18) we have a threefold and a fourfold union, but involved in curious fashion. In: o quantum est auri pereat, potiusque smaragdi (Tib. 1. 1. 51) this union is involved with the que ellipsis; and we restore: o quantum est auri pereat potius, potiusque pereat quantum est smaragdi.

In: peccatum fateor, cum te sic tempore laevo interpellarim (Sat. 2. 4. 4) we have a twofold union followed by a fourfold; but: peccatum fateor is short for: a me peccatum esse tum fateor. We have the same sequence in: hoc unum (sibi peperit), iusso non moritura die (Prop. 4. 6. 64), di melius (rem decrevere); quantus mulier foret una triumphus (v. 65), sed bene Messalam (rem agere), sua quisque ad pocula dicat (Tib. 2. 1. 31), talibus Ilioneus (dixit simul), cuncti simul ore fremebant (Aen. 1. 559). We have the fourfold union doubled in: hoc primum; nec si miserum fortuna Sinonem finxit, vanum etiam mendacemque improba finget (2. 79-80); and in: at pater ut gnati, sic nos debemus amici si quod sit vitium non fastidire (Sat. 1. 3. 43). We have the opposite order in: hic nuptarum insanit amoribus, hic puerorum (1. 4. 27); forte epos acer ut nemo Varius ducit (1. 10. 44); quibus haec, sunt qualiacumque, adridere velim (1. 10. 88), where sunt qualiacumque is for: sint haec qualiacumque sunt; nunc hiemem inter se luxu, quam longa, fovere (Aen. 4. 193), where quam longa is for: tam longam quam longa est. This order is frequent in prose, as in: quam plurimis, modo dignis, se utilem praebeat (Off. 1. 92. 26), perventum inde ad frequentem cultoribus alium, ut inter montanos, populum (Liv. 21. 34. 1). We have the twofold union doubled in: quid mi igitur suades? ut vivam Maenius? aut sic ut Nomentanus? (Sat. 1. 1. 101), and two enclosing a fourfold one in: huc tandem concede; haec ara (simul) tuebitur omnes, aut moriere simul (Aen. 2. 523).

We have a twofold union shortened from four standing alone in: sic placitum (res esse futuras) (Aen. 1. 283) and: Aeneas haec de Danais victoribus arma (Apollini voveo) (3. 288). We have it shortened to one in: dissimulant (se velle dextras iungere) (1. 516), non tali auxilio, nec defensoribus istis tempus eget, non, si ipse meus nunc adforet Hector (2. 521-2), where non is for: non Hectore defensore egeret tempus; and so in: concurritur (Sat. 1. 1. 7) and tabescat (v. 111). In: sum deus; est nostri sanguinis ista fides 275 (Prop. 4. 6. 60) sum deus seems short for: sum nunc—deus factus—qui homo—olim fuerim. While we have twofold unions, the shortening of which from fourfold ones is not obvious, as in: quid faciam, si furtum fecerit? (Sat. 1. 3. 94), or: fuimus Troes, fuit Ilium et ingens gloria Teucrorum (Aen. 2. 325), or: noris nos; docti sumus (Sat. 1. 9. 7), they are comparatively rare.

Of the books of the Aeneid six begin with a sentence filling from five to seven verses; six with a shorter and less complex sentence of from one verse to two verses and a half. Let us look at the structure of these more simple beginnings.

The second Aeneid begins: conticuere omnes; intentique (omnes) ora tenebant. We have a union of two here, followed by one of three shortened from four. Book IV begins:

At regina gravi iamdudum saucia cura

Volnus alit venis et caeco carpitur igni;

where we have a fourfold union: regina saucia gravi cura—iamdudum—volnus alit—venis; and a twofold: carpitur—caeco igni, with which regina and iamdudum are to be supplied. Book VI begins:

Sic fatur lacrimans classique inmittit habenas,

Et tandem Euboicis Cumarum adlabitur oris.

It seems a sentence of threefold structure, each of whose parts is a union of three terms, the last being: tandem—adlabitur—Euboicis Cumarum oris. But we must remember that to each must be supplied the subject Aeneas. The seventh Aeneid begins:

Tu quoque litoribus nostris, Aeneia nutrix,

Aeternam moriens famam Caieta dedisti.

It consists of a sentence of fourfold structure: tu quoque—litoribus nostris—aeternam famam—moriens dedisti, and a description of the subject shortened from four to three: Caieta—Aeneia—olim—nutrix. The ninth begins:

Atque ea diversa penitus dum parte geruntur,

Irim de caelo misit Saturnia Iuno

Audacem ad Turnum;

—a sentence of twofold structure, each part being fourfold. We may analyse: atque dum—ea—geruntur—diversa penitus parte ‖ Saturnia Iuno—Irim—audacem ad Turnum—de caelo misit. The eleventh Aeneid begins: Oceanum interea surgens Aurora reliquit; to be divided: interea—Aurora—surgens reliquit—Oceanum. We have in these six initial sentences five fourfold unions, four shortened from four to three, and one twofold. We can see that the unions of two and four terms play a great part in poetic phrasing.


This becomes still more striking when we turn to the lyrics of Horace. I take a stanza marked by the absence of λέξεις:

Parcius iunctas quatiunt fenestras

Ictibus crebris iuvenes protervi,

Nec tibi somnos adimunt, amatque

Ianua limen (Od. 1. 25. 1-4).

You notice how the first and second verses are of four words each, constituting four pairs, arranged alternately in the first verse, but in order of pairing in the second, and making up a sentence of fourfold structure; iuvenes protervi—ictibus crebris—iunctas fenestras—parcius quatiunt. The second is threefold for fourfold: (iuvenes)—tibi—somnos—neque adimunt, and the third is threefold for twofold: ianua—amat—limen; for amat limen is poetic for claudi solet.

The second stanza is still more clearly fourfold in structure, if less strikingly fourfold in the arrangement of words:

Quae prius multum facilis movebat

Cardines; audis minus et minus iam:

Me tuo longas pereunte noctes,

Lydia, dormis?

It consists of three unions of fourfold structure: quae—multum facilis—prius—multum movebat cardines ‖ audis—minus—et minus—iam ‖ tu Lydia—dormis—longas noctes—me tuo pereunte.

Let us turn to the next ode, where we meet the Alcaic strophe:

Musis amicus tristitiam et metus

Tradam protervis in mare Creticum

Portare ventis, quis sub Arcto

Rex gelidae metuatur orae,

Quid Tiridaten terreat, unice

Securus. (Od. 1. 26. 1-6).

If we leave the λέξεις: out of account, we have four words in each verse. The whole sentence may be regarded as threefold in structure, consisting of a principal clause, and two questions each dependent on unice securus. Each part is threefold and may be analysed: Musis amicus—protervis ventis—tristitiam et metus—tradam portare (= portanda) in mare Creticum ‖ unice securus—quis rex gelidae orae—metuatur—sub Arcto ‖ unice securus—quid—Tiridaten—terreat.

Turn now to the first two verses of the Odes:

Maecenas atavis edite regibus

O et praesidium et duke decus meum.

We notice (1) that: O Maecenas is distributed between the first and second verses, (2) that from dulce we must imply firmum with praesidium, 277 (3) that meum is evidently poetic for mihi. So we have: O Maecenas—atavis—edite—regibus ‖ O Maecenas—et praesidium firmum—et dulce decus—mihi.

Let us now turn to the first two verses of Propertius:

Cynthia prima suis miserum me cepit ocellis

Contactum nullis ante cupidinibus.

We notice at once that the pentameter is of four words; when we join Cynthia prima, suis ocellis, miserum me, the fourfold structure of the hexameter is clear. When we join nullis cupidinibus, the structure of the pentameter is threefold, but it must be attached to me, restoring its fourfold character. Let us turn to the first distich of Tibullus’s third elegy:

Ibitis Aegaeas sine me, Messala, per undas;

O utinam memores ipse cohorsque mei.

The structure is twofold; and we may analyse:

O Messala—ibitis—sine me—per undas Aegaeas,

Utinam—o ipse cohorsque—memores sitis—mei.

Though we are not dealing specially with prose diction, we may here analyse the structure of the first three sentences of the Bellum Gallicum: Gallia—est divisa—omnis—in partes tres ‖ quarum—unam—incolunt—Belgae ‖ (quarum)—aliam—(incolunt)—Aquitani ‖ (quarum)—tertiam—(incolunt—ei) ‖ qui—ipsorum lingua—Celtae—(appellantur) ‖ (qui)—nostra (lingua)—Galli—appellantur. ‖ Hi omnes—lingua—institutis—legibus—inter se—differunt. ‖ Gallos—ab Aquitanis—Garumna flumen—(dividit) ‖ (Gallos)—a Belgis—Matrona et Sequana—dividit. I have set est with divisa, where it belongs, but Caesar set it between Gallia and omnis, marking off omnis as an independent term. We see in the second sentence three fourfold unions; for with lingua—institutis—legibus severally, we must supply the three remaining terms hi omnes—inter se—differunt. The three sentences are made up of eleven unions, all of them fourfold.

We notice how fourfold unions are repeated: in the opening verses of the sixth Aeneid we have a good example of four threefold unions, followed by one fourfold, and another threefold:

Sic fatur lacrimans ‖ classique inmittit habenas ‖

Et tandem Euboicis Cumarum adlabitur oris. ‖

Obvertunt pelago proras; ‖ tum dente tenaci

Ancora fundabat naves ‖ et litora curvae

Praetexunt puppes.

But Aeneas is to be supplied with the first three, Aeneadae with 278 the fourth, and tum with the sixth, making them all fourfold. We have in:

Te Dacus asper, ‖ te profugi Scythae ‖

Urbesque gentesque ‖ et Latium ferox

Regumque matres barbarorum ‖ et

Purpurei metuunt tyranni ‖

Iniurioso ne pede proruas

Stantem columnam ‖ neu populus frequens

Ad arma cessantes ‖ ad arma

Concitet ‖ imperiumque frangat (Od. 1. 35. 9-16),

a like series of unions really fourfold, but apparently twofold and threefold; and in:

Cum rapies in ius, malis ridentem alienis,

Fiet aper, modo avis, modo saxum, et cum volet arbor (Sat. 2. 3. 73-4),

we seem to have a series of three unions apparently threefold with three pairs to balance them.

We noticed how the ellipsis with que arose from unions like: nigris oculis nigroque crine, or: caedebant pariter pariterque ruebant. We find the distribution of pairs occurring in the former order in: parcius iunctas quatiunt fenestras. We have this alternate distribution in:

Non ego Myrmidonum sedes Dolopumve superbas

Aspiciam, aut Graiis servitum matribus ibo (Aen. 2. 785-6),

Me famulo famulamque Heleno tramisit habendam (3. 329),

Ex utraque pari malarum parte profusa est (Lucr. 1. 88),

where parte must be taken twice as: ex utraque parte malarum pari parte (= modo), a case of zeugma. In: non, ita me divi, vera gemunt, iuverint (Catull. 66. 18) we have this alternation applied to a threefold and a fourfold union, as though they were each twofold.

We have the second form in: integer vitae scelerisque purus (Od. 1. 22. 1), latis equa trima campis (3. 11. 9), and a shorter form of this arrangement in: meam canto Lalagen (1. 22. 10), curis vagor expeditis (v. 11), silva lupus in Sabina (v. 9). This form is favoured by Catullus in his Berenice’s Lock, where we read: omnia qui ‖ magni dispexit lumina mundi (v. 1) in the fourfold, and: dulcis amor ‖ gyro devocet aerio (v. 6) in the threefold form; and so in: ut cedant ‖ certis sidera temporibus (v. 4), qui vix sero ‖ alto mergitur Oceano (v. 68), and: levia protendens bracchia ‖ pollicitast (v. 10). We have these threefold unions paired in: qua rex tempestate ‖ novo auctus hymenaeo (v. 11) and: quis ego pro factis ‖ caelesti reddita coetu (v. 37).



In both the ellipsis with que and the usual forms of distribution we have examples of the expression of a union really fourfold by three terms; so that we can readily understand the full meaning of: fortes creantur fortibus et bonis (Od. 4. 4. 29). When the pair of words, one of which is omitted, are the same, as in: arma (ferte), viri, ferte arma (Aen. 2. 668), or: nate fuge (fuge) nate (2. 733), we feel that the omission is easy and natural. Easy too are most of the ellipses where the word is simply varied in inflexion, as in: Gratia (nuda) nudis iuncta sororibus (Od. 3. 19. 17), ubi acris invidia (viget) atque vigent ubi crimina (Sat. 1. 3. 61), adeo maxima quaeque (maxime) ambigua sunt (Ann. 3. 19. 3). Sometimes there is doubt as to which was the omitted word; Servius’s note to: (credas) monies concurrere montibus altos (Aen. 8. 692), alii altis legunt, unum tamen est, makes this clear. When we have the opposite pleonasm, as in: res si qua diu mortalibus ulla est (10. 861), where ulla is pleonastic, we pass it over without perceiving it. Indeed in: sive reges (divites) sive inopes erimus coloni (Od. 2. 14. 12), where the word omitted is the opposite of that expressed, we are apt to overlook the omission.

Easy is the ellipsis of eis in: atqui licet esse beatis (Sat. 1. 1. 19), but when we come to: dederim quibus esse poetas (1. 4. 39), many editors write poetis, not perceiving the ellipsis; for the fourfold expression is: quibus dederim eos esse poetas, giving a meaning quite different from that they assume. So natural is the ellipsis in: his me consolor (me) victurum suavius (Sat. 1. 6. 130) that, when we have the full expression in: ut tandem agitando se se movere (Liv. 21. 58. 10), we take se se as equivalent to se. The Romans seem to have fallen into the same mistake; for from such constructions seems to have arisen the use of sese for se in Latin. This was the easier, as in conversation and poetry se is often omitted with the verb, as in: accingunt omnes operi (Aen. 2. 235), and the enclitic se attaches itself quite naturally to the preceding word, becoming thus the second syllable of the new word sese. We have a good example of the 280 opposite, the use of se for se se in: iureiurando obstrinxit se non excessurum (Tac. Ann. 1. 14. 6).

We do not feel the ellipsis in: mixtae (mixtis) pueris puellae (Od. 4. 11. 10), as the same ellipsis is usual in English, e.g. in ‘good mixed with bad’. We have the same feeling about: procella velum (adversum) adversa ferit (Aen. 1. 103), adversi rupto seu quondam turbine venti (cum adversis) confligunt (2. 416), genibus (adversis) adversae obluctor harenae (3. 38), qui se ignotum (ignotis) venientibus ultro . . . obtulerat (2. 59). And generally, when it is the first of the pair that is omitted, we are less apt to notice the ellipsis, as in: materies quia rebus (certis) reddita certa est (Lucr. 1. 203), per (tot) varios casus, per tot discrimina rerum (Aen. 1. 204), cernes urbem (promissam) et promissa Lavini moenia (1. 258), scaenis decora alta (futura) futuris (1. 429), trabes . . . devolvunt (alii), alii strictis mucronibus imas obsedere fores (2. 449), date (lymphas), volnera lymphis abluam (4. 683), amicus dulcis . . . cum (meis) mea compenset vitiis bona (Sat. 1. 3. 70), lanea et effigies erat (altera), altera cerea (1. 8. 30), cum Velabro (omni) omne macellum (2. 3. 229).

Perhaps it is for the same reason that we are slow to see the ellipsis at the beginning of a poem, as in: (commotum) motum ex Metello consule civicum (Od. 2. 1. 1), or of a phrase in: mandata (data) Clementi centurioni quae perferret (Ann. 1. 26. 1). Tacitus writes: datis mandatis (3. 8. 1), and probably in filling the ellipsis Horace would have read: (commotum) motum. The ellipsis is easy to supply in: causasque (causis) innecte morandi (Aen. 4. 51), as causas comes first; and so in: curvam (curvis) compagibus alvum (2. 51). More involved is it in: si . . . perituraeque addere Troiae teque tuosque (perituros) iuvat (2. 661), huic ego volgus errori (errorem) similem cunctum insanire docebo (Sat. 2. 3. 63); and there is a slight variation of sense in: tum litore funem (laxatum) deripere excussosque iubet laxare rudentes (Aen. 3. 267). This variation becomes more marked in: magni quo pueri magnis e centurionibus orti (Sat. 1. 6. 73), and for: ductus Neptuno sorte sacerdos (Aen. 2. 201) the full expression would be: factus Neptuno sorte ducta sacerdos. So for: cui fata parent (2. 121) the full expression is: cui oracula mortem parent, and in fata we have an amphibole, whereby it serves as both subject and object to parent.

Ut is at times omitted in the predicate, giving greater vividness to the comparison, says Prof. Clement Smith, by completely identifying the subject with it, as in: totidem plagis consumimus hostem, lento 281 (ut) Samnites . . . duello (Ep. 2. 2. 97). The omission is at times due to the reduction of four terms to three, as in: quid mi igitur suades? ut vivam (ut) Maenius? aut sic ut Nomentanus? (Sat. 1. 1. 101). We have the first ut omitted in: illuc unde abii redeo, (ut) nemo ut avarus se probet (1. 1. 107). The fact that in each case the ut expressed has developed a marked difference of meaning from that omitted, is parallel to what we noted above. Probably in our first example the omission of ut is due to totidem in the preceding clause. It is easier in: (ita) dives ut metiretur nummos; ita sordidus ut se non umquam servo melius vestiret (1. 1. 95) because of the second expression being in full. Nor is there difficulty in: furit ac velut ursus . . . recitator acerbus . . . occidit legendo . . . plena (velut) cruoris hirudo (A. P. 476), simplicior quis (est) et est (talis) qualem me saepe libenter obtulerim tibi (Sat. 1. 3. 63), Antoni (sic amicus) non ut magis alter amicus (erat) (1. 5. 33). But much more difficulty has been offered by: frater erat Romae consulti rhetor (et amore adeo fraterno) ut alter alterius sermone meros audiret honores (Ep. 2. 2. 87-8).

We have this ellipsis with primus in: cum prorepserunt primis (prima) animalia terris (Sat. 1. 3. 99), qua prima fortuna saluus monstrat (primum) iter (Aen. 2. 387), helping us to understand it in: nam primum quemque necessest (primum) occupet ille locum (Lucr. 1. 389), cum acre fluit frigus, non primam quamque solemus particulam venti (primum) sentire et frigoris eius (id. 4. 260-1), where Munro has wrongly written privam, an emendation of Gifanius. So too in: ut noscas . . . primum iactum fulgoris quemque (primum) perire (5. 291), primum quicquid fulgoris (primum) disperit (5. 284), primum quicquid aquai (primum) tollitur (5. 264), primum quicquid flammarum (primum) perdere semper (5. 304).

The construction was examined with great care by Madvig in his note to: nec enim absolvi beata vita sapientis, neque ad exitum perduci poterit, si prima quaeque bene ab eo consulta atque facta, ipsius oblivione (prima) obruentur (Cic. Fin. 2. 105. 32); and yet he failed to perceive its origin, which is plain from some of the examples he cites. For he gives us: ut et prima quaeque pars, ut exposita est in partitione, sic ordine transigatur (Inv. 1. 33. 23), where ordine is plainly short for primo ordine. Near the end of this same section we have in: quemadmodum igitur hic et ad primam quamque partem primum accessit, et omnibus absolutis finem dicendi fecit, the construction in full, as we should expect to find it in Cicero’s careful 282 prose. But he has it usually in the shortened form, as in: primum quidque explicemus (Fam. 12. 1. 1). Other examples are: ut primus quisque acervos (primos) demittatur per serias ad vasa olearia (Varro R. R. 1. 55. 5), opus quod prima quaeque (primum) perficiendo minui videbatur (Liv. 31. 1. 5), primas quasque aquas (primum) explicat (Sen. N. Q. 6. 17. 2), prima quaeque ut (primum) absolveris, mittito (Plin. Ep. 8. 4. 6), Africus vero prima quaeque congesta pulsu illisa maris (primum) subruit (Curt. 4. 2. 8).

As Madvig clearly saw, it has nothing to do with the similar phrase in: ut exercitui diem primam quamque diceret ad conveniendum (Liv. 42. 48. 4) ‘the earliest day possible’, a development from: primo quoque die ad senatum referant (Cic. Phil. 8. 33. 11), short for: quocumque die primo possint, etc. The phrase: proximum quodque, as Madvig saw, is parallel to primum quodque and must be solved in the same way. We have examples of it in: ne proxima quaeque (primum) amoliendo maioribus gravioribusque aditum ad se facerent (Liv. 33. 12. 11), quid agam si proxima quaeque (prima) relinquunt (Ov. Trist. 5. 2. 39). Again in: igitur antiquissimae cuique (epistulae) primum respondebo (Att. 9. 9. 1) Cicero gives us the full form usually shortened in this construction.

The varied use of tenses with iamdudum, iamdiu, iampridem, iam olim, has given some trouble. Iamdudum, the most common in classical authors, is usually joined with the present, as in: iamdudum et frustra cerno te tendere contra (Aen. 5. 27), where we translate the present by a perfect, ‘I have long perceived’. But I read in Tibullus: iamdudum Syrio madefactus tempora nardo debueram sertis implicuisse comas (3. 6. 63-4). Iam-dudum ‘now for some time’ consists of iam, which usually takes a present, and dudum taking a perfect (or imperfect), as in: egomet dudum Beroen . . . reliqui aegram (Aen. 5. 650) or: quem dudum non ulla iniecta movebant tela (2. 726). It would be natural to expect for: iamdudum video (Sat. 1. 9. 15) the fourfold union: iam video—dudum vidi. This seems to have been ordered: iam dudum video vidi, and then reduced from four to three; of the two verbs video was retained as the stronger and more vivid.

In the English ‘I have long seen’ it is the perfect that is retained; for that we feel to be the stronger form. In Latin, too, when one of a pair is retained it is oftener the second, as in: te maximus Actor (ante gessit), te Turni nunc dextra gerit (Aen. 12. 96-7). So we might expect to find the perfect (or imperfect) joined with iamdudum 283 as well as the present. And when we turn to archaic Latin we find the perfect quite frequently, as in: iamdudum, si des, porrexi manum (Pl. Pseud. 1148), iamdudum audivi (Merc. 953), iamdudum factumst quom abiisti domo (Trin. 1010), quia non iamdudum ante lucem ad aedem Veneris venimus (Poen. 318), iamdudum, mulier, tibi non imprudens advorsabar (Men. 419), Py. An abiit iam a milite? Ch. Iamdudum, aetatem (Ter. Eun. 734), Ct. Ain patrem hinc abiisse rus? Sy. Iamdudum. Ct. Die sodes. Sy. Apud villamst (Ad. 517), audivi, Archylis, iamdudum: Lesbiam adduci iubes (And. 228), to which Donatus’s note is: utrum iamdudum audivi an iamdudum iubes incerta distinctio. He does not know whether to join iamdudum with audivi or with iubes, a difficulty which does not trouble our editors, who here, strange to say, prefer to join it with audivi. Probably in the last two examples we have an amphibole in the use of iamdudum, which can be joined with either perfect or present; as we seem to have in: visa iamdudum prosilit altis diva toris (Val. Fl. 6. 456).

When the continued act is transferred from the present to the past, we have the present with iamdudum or iampridem passing to the imperfect, as in: erat ei de ratiuncula iampridem apud me relicuom pauxillulum nummorum (Ter. Phorm. 37), iamdudum flebam (Ov. Met. 3. 656). The perfect with iamdudum would naturally pass to the pluperfect; and this seems the explanation of debueram in Tib. 3. 6. 64. We have it transferred to the future in: in medios belli non ire furores iamdudum moriture paras? (Luc. 2. 524), iamdudum aetherias eadem reditura sub auras (Stat. Theb. 6. 857). We find it with the imperative in: iamdudum sumite poenas (Aen. 2. 103), where Servius explains it as for quam primum. To understand the syntax here, we must return to the fourfold form: iam poenas sumite; dudum poenas sumpsissetis,—to be rearranged: iamdudum poenas sumite sumpsissetis, and then reduced to three. We have the fourth and not the third term in: ut ego hue iamdudum simitu exissem vobiscum foras (Pl. Stich. 743). In the sense of quam primum it is transferred to the present subjunctive in: candida iamdudum cingantur colla lacertis (Ov. Ars 2. 457); and to the infinitive in: ingenti iamdudum de grege duci iussit (ib. 1. 317). We have iampridem with the imperative in: tollite iampridem victricia tollite signa (Luc. 1. 347), a construction clearly parallel to that just noticed, and pointing to the primary use of duci.

But we often find this reduction from four to three brought about by the omission, not of a second form of the same word, but of 284 a word correlated in meaning or use to that expressed. So in: (eo) parto quod avebas (Sat. 1. 1. 94), nam displosa sonat quantum vesica (tantum) pepedi diffissa nate ficus (1. 8. 46), toga quae defendere frigus, (sit tam) quamvis crassa, queat (1. 3. 15). Both of course are often expressed, as in: nil satis est, inquit, quia tanti quantum habeas sis (1. 1. 62). Very usual in Tacitus is such an ellipsis with comparatives and superlatives, as in: quem haud fratris interitu (magis) trucem quam remoto aemulo aequiorem sibi (futurum) sperabat (Ann. 3. 8. 1), optimus quisque reipublicae cura (maxime) maerebat (3. 44. 2). We have it with numerals in: satis una superque vidimus exscidia et captae (semel) superavimus urbi (Aen. 2. 643).

We find it with words not thus related in the following examples, where I bracket the word omitted, and italicize the word suggesting it to the reader: facti de nomine Byrsam (appellatam) (Aen. 1. 367), lyra (dulci) vel acri tibia (Od. 1. 12. 1), tu (supremus) secundo Caesare regnes (1. 12. 51), ob Italiam (negatam) terrarum clauditur orbis (Aen. 1. 233), fortia facta patrum, series longissima rerum (gestarum) (1. 641), et breviter (narratum) Troiae supremum audire laborem (2. 11), artificumque manus inter se (certantium) operumque laborem miratur (1. 455), cognatos nullo natura labore (tuo) quos tibi dat (Sat. 1. 1. 88), tu semper urges flebilibus modis (lugere) Mysten ademptum (Od. 2. 9. 9), tum vero manifesta fides (mala), Danaumque patescunt insidiae (Aen. 2. 309), non haec, o Palla, dederas promissa parenti (precanti) cautius ut saevo velles te credere Marti (11. 153-4).

In several of these we have two or more words to suggest the word to be supplied, but in each they are joined and present a single idea. That is not the case in the following, where two separate words in the phrase or sentence join in suggesting to the reader the word to be supplied: tendoque supinas ad caelum cum voce (sublata) manus (Aen. 3. 177), vocat (ad mortem) lux ultima victos (2. 668), fundamenta quatit totamque a sedibus (imis) urbem eruit (2. 611), omnis spes Danaum et coepti fiducia belli (bene gerendi) (2. 162), mihi parvus lulus sit comes, et longe (secuta) servet vestigia coniunx (2. 711), ingens a vertice pontus (deruens) in puppim ferit (1. 114), quae vos a stirpe parentum (olim amotos) prima tulit tellus, eadem vos ubere laeto accipiet reduces (3. 94-6), non me tibi Troia externum tulit, aut cruor hic (alienus sanguis hoc) de stipite manat (3. 43), at regina dolos praesensit motusque excepit prima futuros omnia (etiam) tuta timens (4. 296-8), where etiam is Servius’s suggestion.

In return we have two ellipses suggested by a single word expressed 285 in: fortunae (malae) miseras auximus (prava) arte vias (Prop. 3. 7. 32), longa (dictu) est iniuria, longae ambages; sed summa sequar (narrando) vestigia rerum (Aen. 1. 341), cithara crinitus Iopas (tectis) personat aurata (quae carmina) docuit quem maximus Atlas (1. 741).

In each of the following examples we have two ellipses, each suggested by a distinct word in the context: nos, tua progenies caeli quibus adnuis arcem (nostram futuram) (Aen. 1. 250), adparent rari nantes in gurgite vasto, (obscura videntur) arma virum tabulaeque et Troia gaza per undas (1. 118-19), illi me comitem . . . pauper in arma pater primis (et belli et pucritiae) huc misit ab annis (2. 87), arduus (exstans) armatos mediis in moenibus astans fundit equus, victorque Sinon incendia (caedibus) miscet (2. 328-9), sanguine quaerendi reditus, animaque litandum Argolica . . . obstipuere animi . . . cui fata parent (Argivo), quem poscat Apollo (inmolandum) (2. 118-21), non haec tibi litora (petere) suasit Delius, aut (hic) Cretae iussit considere Apollo (3. 161-2), quem sese ore ferens, quam forti pectore et armis (validis instructus) (4. 11), et nos tela (certa), pater, ferrumque haud debile dextra (gerimus et) spargimus (12. 50), hac (aequa) lege (concessa) in trutina ponetur eadem (Sat. 1. 3. 72), ut tuto ab atris corpore viperis dormirem et (atris) ursis . . . non sine dis (tuentibus) animosus infans (Od. 3. 4. 17-20), where ater varies in sense, being felt in the ellipsis in its proper meaning as the colour of bears, but figurative in the context of the poison of the vipers. In: donec ornus . . . supremum congemuit traxitque (secum) iugis avolsa ruinam (aliarum arborum) (Aen. 2. 631) the last ellipsis is that suggested by Servius, and it seems justified by the secum evidently implied in traxit, as well as by the ornus avolsa, which is the subject of the sentence.

In: quid tantum Oceano properent se tingere soles hiberni, vel quae (= unde tanta) tardis mora noctibus (hibernis) obstet (Aen. 1. 745-6) we have a twofold ellipsis arising from three words italicized in the context; and the same seems true of: Iuppiter omnipotens, precibus si flecteris ullis, adspice nos (voltu laeto); hoc tantum (precor) (2. 689-90). In: dum pecori lupus (inimicus pascua turbaret) et nautis infestus Orion turbaret hibernum mare (Epod. 15. 7-8) Horace has so nicely balanced his threefold ellipsis with the three terms suggesting it that it is not hard to trace. In: quanto rectius hoc, quam tristi laedere versu Pantolabum (Sat. 2. 1. 21) the last four words clearly suggest: alto laudare versu Caesarem as the phrase for which hoc stands; and in: saltat Milonius, ut semel icto accessit fervor capiti, numerusque (additur) lucernis (iam incertis) (2. 1. 24-5) we have 286 a threefold ellipsis suggested by the three words italicized. In: hoc (iure = modo) etenim sunt omnes iure (= in iudicio) molesti, quo (iure = modo) fortes (molesti sunt), quibus adversum bellum incidit (Sat. 1. 7. 10-11), where the last phrase seems equivalent to the prose: in bello, we have a curious alternation of meaning in iure. With hoc and quo it takes a sense closely allied to that seen in: quo iure quaque iniuria praecipitem me in pistrinum dabit (Ter. And. 214) ‘rightly or wrongly he will send me post-haste to the mill’, while with molesti it seems to stand for in iudicio; cf. in ius ambula (Phorm. 936); and to be opposed to: in bello, here amplified by Horace as we have indicated.

Tacitus in his prose is a close and skilful imitator of Virgilian diction; but though his brevity disposes him to the use of ellipses, he has not used this figure so much as he has some closely allied with it. I cite here some examples of his ellipses: (se) moriturum potius quam (id commissurum ut) fidem exueret clamitans (Ann. 1. 35. 5), etiam (erant) quorum diversa (essent) oppida, (qui) tamen obvii (facti) et victimas (mactantes) atque aras dis manibus statuentes, lacrimis et conclamationibus dolorem testabantur (3. 2. 3), nam secutae leges, etsi aliquando in maleficos ex delicto (commisso), saepius tamen dissensione ordinum et (cupidine) apiscendi inlicitos honores aut (patria) pellendi claros viros aliaque ob prava latae sunt (3. 27. 2). In: raro ea tempestate et e vetere memoria facinore decimum quemque ignominiosae cohortis sorte ductos (= sorte ducta electos) fusti necat (3. 21. 1) we may paraphrase raro . . . facinore thus: exemplo eo tempore raro et vetere ex memoria prave adhibito. In: eam condicionem esse imperandi ut non aliter ratio constet quam si uni reddatur (1. 6. 6) the latter half, ut . . . reddatur, is a union of question and answer skilfully involved to hide the real structure. We may arrange the whole sentence in threefold form: imperandi condicionem esse eam: ut imperio Romano constet ratio? Non aliter quam si uni (i.e. principi) reddatur (ratio). The reader is likely to be led astray at first by the apparent union of ut with imperandi.

Of a pair of objects, qualities, or relations, opposed to each other, one may be omitted which the other naturally suggests, and often so naturally that we are but obscurely, if at all, conscious of the omission. We have the fourfold groups of such objects without any omission in: ego quid sit ater Hadriae novi sinus, aut quid albus peccet Iapyx (Od. 3. 27. 19) and: undique ex insidiis barbari a ronte a tergo coorti, comminus eminus (eos) petunt (Liv. 21. 34. 6). 287 But we read: omnem oram . . . partim renovandis (veteribus) societatibus, partim novis instituendis Romanae dicionis fecit (21. 60. 3), where the veteribus is so plainly implied in novis and renovandis that we hardly notice its omission. The following are also simple examples: quaerenti talibus ille (respondit) suspirans (Aen. 1. 370), patrios foedasti (fili) funere voltus (2. 539), superstes restarem ut (filio) genitor (11. 160), at non Evandrum potis est vis ulla tenere (domi), sed venit in medios (11. 148), hinc spargere voces in volgum ambiguas, et quaerere (doli) conscius arma (2. 99), nec iam amplius armis (eas adgredi), sed votis precibusque iubent exposcere pacem (3. 260).

Though this figure is not common in Augustan prose, I have noted a fine example of it in Cicero’s oration for Milo: quam ob rem ilia arma, centuriones, cohortes non periculum nobis, sed praesidium denuntiant (3. 1). We should have expected: non periculum nobis denuntiant, sed praesidium promittunt; but Cicero in the bitter irony that marks his account of Pompey’s measures to protect the court in this trial, measures that so overawed him as to cause the failure of the oration he then delivered, has chosen in the oration he wrote later to transfer denuntiant to praesidium, giving us a marked oxymoron and a variety of zeugma that reached its highest development in Tacitus’s prose. I have noted from Tacitus: Seio Tuberoni legato tradit equitem campumque . . . quod arduum sibi (sumpsit), cetera legatis permisit (Ann. 2. 20. 2), which Furneaux calls a zeugma; and: solum veneni crimen visus est diluisse, quod ne accusatores quidem satis firmabant, in convivio Germanici, cum super eum Piso discumberet, infectos (huius) manibus eius (= illius) cibos arguentes (3. 14. 2), where eius seems to fill the place both of huius, i.e. Pisonis, and illius, i.e. Germanici, and to be a difficult zeugma.

We may note this threefold example in Propertius:

Non tulit haec Paetus, stridorem audire procellae

Et duro teneras laedere fune manus,

Sed (gaudebat) Chio thalamo (dormire) et Oricia terebintho

(recumbere quietum) et fultum pluma versicolore caput (3. 7. 47-50).

We have the following in Martial:

Hoc ego maluerim quam si mea carmina cantent

Qui Nilum ex ipso protinus ore (ad ultimos fontes) bibunt (7. 88. 6);

and in the epigram addressed to the Emperor Nerva:

Et te (imperatorem) privato cum Caesare Magnus (privatus) amabit (11. 5. 11).


We have the following two interesting examples in Lucretius:

Et quae res nobis vigilantibus obvia mentes

Terrificet morbo adfectis, somnoque sepultis (quamvis sanis) (1. 132-3).


haud igitur quidquam procedere posset,

Principium quoniam (pro)cedendi (cedendo) nulla daret res (1. 338-9),

‘since nothing could give a beginning of advance by its retreat’, a solution that has occurred to no one, though an examination of the reason for omitting the preposition in composition will make it seem probable.

We have found few problems of any difficulty connected with this variety of ellipsis in Virgil; but one, which I will now state, raises the question how far that is due to Virgil, how far to emendations of his text by later grammarians. Hyginus in Gellius (10. 16. 14) quotes Aen. 6. 840 thus:

Ultus avos Troiae, templa †intemerata Minervae.

The obelisk was set before intemerata by Martin Hertz, my old teacher, who thus indicated that he did not understand the reading, but that it was that of the best manuscripts of Gellius. The courage and veracity that moved him to this are both admirable in themselves and of the highest value as an example to students of Latin. When we turn from the text of Gellius to that of the Aeneid, the problem is no longer so simple. All Virgil manuscripts give the reading adopted by Ribbeck: ultos avos Troiae, templa et temerata Minervae; and this is the reading of Servius, who explains temerata: per stuprum Cassandrae.

And yet we have reason to pause; Gellius’s citation of the verse is the oldest authority we have on the reading. The question is, which of the two readings proceeded from Virgil, which was the emendation of some grammarian. It cannot be a question of the careless error of some scribe; Gellius prized a correct text too highly for that. The more clearly we show that all students of letters among the ancients knew of Ajax’s violation of Minerva’s fane, the less likely becomes the change from Virgil’s temerata to the intemerata of some unblessed grammarian. It is far more likely that some grammarian in early days determined to relieve Virgil’s text from the apparent inconsistency involved in: intemerata, and emended it to: et temerata. When we turn to the figure with which we are now busied, we see at once that: templa intemerata Minervae will be the natural reduction 289 to three of the fourfold: templa temerata intemeratae Minervae; for we have noticed that in determining the resultant of a pair like temerata intemeratae the second term is apt to have the greater influence. But you will ask, What could: templa intemerata Minervae mean to Virgil’s reader in his day? It seems quite evident to me that the reader of Virgil in Virgil’s day understood the common reduction of the threefold from the fourfold union, and would mentally substitute the latter for the former here. I have further shown how often we supply with an adjective the participle futurus, giving us here templa Minervae intemerata futura. What would be the effect of the vengeance we have here described, but to ensure that Minerva’s shrine would be inviolate for the future? Of such varied interest are the ideas called up in the mind of the reader by the reading intemerata, that compared with it temerata, the text of Servius and the manuscripts, seems plain and prosaic.

As we saw in: temerata intemeratae, at times from a union of two words, one of which is syntactically dependent on the other, the governing word is omitted and the dependent retained and put in the case, or in the mood and tense of the word omitted. We have a good example in: Teucrum arma quiescant et Rutuli (Aen. 12. 79), where Rutuli is for arma Rutulorum, and where variety of expression seems to be the aim rather than brevity, which seems the poet’s object in most examples of this change. So in: stupet Albius aere (Sat. 1. 4. 28) for: aeris splendore, lituo tubae permixtus sonitus (Od. 1. 1. 23), non laudis amor, nec gloria cessit (Aen. 5. 394), ille ducem haud timidis vadentem passibus aequat (6. 263).

For an adverb joined with a participle we have an adjective substituted, that represents the adverb transferred to the syntactical function of the omitted participle, in: equidem per litora certos dimittam (1. 576) for: certo eos inventuros; si quando adsideret, atrox ac dissentire manifestus (Ann. 2. 57. 4) for: manifeste visus; an falsa haec in maius volgaverint (3. 12. 6) for: falso credita. In: crateres auro solidi (Aen. 2. 765) for: auro solido facti, we have the ablative of specification developed through hypallage from the ablative of material; as in: ductus Neptuno sorte sacerdos (2. 201) for: factus . . . sorte ducta sacerdos, hypallage lies at the root of the change. In: degenerem Neoptolemum (2. 549) we have an adjective formed from the phrase: de genere (= de decore generis) declinantem in much the same way that the verbs are formed that we notice next.

In: Gabinium tres adhuc factiones postulant (Cic. ad Q. Fr. 3. 1. 15. 5) 290 and: Ancharius . . . Cordum proconsule Cretae postulaverat repetundis (Ann. 3. 38. 1) the use of: postulare Gabinium for: Gabinio diem postulare is a natural shortening arising from the long and familiar use of diem postulare ‘to impeach’; such phrases tend to expression by a single term, and they would in such case shift the object of their action from the dative to the accusative. But in: vox hominem sonat (Aen. 1. 328) for sono indicat, we have a verb formed from a noun dependent on the verb omitted, and giving the meaning primarily given by this noun joined with the verb now omitted.

In: ut ridentibus adrident, ita flentibus adsunt (flentes) humani voltus (A. P. 101) adrident seems short for adsunt ridentes—a union of a present participle with the verb to form the new verb. In; quibus haec . . . adridere velim, doliturus si placeant spe detenus nostra (Sat. 1. 10. 89) adridere seems to be for adridentes adire (= adire ut adrideant); and spe nostra is the old and usual expression for quam speramus, being the opposite of this construction, perhaps induced by such resolutions as: dicta dabas (Aen. 10. 600) and: discrimina dabat (10. 382). It is a present participle too that is involved in: tamquam parum ambitiose filium ducis gregali habitu circumferat (Ann. 1. 69. 5) for: circumferens obtendat militibus; but the past participle in: in flammam iugulant pecudes (Aen. 11. 199) for: iugulatas iniciunt, and in: ille autem impavidus partes cunctatur in omnes (10. 714) for: cunctatus (or cunctans) ruere meditatur.

But it is the gerundive in: cum corpora . . . prensa. . . frangeret ad saxum (3. 625) for: pelleret frangenda, and the gerund in: in magicis sacra piare focis (Prop. 1. 1. 20) for: sacra facere piando, and in: atram prorumpit ad aethera nubem (Aen. 3. 572) for: mittit prorumpendo. This seems the construction involved in: correctus Bestius (Hor. Ep. 1. 15. 37) for: corrigendo factus Bestius ‘become a Bestius in his censures’; for in Persius too: Bestius urget doctores Graios (6. 37) Bestius seems a censor whose severity outdoes the Stoics.

There has been doubt about the spelling and meaning of nequidquam ‘in vain’, which seems in its primary sense to be ‘not a whit’, ‘nothing whatever’. It is still used in its primary sense in: secernere autem a corpore animum nequidquam est aliud quam emori discere (Cic. Tusc. 1. 75. 31). Cicero in deriving nequitia, which is from nequam (utilis) ‘not useful to any extent’, says: nequitia ab eo, quod nequidquam est in tali homine; ex quo idem nihili dicitur (3. 18. 8), and it is useless to try to give any other force to nequidquam here but that plainly indicated by its composition (ne—quidquam). 291 But when we turn to its use by the poets, we feel that we are at once involved in ambiguity. Are we to render Horace’s: an male sarta gratia nequidquam coit et rescinditur? (Ep. 1. 3. 32) ‘or does the friendship ill-patched-up utterly fail to join and pull apart again’, or ‘join in vain’? Porphyrio favours the former rendering; his note is: nequidquam pro non, and I feel that this sense is better. But he refers us to: et nunc nequidquam fallis dea (Aen. 12. 634), to which Servius’s note is: nequicquam fallis pro non fallis. I must translate ‘and now thy godhead escapes me not a whit’, i.e. I am fully aware that thou art a goddess. But Mackail, following the usual sense of nequidquam translates ‘and now thy godhead conceals itself in vain’, and the passage will bear that meaning too. But the other sense seems to have been that obvious to the Roman reader.

In: quia non firmus rectum defendis, et haeres nequidquam caeno cupiens evellere plantam (Sat. 2. 7. 27), shall we translate nequidquam cupiens ‘vainly wishing’ or ‘striving in vain’ as Lonsdale and Bryce do? No doubt our inconstant friend would like to have that generally believed, but the real trouble is expressed by the primary meaning of nequidquam: he is not in reality bent on doing anything of the kind; si firmus esset the result would be very different. We have here a purposed ambiguity on Horace’s part, giving a fine effect to his verse, which, when you reject the origin I suggest for nequidquam in consonance with the testimony of the ancients, you lose entirely. Let us turn to:

Sic fatus senior, telumque imbelle sine ictu

Coniecit, rauco quod protinus aere repulsum,

Et summo clipei nequidquam umbone pependit (Aen. 2. 544-6).

Again Servius explains nequidquam as non, and refers to Persius (2. 51). He adds that some wish to give summo . . . pependit the sense, that the shaft repelled from the boss did not even stick to the surface of the shield, so as to hang there. Others thought the phrase a vain repetition in that case; but Servius does not agree with this and thinks that to: aere repulsum Virgil might well add: nec de clipeo pependit. The idea of the weapon hanging in vain from the boss seems to have occurred to none of them; how could it hang suspended from the surface of the boss (summo umbone)? The weak weapon strikes full on the boss, which sings hoarse from the blow, and is repelled; nor does it hang from the burnished brass; if it had struck the hide of which the rest of the shield was made it might have hung thus suspended, but even this poor satisfaction 292 was denied Priam. There seems little doubt that Servius is right in explaining nequidquam here as non.

But how does nequidquam get its common meaning of ‘in vain’? Let us turn to: nequidquam deus abscidit prudens Oceano dissociabili terras (Od. 1. 3. 21), where there can be no question that nequidquam means, not non, but frustra. We noticed that in Prop. 1. 1. 20 piare was used for facere piando; here abscidit is for effecit abscindendo, ‘the god has effected nothing whatever by sundering the lands from the Ocean’. Let us turn to: ne istuc nequidquam dixeris in me tam indignum dictum (Pl. Asin. 698); here dixeris is for effeceris dicendo. And so in: qui ipse sibi prodesse non quit sapiens, nequidquam sapit (Enn. F. 282, M.) for nequidquam efficit sapiendo. In: hodie sero ac nequidquam voles (Ter. Heaut. 344) sero is for: sero et re infecta voles. Different is the solution in: neque illum flava Ceres alto nequidquam special Olympo (Geo. 1. 96), where with nequidquam (ad)spectat we musl supply adfutura (= profutura). So in: cernimus adstantes nequidquam (adfuturos) lumine torvo Aetnaeos fratres (Aen. 3. 678). In: Rufe mihi frustra ac nequidquam credite amico (Catull. 77. 1) the magno cum pretio of the next verse points to an ellipsis of proficienti with nequidquam. And in: donec deceptus et exspes nequidquam fundo suspiret nummus in imo (Pers. 2. 51), where Servius tells us that nequidquam is for non, the nequidquam expressed is probably for non, but the sentence is elliptical and we must supply nequidquam wilh suspirando in its later meaning of ‘in vain’, giving us: donec deceptus et exspes (nequidquam suspirando) nequidquam fundo suspiret nummus in imo ‘till baffled and hopeless from sighing in vain the penny at the bottom of his chest ceases from its sighing’; where we have a good example of the expression of four terms by two.



The ambiguity of which we have such a fine example in Persius’s use of nequidquam is often associated in Latin poetry with the amphibole, and is sometimes evolved from that figure. Murray defines the amphibole as an ambiguity, but in the Roman grammarians the term has a very special application, as the use of a single word in two different relations in the same clause or sentence. We read in Horace: male verum examinat omnis corruptus iudex (Sat. 2. 2. 8), of which Acron tells us: male ἀμφίβολον: aut male verum, i.e. non verum, . . . aut male examinat, i.e. non potest de sapientiae bono nequitia iudicare. It seems to me that here, as in several cases, it might be made threefold, and extended to corruptus as well, giving us: ‘the depraved verdict of the badly warped juror strays far from the truth’. Acron refers us to: cum sic unanimam adloquitur male sana sororem (Aen. 4. 8), where Servius tells us that male may be for minus or for perniciose. In this verse we may take male sana as minus sana, and male adloquitur as perniciose adloquitur, and translate: ‘when the queen far from sane with consequences disastrous to herself addresses her loving sister’.

That this figure is primarily a special form of the reduction of a fourfold union to three terms seems plain from: cedamus Phoebo et moniti meliora sequamur (Aen. 3. 188) ‘let us yield to Apollo, and follow the better way he counsels’ (Mackail); we have here clearly a shortening from: moniti meliora meliora sequamur, as Servius clearly sees. So in: volans alte raptum cum fulva draconem fert aquila (11. 751) for: alte volans alte raptum, etc., and in: lustrabat studio recolens (6. 681). In this figure the word, though in its various relations it may change its case, gender, or number, retains exactly the same form; when the figure passes to verbs, we call it, not amphibole, but zeugma.

As I began by citing examples of this figure with the adverb, I add a few more: solve senescentem mature sanus equum (Ep. 1. 1. 8) ‘be wise in time and loose betimes, &c.’, rite repertum carpe manu (Aen. 6. 145). In:

Has equidem memorare tibi atque ostendere coram

Iampridem, hanc prolem cupio enumerare meorum (6. 716-7),


the position of iampridem between the clauses to which it is related is parallel to that of alte in: alte volans raptum. In: bis patet in praeceps tantum tenditque sub umbras (6. 578) bis tanlum is so distributed between the two clauses to which it is to be related, that its terms constitute the first words of each, and of each colon of the verse. In: torquet agens circum (1. 117) circum is not yet a preposition; but in: laurus erat tecti medio in penetralibus altis (7. 59) the preposition in is to be taken with medio as well as with penetralibus. In: emicat, adrectisque fremit cervicibus alte luxurians (11. 496) the connexion between the first and last words of the verse is so usual with Virgil and Horace, that we must join emicat with alte, giving it a threefold relation here, with emicat, fremit, and luxurians.

For pronouns I note: Annam cara mini nutrix huc siste sororem (4. 634), where mihi is to be joined with huc siste as well as with cara. In: haec ubi nos praecepta iubent deponere dona (6. 632) haec, the first word in the verse, must be joined in its usual sense with dona the last, but in the sense of ‘our’ with praecepta the last word of the first colon. In: o genetrix, quo fata vocas, aut quid petis istis (9. 94), to which Servius notes: istis utrum precibus an navibus?, we must join istis (= istis precibus) as an ablative with vocas, and as a dative (= istis navibus) with petis. In:

Quid struat his coeptis, quem, si Fortuna sequatur,

Eventum pugnae cupiat, manifestius ipsi

Quam Turno regi, aut regi apparere Latino (8. 15-17),

we have one of Virgil’s hidden meanings. It is obvious and natural to refer ipsi to Aeneas. It may seem too obvious on reflexion; but, as Sidgwick points out, it may be a skilful irony for: nos minime novimus ‘he knows, we don’t’. But Servius’s note is: quem finem suae velit esse victoriae, ipsum melius nosse, qui iam antiquus est hostis. This reference of ipsi to Diomede, and not to Aeneas, will seem preferable, when we consider how Rome subjected Magna Graecia, and remember the historical purpose of Virgil in the Aeneid.

In: quia non sentis quod clamas rectius esse (Sat. 2. 7. 25) rectius esse goes with sentis as well as with clamas; and in: casus . . . insontis amici (Aen. 5. 350) casus has the double meaning of ‘case’ and ‘fall’, and in either sense insontis seems to belong to it as well as to amici. In: fecerat et viridi fetam Mavortis in antro procubuisse lupam (8. 630), as Servius notes, Mavortis, the first word of the second colon, is to be joined with its last word lupam as well as with: in antro. In: rerumque ignarus imagine gaudet (8. 730) the connexion of 295 rerum with both ignarus and imagine is very plain; Aeneas does not yet know the achievements of Rome, but he is delighted with their portrayal on his shield. In: illam Terra parens, ira irritata deorum, extremam, etc. (Aen. 4. 178-9) deorum is dependent on extremam as well as on ira, as is clear from the arrangement which puts illam at the beginning and deorum at the end of the verse in juxtaposition with extremam. Servius, moreover, sees an amphibole in the relation of deorum to ira; for this genitive may be subjective or objective, signifying either the wrath of the Gods at the Giants, Earth’s offspring, or the wrath of Earth at the Gods for destroying them.

In: Herculeis sopitas ignibus aras excitat (8. 542) Servius notes the hypallage in: Herculeis ignibus for Herculeas aras; this hypallage seems due to the double use of ignibus, as an ablative of means with excitat, and as an ablative of respect with sopitas; in the last connexion the transference of Herculeas would be easy. In: bellis hoc victor abibat omnibus (10. 859) bellis is to be joined with victor as an ablative of manner, and with abibat as the proper ablative. In: exceptus tergo consueta locavit membra (10. 867) we have a triple amphibole in: tergo, which is in threefold relation with consueta, exceptus, and membra locavit; in the first it is a dative proper, but in the other two it seems the Virgilian dative for: in tergum. In: celsam oppugnat qui molibus urbem (5. 439) Servius sees the amphibole: aut celsam molibus, aut quae molibus oppugnatur, is his note. As: celsam molibus is a poetic variety for: celsis molibus, we should rather take as the alternative: aut quae celsis molibus oppugnatur. We have a threefold amphibole in:

Namque tibi reduces socios classemque relatam

Nuntio et in tutum versis Aquilonibus actam (1. 390-1),

where: in tutum must be understood with reduces and relatam. In: petit ante alios pulcerrimus omnes Turnus (7. 55) ante alios omnes must be joined with both petit and pulcerrimus. So in: dum pluit in terris, ut possint sole reducto (in terris) exercere diem (10. 807) ‘put in a busy day on their farms’; and in: obferebatque familiam (in tormenta) reus et ministros (Germanici) in tormenta flagitabat (Ann. 3. 14. 3).


rex ipse Latinus

Ni dare coniugium et dicto parere fatetur,

Sentiat et tandem Turnum experiatur in armis (Aen. 7. 432-4),

we have an amphibole in: in armis, which must be joined with 296 (expertus) sentiat as well as with experiatur, and a curious ambiguity in: dicto parere. At first sight dicto seems to be Turnus’s command, but the oxymoron implied in making the king obey leads us to prefer as Virgil’s hidden meaning the older sense of parere ‘to be on hand’; for parere is the neuter of parare and related to it as is placere to placare. In that case dicto is Latinus’s promise to Turnus, and: dicto parere is equivalent to: promisso stare.

In: corripiunt (spiris ingentibus) spirisque ligant ingentibus (2. 217) and: quid aeternis (consiliis) minorem consiliis (aeternis) animum fatigas? (Od. 2. 11. 11) we have amphiboles consisting each of a noun with its adjective in synchysis. To: quam (vallem) densis frondibus atrum urget utrimque latus (Aen. 11. 523) Servius’s note is: utrum densis frondibus urget, an densis frondibus atrum? We have the amphibole distributed in: crescit occulto velut arbor aevo fama Marcelli (Od. 1. 12. 45), the full form being: crescit occulto (aevo) velut arbor (occulto) aevo fama Marcelli; and in: tutela vel ipsis certior arcebat muris (Sil. 12. 64), short for: tutela vel ipsis (muris) certior arcebat (ipsis) muris (Hannibalem). So in: utrumque sacro digna silentio mirantur umbrae dicere (Od. 2. 13. 29), short for: utrumque sacro (silentio) digna silentio (sacro) mirantur, etc., and: accipe si vis, accipiam tabulas (Sat. 1. 4. 14), short for: accipe si vis (tabulas), accipiam (si vis) tabulas; where the first si vis is our ‘please’, but the second is ‘if you will’.

In: navis quae . . . debes Virgilium finibus Atticis reddas incolumem (Od. 1. 3. 6) the position of Virgilium finibus Atticis indicates its connexion with both debes and reddas. Very difficult is:

Quodque vehunt prorae Centaurica saxa minantes,

Tigna cava et pictos experiere metus (Prop. 4. 6. 49-50).

Virgil described it as follows:

Tanta mole viri turritis puppibus instant (Aen. 8. 693),

to which Servius’s note is: nam Agrippa primus hoc genus turrium invenit, ut de tabulatis subito erigerentur, simul ac ventum esset in proelium, turres hostibus improvisae, in navigando essent occultae. We shall write Virgil’s verse in full thus: viri tanta mole (Centaurica) turritis puppibus (minantes tanta mole saxorum) instant, and Propertius’s: quodque vehunt prorae (inmanes species Centaurorum) Centaurica saxa minantes, etc. It is characteristic of Virgil’s style to give us the unexpected in designating such shapes as viri, and not Centauri.


Aenean credam quid enim fallacibus auris

Et caeli totiens deceptus fraude sereni? (5. 850-1),


Sidgwick’s note is: ‘fallacibus auris must be dative after credam, and not ablative with deceptus; otherwise credam has no dative, and quid enim is awkwardly lost in the sentence. That being so, what is et?’ The text is not certain; Servius read caelo, with some of the best codices, and Ribbeck has adopted this reading. If we follow him we shall have no difficulty with et, which joins the datives auris and caelo. But caeli agrees well with sereni, the beginning with the end of the verse; if we read caelo, sereni will be short for sereni caeli. Acron read caeli (ad Od. 1. 5. 6), and Servius adds: alii legunt ‘deceptus fraude caeli sereni’. It seems the more difficult reading, and is supported by most of the manuscripts. The difficulty is that stated by Sidgwick: how can auris be both dative with credam and ablative with deceptus?

We have already seen how in Aen. 9. 94 istis seems as well for istis precibus with vocas, an ablative, and istis navibus with petis, a dative. In: certum voto pete finem (Ep. 1. 2. 56) ‘in your prayer seek a definite limit for your petitions’, i.e. ‘pray to the gods for the contented mind’, we have a similar amphibole. So in: auxilio laetos dimittam (Aen. 8. 171) ‘you rejoicing in my aid I will send away to the Etruscans, who are to be your allies’, auxilio must be joined with laetos in the ablative, and with dimittam in the dative. Parallel to this is: poscor Olympo (8. 533), to which Servius’s note is: aut de Olympo poscor, aut certe in Olympum poscor. To: neque finitimo Mezentius umquam huic capiti insultans tot ferro saeva dedisset funera (8. 569-71) Servius’s note is: aut in finitimo, aut finitimo capiti. In: extulit os sacrum caelo tenebrasque resolvit (8. 591) Sidgwick prefers to take caelo as the ablative with extulit as ‘making more natural sense’, but its position in the verse plainly shows that it is to be joined with both extulit and resolvit, being with resolvit the dative of advantage, and with extulit the Virgilian dative for: in caelum. In: his informatum manibus iam parte polita fulmen erat (8. 426) Servius explains manibus as: in manibus; Sidgwick translates ‘shaped by these hands’, making it depend on informatum, where it is probably a dative. The order of the words favours Sidgwick, but it is quite clear that the thunderbolt is still in Vulcan’s ‘hands’, and has been only partly polished by them. In:

quo se multis cum milibus heros

Consessu medium tulit exstructoque resedit (5. 289-90),

Servius arranges thus: quo se Aeneas medium tulit cum multis milibus et exstructo consessu resedit. But the position of consessu shows 298 that it must be constructed with (in)tulit as well as with resedit, being an ablative with resedit and a dative with intulit. The phrase: quo consessu seems parallel in syntax to: unde domo (Ep. 1. 7. 53), and to: huc viciniam (Ter. And. 70), where, however, the best manuscript has: huic viciniae. In: paulatim adnabam terrae iam tuta tenebam (Aen. 6. 358) terrae seems dative with adnabam, but genitive with tuta.

In: ne turbata volent rapidis ludibria ventis (6. 75) we have rapidis ventis in a threefold relation, to turbata as an ablative of cause, to volent as an instrumental ablative of the way along which, and to ludibria as a dative. So in: da, non indebita posco, regna meis fatis Latio considere Teucros (6. 66-7) meis fatis is a dative with da and indebita, and an ablative with considere for: ex meis fatis. The pregnancy of meaning involved in such constructions is plainly part of the poet’s aim, and in sermo little attention is paid to grammatical categories like the dative and ablative; indeed it may be questioned whether the term: ablativus had come into use in Virgil’s day, though the tradition is that it was devised by Caesar. But the important thing in relation to amphibole is evidently identity of form; and the dative and ablative are identical in form always in the plural, and very often in the singular.

Not that this confusion in grammatical categories is confined to these two cases. We have a like confusion between the feminine singular and the neuter plural in: non adversata petenti adnuit (4. 127), where we may take adversata as the neuter plural, the object of adnuit, and as the feminine singular in agreement with Venus, its subject. This is exactly parallel to the amphibole which Servius notices in: classica iamque sonat (7. 637); his note is: bene posuit amphiboliam; nam classicum dicimus et tubam ipsam et sonum. He clearly means that here too classica is both subject and object; and we must translate: ‘the trumpets sound the war-notes’. When we remember the relation of the feminine singular to the neuter plural, how at this very time in popular speech most neuter plurals were becoming feminine singulars, a process we see in: interea servitia repudiabat, cuius initio ad eum magnae copiae concurrebant (Sall. Cat. 56. 5), we shall not be surprised to find in Virgil’s verse unequivocal signs of this tendency.

A note of Servius to: sic te ut posita crudelis abessem? (Aen. 4. 681) expresses doubt whether crudelis refers to Anna or Dido, i.e. whether it is nominative or vocative. To: nomine Dido saepe vocaturum 299 (4. 383) his note is: Dido potest et vocativus esse et accusativus. In: ille ictum venientem a vertice velox praevidit (5. 444) he sees that velox may agree with ille, or may be the accusative of cognate notion with venientem, giving the sense of velociter. In: sed magnum metuens se post cratera tegebat (9. 346) it seems well to take magnum with cratera as well as with metuens. Virgil’s Latin is still a living language, and not intended merely as a mental gymnastic virginibus puerisque; and the Roman grammarians recognize in him this tendency to ambiguity which our scholars are at such pains to ignore. In: meritumque malis advertite numen (4. 611) our scholars take malis as neuter; Servius’s note to meritum is: quod mali merentur, clearly indicating an amphibole in: malis. To: Acrisioneis Danae fundasse colonis (7. 410) Servius’s note is that Acrisioneis here is the patronymic of Danae, and not to be joined with colonis; the position of these words at the beginning and end of the verse is clear proof to me that his negative is wrong, just as the juxtaposition of Acrisioneis and Danae prove him right in his positive assertion.

In: ille agmine longo . . . serpens (5. 90-1) with ille we may supply anguis from v. 84; but it is far more consonant with Virgil’s style to take serpens both as substantive and as verb. In: se . . . aufert, linquens multa metu cunctantem et multa parantem dicere. Suscipiunt famulae collapsaque membra . . . referunt (4. 389-92), to which Servius’s note reads: multi pro ‘relinquens Aeneam’, alii pro ‘deficiens’ accipi volunt more antiquo, sicut ‘delinquere’ pro ‘deficere’, it is plain that from linquens we must supply linquentis with suscipiunt membra; the linquens expressed is for relinquens, but the linquentis to be supplied is for delinquentis ‘fainting’. To: adsit laetitiae Bacchus dator (1. 734) Servius notes: bene autem addidit dator laetitiae, quia est et dator furoris. But the obvious construction of laetitiae is not that of genitive with dator, but of dative with adsit. To: quantulum summae curtabit quisque dierum? (Sat. 2. 3. 124) Acron notes: de summa curtabit, from which it is clear that summae is an amphibole, being for the genitive dependent on quantulum, and the dative dependent on curtabit; quantulum summae curtabit is short for: quantulum summae summae decurtabit. But in: stridor ferri tractaeque catenae (Aen. 6. 558), where Servius notes that catenae may be a genitive singular or a nominative plural, there is probably no shortening, and the amphibole is in the ambiguity. In: et pater ipse suo superum iam signat honore (6. 780) Servius makes superum the accusative singular with signat; in reality it 300 has a threefold relation, being also the genitive plural with pater and with honore, as their position at the beginning and end of the verse clearly indicates. So in: arserit Euhadne flammis iniecta mariti (Mart. 4. 75. 5) flammis is to be joined with arserit as well as with iniecta mariti.

But the variety in meaning marking the amphibole may depend, not on a confusion of inflexions, or even of parts of speech. In: his informatum manibus iam parte polita fulmen erat (8. 426) one of the meanings of manus was that we have in: nos aera, manus, navalia demus (11. 329), a sense exactly paralleled in English. Such uses involve a lack of propriety of meaning, that is a distinctive mark of Latin poetry as opposed to Latin prose. In: quare age et armari pubem portisque moveri laetus in arma para (7. 429-30) Servius notes that laetus cannot be properly used here of Turnus, who has just been deprived of his promised bride; it is, he thinks, for alacer; for: alacer quamvis tristis, perhaps we might better phrase it. In a fourfold union not reduced to an amphibole we have a like ambiguity in: quam pro me (i.e. pro vita mea) curam geris, hanc precor, optime, pro me (i.e. pro mea fama) deponas (12. 48-9).


ea vox audita laborum

Prima tulit finem, primamque loquentis ab ore

Eripuit pater ac stupefactus numine pressit (7. 117-19),

at first sight we seem to have a repetition of prima and primam in a slightly altered sense; ‘not till then’ and ‘at the first’ Sidgwick proposes. But a glance at v. 116 shows that: mensas consumimus, the vox in question, was the last, not the first, that fell from the lips of Julus; but that his father was the first to catch it up and stay him from further speech; so that we should have: ultimamque loquentis ab ore eripuit pater primus, and primam here is an amphibole for ultimam primus. To go back to: patulis nec parcere quadris (v. 115), the bread bitten into is described as orbis in v. 114, but is here called quadra to elaborate the jest in: mensas consumimus; for quadra ‘the square’ is a word used for a table as well as for a loaf, as we see in: ut bona summa putes aliena vivere quadra (Juv. 5. 2). The prayer Aeneas offers for the consummation of the omen is followed by this sign from highest Jove:

Hic pater omnipotens ter caelo clarus ab alto

Intonuit, radiisque ardentem lucis et auro

Ipse manu quatiens ostendit ab aethere nubem (vv. 141-3).


Here clarus is both of the sound of the thunder-peal and of the bright glow of the cloud from which the lightning flashes. So in: ne populum extrema totiens exoret harena (Ep. 1. 1. 6) most scholars translate extrema harena ‘from the arena’s edge’, but Lucian Müller calls attention to the fact that it was customary to ask for the missio at the close of the performance. Evidently we have an amphibole in extrema, which Müller does not suspect however. So in: dicta tibi est lex (Ep. 2. 2. 18), where some translate ‘you were told the conditions of sale’, some ‘you have your law’, we have an amphibole in lex.

So too in: quam sedem somnia volgo vana tenere ferunt (Aen. 6. 283) Servius’s note is: volgo temere, passim, catervatim; aut volgo ferunt. Volgo he connects with tenere in the sense of ‘in swarms’, and with ferunt in the sense: ‘the common report is’. In: nec dis nec viribus aequis (5. 809), while the meaning of viribus aequis is plain, Servius explains: nec dis aequis as: dis iniquis, i.e. adversis. So in: quin et supremo cum lumine vita reliquit (6. 735), which Mackail translates: ‘when the last ray of life has gone’, we may recall the use of lumine in: vix lumine quarto (6. 356), and translate: ‘when on their last day life has departed with its latest ray’. In: nec dextrae erranti deus afuit (7. 498) ‘a present deity suffered not his hand to stray’ (Mackail), while we connect neque with erranti as well as with afuit—a very obvious amphibole—we should also regard: dextrae erranti as short for: dextrae iuvenis errantis, and recall: ille . . . errabat silvis (v. 491); when we must translate: ‘nor did the fury, on hand for mischief, allow the shaft of the roving youth to miss its mark’. To: saevit medio in certamine Mavors caelatus ferro (8. 701) Servius’s note is: caelatus ferro, aut in armis locatus, aut de ferro sculptus; either ‘clad in mail’ or ‘chased in steel’. In: supposta furto Pasiphae (6. 24) Servius explains supposta furto as: furtim inclusa in vaccam ligneam, where furto is for dolo; but he also explains furto as for adulterio, comparing: Martis dolos et dulcia furta (Geo. 4. 346), when furto is no longer an ablative of manner but a dative of goal with supposta. In:

A quo post Itali fluvium cognomine Thybrim

Diximus; amisit verum vetus Albula nomen (Aen. 8. 331-2),

we are at first inclined to take cognomine in the general sense it has in: clari cognominis Albam (8. 48), but we presently find that the cognomen Thybris is opposed to the verum nomen Albula. Cognomen really means nickname (= eke name), and we get its force here 302 best if we supply this: quo cognomine adsumpto amisit verum nomen. In:

Talibus Aeneadae donis dictisque Latini

Sublimes in equis redeunt (7. 284-5),

it is clear that sublimes is to be joined with: donis dictisque ‘gifts and promises’ as well as with: in equis, and is really short for elati et sublimes. So in:

At pius Aeneas ingenti mole sepulcrum

Imponit, suaque arma viro remumque tubamque (6. 232-3),

imponit is short for: imponit, ubi insculpsit; and we have passed to zeugma.

In: si quis bella tibi terra pugnata marique dicat (Ep. 1. 16. 25) tibi has been removed from dicat and given a place next to bella pugnata marique to afford a shadow of excuse for the mistake Horace assumes the hearer might make in imagining that he, and not Augustus, is the hero referred to in tibi. In: quattuor a stabulis praestanti corpore tauros avertit (Aen. 8. 208) in avertit at first the reader will see only the usual meaning of ‘drives off’, the meaning Catullus transfers in sport from the ram to the fleece in: auratam optantes Colchis avertere pellem (64. 5); but as he reads on: atque hos . . . cauda in speluncam tractos versisque viarum indiciis raptos (vv. 210-11), he sees that by this use of avertit Virgil was paving the way to present his reader with this picture of Cacus dragging the cattle by their tails into his den. In: alte consternunt terram concusso stipite frondes (4. 443) the manuscripts have altae, but both Macrobius and Servius read alte, and Servius explains it as: iugiter, diu; aut ex alto cadentes. We may adopt the reading of the scholiast, while we remember how often ae and e are confused in manuscripts, and take care never to read altae as if it were written altai, a thing Virgil was plainly never guilty of. Evidently there is an amphibole in alte, which we shall understand if we restore: altae consternunt terram concusso stipite alte concussae frondes ‘thick strew the ground the leaves shaken from on high from the shaken trunk’. More difficult is:

Ergo iter inceptum celerant rumore secundo.

Labitur uncta vadis abies (8. 90-1).

In the last phrase we have an amphibole in uncta, which when connected with abies has the meaning of picta, but when joined with labitur, that of celeriter. But the real difficulty is in the reading: Rumone secundo, which Servius does not adopt in his text because of its difficulty, though he attempts to explain it. Many of our modern 303 scholars explain: rumore secundo, as though it were like: clamore secundo (10. 266) but we can readily see how far from probable this is, when we consider that they are rowing up an unknown river through a land swarming with hostile Latins. Some understand it as for: cursu secundo, rumore being the subdued murmur of the current round the keel; but I have no parallel to this use. Servius does not venture even so far as this; he explains rumore secundo as bona fama, since they avoided depredations on the banks. Did they really?—we seem to recall a story of a sus alba cum fetu, told just before this. We have the reading: Rumone secundo not merely in Servius, but in the Codex Meliceus and the Romanus, where the n has been corrected to r by a later hand. Rumon, the oldest name of the Tiber, Virgil gives us nowhere else; and from his usual practice it seems likely that he would find a place in his poem for this name. But the course of the boat is not down the Tiber, flumine secundo, but up the Tiber, flumine adverso. This objection is obviated by the meaning of celerant, which is here, for celerare videntur, an ambiguity. Servius meets the objection thus: descendentis celeritate prope conscendit; and he cites Statius about the discus: ille citus sublime petit, similisque cadenti crescit in adversum (Theb. 6. 682-3) ‘swift it makes for the sky, and with all the speed of a falling object it progresses in the opposite direction’. He means that such is the speed of the Tiber’s downward sweep, that the rowers feel, as they breast the current, that they are speeding rapidly on—a play on appearance opposed to reality quite in Virgil’s style. We may compare the arrow which: celeres incognita transilit umbras (Aen. 12. 859); it is not the shades that are swift, but the shaft that bounds unperceived across them.

We have this amphibole extended to threefold use in: hic gravis Entellum dictis castigat Acestes (5. 387), where when joined with Acestes gravis means ‘stately’, with dictis ‘lofty’ or ‘dignified’, and with castigat ‘severe’. So in:

I nunc, ingratis offer te, inrise, periclis;

Tyrrhenas i sterne acies, tege pace Latinos (7. 425-6),

ingratis with offer te is ‘against thy inclination’, with periclis ‘that bring thee no thanks’, and with tege pace ‘without gratitude from them’.



The ascendancy of Virgil among the ancients and in the Middle Ages seems to have been due, not merely to his poetical genius, but to his supremacy in the learning of his day. He was not merely a great epic poet; he was universally believed to be a master in grammar, oratory, philosophy, theology, mythology, history, astronomy, magic, to such a degree that in the Middle Ages he was held to be the Master Magician. Throughout the Renaissance and the consequent advance of learning and literature in modern Europe he continued to hold a supreme place: Dante, Petrarch, Bossuet, Addison, Burke were among his devoted students; and it was only with the revival of classical studies in Germany in the nineteenth century that his claim to supremacy even among Latin poets began to be questioned. The obvious cause of this revolt, I read in Professor Sellar’s Virgil (p. 74), ‘is the great advance made in Greek scholarship during the present century’. Perhaps, if this had been associated with a like advance in Latin scholarship, Virgil’s reputation might have fared better. Of course we do not expect to trace in his poems the discoveries of Newton, or Dalton, or Tyndall; though the way in which Lucretius’s great poem has been illustrated by some of these is striking enough. But we do expect to find him abreast with the science which an average man of his day or ours can acquire by the exercise of his senses unaided by further instruments of precision, and by an average intellect. Now Virgil, I have been told, believed and recorded in his great epic that the moon rises in the west at the beginning of the month. For we read:

quam Troïus heros,

Ut primum iuxta stetit, agnovitque per umbram

Obscuram, qualem primo qui surgere mense

Aut videt aut vidisse putat per nubila lunam,

Demisit lacrimas (Aen. 6. 451-5).

There you have it:

primo qui surgere mense

Aut videt aut vidisse putat per nubila lunam


But why: aut videt aut vidisse putat surgere? The expression seems to indicate two classes of observers; and we seem to have here an example of the threefold union, which we have so often been able to trace back to an older fourfold one. Does surgere mean the same with videt as with vidisse putat? Let us turn back to the nearest use of that verb; in: per spes surgentis Iuli (v. 364) surely the idea is not: ex solio surgentis, but rather: in adulescentiam crescentis. And what is there so novel or unheard of in associating crescens with luna? Why then not take surgere in its sense of crescere with videt? and in its usual sense with vidisse putat? He who casts a momentary and careless glance westward on the evening the new moon appears, and again a couple of evenings later, thinks he sees the new moon rising in the west; she is now much higher than when he saw her primo mense. But with him who really sees the case is different; and we have plainly a case here, where we can restore the old fourfold union: aut videt lunam crescere, aut vidisse putat lunam surgere. Surgere is the second of the pair, the term usually retained in the shortened form; and Virgil likes to surprise his reader. But in this threefold union, resulting from the expression of crescere and surgere by surgere, we have a fine example of the figure we call zeugma, an example which helps us to understand its development.

The examination of a second example, which has not been recognized as zeugma, and does not quite agree with the definition of the figure usually given, may be of interest, especially as it seems to solve a puzzling passage in the Annals. We read: tum consultatum de honoribus, ex quis maxime insignes visi, ut porta triumphali duceretur funus, Gallus Asinius; ut legum latarum tituli, victarum ab eo gentium vocabula anteferrentur, L. Arruntius censuere (1. 8. 4). Scholars have feared to trust the manuscripts here; Nipperdey wanted to omit visi, and Bezzenberger to supply qui after quis, not seeing that both quis and visi naturally refer back to honores, and not forward to Asinius and Arruntius. After visi Tacitus gives the two decrees that he thought most important about Augustus’s funereal honours in a form that seems but slightly varied from that in which they would appear in the Acta Senatus. There they would probably be recorded: ut porta triumphali funus ducatur: Q. Asinius Gallus censuit. Ut legum latarum ab imperatore tituli, gentium victarum nomina anteferantur: L. Arruntius censuit. Tacitus has made some verbal changes and alterations more suo, but his main change in syntax was to bring the two 306 censuits together and make of them a single censuerunt. He had already composed a sentence closely parallel in a most important part of his work: initium mihi operis Servius Galba iterum Titus Vinius consules erunt (Hist. 1. 1. 1). Of course this is far from Cicero’s style; but nothing in the Annals is in Cicero’s style. It is clear that in the use of censuere here we have a second, and in some respects a simpler, form of zeugma, a form depending on variation, not in the word itself, but in its inflexions.

Let us begin with this form. Zeugma (τὸ ζεῦγμα) is primarily the same as the Latin iugum ‘the yoke’; and Liddell and Scott define it for grammar as the figure of speech wherein two subjects are used jointly with the same predicate, which strictly belongs only to one of them. The last clause does not fit some of the most striking zeugmas, as e.g. that I have just given; and we shall see that the structure of the figure is often threefold: i.e. three subjects are joined with a common predicate, which, while fitting all three in a measure, does not fit any one exactly. Ruddiman (II. p. 362 ff. cur. Stallbaum) deals with zeugma at some length, and defines it: cum adiectivum vel verbum diversis nominibus substantivis adiunctum cum propiore expresse convenit, alteri, sed mutatis accidentibus, intelligendum. To the: cum propiore he notes many exceptions, without finding a reason for them, and he is content to confine zeugma to this simplest and least interesting form of it, where the change is merely of the accidentia or inflexions.

He gives us examples of zeugma for gender in: et genus et virtus nisi cum re vilior alga est (Sat. 2. 5. 8), eis otium, divitiae, optandae aliis, oneri miseriaeque fuere (Sall. Cat. 10. 2); for number in: hic illius arma, hic currus fuit (Aen. 1. 16), mediocres poetas nemo novit, bonos pauci (Tac. Or. 10. 1); for person in: paene ille timore, ego risu corrui (Cic. Q. Fr. 2. 9 (8). 2), cum hoc tempore nihilo magis ego, quam vos, subsidio Domitio ire possim (Att. 8. 12A. 3). But we have it also for tense in: hos tibi dant calamos . . . Ascraeo quos ante seni (Buc. 6. 70), for mood in: nisi facient, quae illos aequomst (Ter. Ad. 454). It is not unusual in cases of nouns used as adjuncts of the predicate, as in: conveniunt . . . flumina . . . nescia gratentur (parenti) consolenturne parentem (Ov. Met. 1. 578), imitari (bonos) quam invidere bonis malebant (Sall. Cat. 51. 38), cum ego unaquaque de re dicam, et (unamquamque rem) diluam (Cic. Cluent. 6. 2), quem neque pudet quidquam nec metuit quemquam (Ter. Ad. 84).

The old grammarians classified zeugma according to its position as 307 (1) Protozeugma, as in: tutatur favor Euryalum lacrimaeque decorae (Aen. 5. 343); (2) Mesozeugma, as in: caper tibi salvus et haedi (Buc. 7. 9); (3) Hypozeugma, as in: quamvis ille niger, quamvis tu candidus esses (Buc. 2. 16). Of these the Mesozeugma is the most interesting form; for in: caper tibi salvus et haedi, we trace at once the fourfold form in its most common order, though there is no difficulty in reaching: tutatur favor Euryalum, tutanturque lacrimae decorae, or: quamvis ille niger esset, quamvis tu candidus esses.

While in most cases the zeugma is in agreement with the nearest subject, as we read in his definition of it, Ruddiman sees that in some examples this is not so. In these examples if, following the order of the subjects, we supply the missing predicate, we shall see that in the exceptions it is the second term of the pair that is omitted, while in the regular forms it is the first. This agrees with the treatment by Horace and Virgil of Castor et Pollux and like pairs; and we have seen that in pairs of common names, verbs, adjectives, and adverbs it is usually the second that is expressed and the first that is omitted. So to take Ruddiman’s examples of the exceptions: sic imperium adeptus populum Romanum, vel dicam humanum genus, voti compotem (compos) fecit (Suet. Cal. 13. 1), cum natura loci, tum dolo, ipsi atque signa militaria obscurati (obscurata) (Sall. Jug. 49. 5), ego populusque Romanus populis Priscorum Latinorum bellum indico (indicat) facio (facit)que (Liv. 1. 32. 13). But regular are: quis igitur illum consulem nisi latrones (putat) putant? (Cic. Phil. 4. 9. 4), nil hic nisi carmina (deest) desunt (Buc. 8. 67), sin opportunior fugae collis quam campi (fuerat) fuerant (Sall. Jug. 50. 6), talem nisi tu nulla (pareres) pareret filium (Ten Heaut. 1022). The number of exceptions is no doubt increased by the tendency in Latin to make the subject of the active verb a living sentient person; and so to prefer the first person to the second, as in: ego et rex meus, or: melius ego istud quam vos fecissem.

But in many cases we have a marked zeugma in the meaning of the verb when joined with its different objects, though the verb can be used with either of them in varying senses, nor is there any need of supplying another verb. To begin with a striking example: donec ira et dies permansit (Ann. 1. 68. 6) ‘while their anger was still hot and the daylight lasted’; or: angit inhaerens elisos oculos et siccum sanguine guttur (Aen. 8. 260) ‘holding him fast he squeezes his throat till it is drained of blood, and makes his eyes start out with his strangling grip’. So in: Pergama cum peteret inconcessosque hymenaeos 308 (1. 651); while the meaning of petere varies widely in its union with Pergama and hymenaeos, its use with each is natural and usual. So in: ubi . . . finem portusque tenebunt Ausonios (9. 98), non opibus mentes hominum curaeque levantur (Tib. 3. 3. 21), pariterque animaque rotisque expulit (Ov. Met. 2. 312), aedificiorum et hominum strages (Ann. 1. 76. 1), cibos et hortamina pugnantibus gestant (Germ. 7. 4), misceri cuncta tenebris et armis (Hist. 4. 29. 2). In: deus ipse faces animumque ministrat (Aen. 5. 640), where Ribbeck reads animam with the Codex Romanus, animum, the more spirited reading and that of the rest of the manuscripts, seems justified by Horace’s: vinum, quod verba ministret (Ep. 1. 15. 20). Sidgwick sees a violent zeugma in: hi Fescenninas acies . . . hi Soractis habent arces (Aen. 7. 695-6) ‘these form the lines . . . these hold the battlements’. But if Virgil uses acies in its primary meaning—its root is that of acer, and it meant ‘edge’ or ‘topmost ridge’—there is not here even the simplest form of zeugma, and Virgil is laughing at his editors.

More highly figured seems the zeugma in: sola domum et tantas servabat filia sedes (7. 52) ‘a daughter was the sole stay of the household and prop of so great a line’. In: hinc saevitiam centurionum et vacationes munerum redimi (Ann. 1. 17. 6) there is a marked opposition between buying what is pleasant and buying off cruelty. In: iam galeam Pallas et aegida currusque et rabiem parat (Od. 1. 15. 12) we have a fourfold zeugma: Pallas fits on her helmet, buckles on her shield, harnesses her chariot, and whets her wrath. But only in the last verb do we feel that the prose must be different, probably incendit. So in: iam parce sepulto, parce pias scelerare manus (Aen. 3. 41-2) we have a close approach to a formal zeugma: noli is the proper prose word for which parce in the second clause is in poetic use. Hendiadys seems involved in: fugam Dido sociosque parabat (1. 360) for: socios Dido parabat fugitura, in: castra . . . aciemque movebat (11. 446) for: ex castris suos in aciem ducebat, in: quo (voltu) caelum tempestatesque serenat (1. 255) for: caelum tempestatibus turbatum serenat. Servius tells us there is a hysteron proteron in: moresque viris et moenia ponet (1. 264), where I am led to believe from Propertius’s use of: ponere iura (3. 9. 24) that ponere mores too was in use in colloquial Latin.

While in the examples just quoted there is a marked zeugma or yoke of meanings combined in a single verb, in most the verb is capable of expressing each of them with propriety, and we feel no need of calling up in our minds another verb to represent fully and 309 exactly the idea intended. But in the examples that follow we do feel such a need, and the careful reader usually does supply a second verb. The simplest form of zeugma here seems that presented in: Irim demisit Olympo quae luctantem animam (solveret) nexosque resolveret artus (Aen. 4. 695); for in poetry the prefix is often omitted where it is needed, and at times is expressed where it is superfluous. We have further examples of this kind in: Iliacas igitur classes (sequar) atque ultima Teucrum iussa (ex)sequar? (4. 538), protinus et graves iras (condonabo) et invisum nepotem . . . Marti redonabo (Od. 3. 3. 33), dum terras (incolunt) hominumque (ex)colunt genus (Ep. 2. 1. 7), et consulti patres integrum id negotium ad imperatorem (referendum) distulerant (Ann. 3. 52. 3). I have already spoken of: haud igitur quidquam procedere posset, principium quoniam (pro)cedendi (cedendo) nulla daret res (Lucr. 1. 338-9), where I will ask my reader to compare vv. 372, 378, 379, and especially v. 374, which may well be short for: quo possint cedentes (procedentibus) confluere undae. We have a parallel to Ep. 2. 1. 7 in: luce demum, postquam dux et miles (internoscebantur) et facta (cog)noscebantur (Ann. 1. 39. 8).

But in many cases different verbs must be supplied. In: virginis os (habens), habitumque gerens et virginis arma (Aen. 1. 315), while habere seems proper for the person and gerere for what is worn, yet such uses as: formam similem gerit (Lucr. 4. 50) or: umbrata gerunt civili tempora quercu (Aen. 6. 772) make me hesitate to affirm that a Roman would supply a second verb here. So in: quem sese ore ferens, quam forti pectore et armis (se gerens) (4. 11). We have a pair of verbs standing in the relation of cause and effect represented by that which expresses the cause in: crudeles aras (ostendit) traiectaque pectora ferro nudavit (1. 356); Sychaeus had trusted the sanctity of altars that should have been sacred, and so fell a ready victim to Pygmalion, is my reading of Servius’s note: crudeles: epitheton hoc de causa est, nam arae piae sunt. So too in: longa tibi exsilia (ferenda) et vastum maris aequor arandum (2. 780), inclusos utero Danaos (liberat) et pinea furtim laxat claustra Sinon (2. 259), vocemque volens (emittit) atque ora resolvit (3. 457), disce puer virtutem ex me . . . fortunam (pete) ex aliis (12. 435). It seems the effect in: quod litore currum (evertere) et iuvenem monstris pavidi effudere marinis (7. 780), his fretus non legatos (praemisi) neque prima per artem temptamenta tui pepigi (8. 144), Tyrrhenamque fidem (sollicitare) aut gentes agitare quietas (10. 71), instans operi regnisque futuris (providens) (1. 504).


We have a single verb representing a pair that denote concomitant elements in the course of an action in: sic memorans umeros (amplectebatur) dextrasque tenebat amborum (9. 250), Phrygios . . . duces (interfice) pictasque exure carinas (7. 431), caeruleae cui terga notae (variabant) maculosus et auro squamam incendebat fulgor (5. 88), neu magis irae vestrae (pareatis) quam famae consulatis (Sall. Cat. 51. 8), tunc Arminius . . . reciperatam libertatem trucidatas legiones (commemorabat), spolia adhuc et tela Romanis derepta in manibus multorum ostentabat (Ann. 2. 45. 4), ius naturae (commemorant), labores educandi adversus fraudem et artes et brevitatem adoptionis enumerant (Ann. 15. 19. 2). We have a double zeugma in: consistere iussis militibus Italiam ostentat subiectosque . . . campos, (docet) moeniaque eos tum transcendere non Italiae modo, sed etiam urbis Romanae (conscendere) (Liv. 21. 35. 8). We have a threefold zeugma in: illa tibi Italiae populos (enumerabit) venturaque bella (narrabit) et quo quemque modo fugiasque ferasque laborem expediet (Aen. 3. 458-9) and one far more involved in: quod scelus aut Lapithas tantum (committentes), aut (quo scelere commisso) Calydona (tantas poenas) merentem? (7. 307).

In prayer or intercession to a lord or conqueror on earth or to a god in heaven a common posture is that represented in: et duplices tendens ad sidera palmas talia voce refert (1. 93). In silent prayer it is the eyes that are raised with the hands, and when the hands are bound we have:

Ad caelum tendens ardentia lumina frustra,

Lumina, nam teneras arcebant vincula palmas (2. 405-6),

where tollens would be the proper verb with lumina, but tendens with palmas. So we have tendo for tollo et tendo in:

illi has ego voces

Qua datur, hasque manus . . . a litore tendo (Val. Fl. 7. 269-70).

But more usual is the form we see in:

Et duplices cum voce manus ad sidera tendit (Aen. 10. 667),


tendoque supinas

Ad caelum cum voce manus (3. 176-7),

where cum voce is subordinated to manus, as we should expect when tendo alone is expressed. In: caelo palmas cum voce tetendit (2. 688) we have the same order we find in: manus ac supplices voces ad Tiberium tendens (Ann. 2. 29. 2), where the verb belongs to the more distant object; we have it with the nearer object in: liberti etiam ac servi patrono vel domino, cum voces, cum manus intentarent, 311 ultro metuebantur (3. 36. 1). But we have the verb belonging to the more distant object in: nec patris Anchisae cineres manesve revelli (Aen. 4. 427) for: cineres revelli manesque laesi; quin potius pacem aeternam pactosque hymenaeos exercemus? (4. 100) for: pacem exercemus et hymenaeos inimus; dedit iura quis pace et principe uteremur (Ann. 3. 28. 3) for: quis usi pace et principe frueremur.

In: sperat infestis, metuit secundis alteram sortem bene praeparatum pectus (Od. 2. 10. 13-14) we have a fourfold union expressed in full, of which we might have expected the related terms infestis . . . secundis to be expressed by the second term. In: quod arduum sibi (sumpsit), cetera legatis permisit (Ann. 2. 20. 2) we have a fourfold union thus reduced, giving a zeugma where the verb is used to express its opposite as well. We have similar zeugmas, where the verb expresses its opposite in part or in full, in: seu pacem (agam) seu bella geram (Aen. 9. 279), nec iam amplius armis (rem gerere), sed votis precibusque iubent exposcere pacem (3. 260-1), saepe (festinabat) velut qui currebat fugiens hostem, persaepe (incedebat) velut qui Iunonis sacra ferret (Sat. 1. 3. 10), demigrant Helicone deae quatiuntque novena lampade sollemnem thalamis coeuntibus ignem et de Pieriis (spargunt) vocalem fontibus undam (Slat. Silv. 1. 2. 4-6), nam sicut (non licuit) durare in hanc beatissimi saeculi lucem . . . ita festinatae mortis grande solacium tulit (Tac. Agric. 44. 5), utque ad fallendum silentio (agebant), ita coepta caede, quo plus terroris adderent, cuncta clamoribus miscebant (Hist. 5. 22. 3), datum id non modo precibus Artabani, sed (additum) contumeliae Pisonis (Ann. 2. 58. 3), simul nobilitatem domus (laudibus extulit), etiam ipsius quoquo modo meriti gravem casum miseratus (3. 17. 1), Agrippina, quae filio dare imperium (ausa est), tolerare imperitantem nequibat (12. 64. 6).

In many zeugmas we find a use of the special for the general, a use highly conducive to poetic vividness; with one of the subjects is joined a verb denoting a special form peculiar to it of the action in question, and this special form is understood with the subject or subjects to which it does not properly belong. We have examples of this in: retia rara, plagae, lato venabula ferro, Massylique ruunt equites et odora canum vis (Aen. 4. 131-2), where ruunt, used with equites, cannot be properly joined with retia or venabula, with which we must understand cito efferuntur. So in: inceptoque (perseverat) et sedibus haeret in isdem (2. 654), nunc hos, nunc illos aditus (tentat) omnemque pererrat arte locum (5. 441), hic thalamum invasit natae velitosque (iniit) hymenaeos (6. 623), ut celsas videre 312 rates (adire) atque inter opacum adlabi nemus (8. 108), non ut tela tamen (iaceret), non ut contenderet arcum (12. 815), haec Proteus, et se iactu dedit aequor in altum . . . at non Cyrene (filium reliquit) (Geo. 4. 528-30), coniurata tuas rumpere nuptias et (evertere) regnum Priami vetus (Od. 1. 15. 7), Esquilias quidem ab hoste prope captas (nemo defendit), et scandentem in aggerem Volscum hostem nemo submovit (Liv. 3. 67. 11), Thraecias urbes (visit), mox Propontidis angustias et os Ponticum intrat (Ann. 2. 54. 2), primo boves ipsos, mox agros (dedebant), postremo corpora coniugum aut liberorum servitio tradebant (4. 72. 4), grates dis (agens), atque ipsam recentis casus fortunam celebrans (15. 34. 2).

The opposite of this use of the special for the general was to be expected; and we have some interesting examples of it. In:

qui numina Phoebi,

Qui tripodas, Clarii laurus, qui sidera sentis

Et volucrum linguas et praepetis omina pennae (Aen. 3. 359-61),

we have a fourfold zeugma, where the general term sentis, though not properly adapted to union with any one of its objects, is used to represent four verbs, which with Servius’s aid I venture to express thus: qui numina Phoebi suscipis, qui oracula Pythii nosti, qui astrologiae peritus es, et volucrum linguis et pennae praepetis omnibus calles augur. In the following examples I have set in brackets the special verbs for which the general verb seems to be used: quos lucos et aquae (subterfluunt) subeunt et aurae (perflant) (Od. 3. 4. 8), iura (dant) magistratusque (creant) legunt sanctumque senatum (Aen. 1. 426), where Virgil has the verb usually selected by the beginner for his Latin prose, sacra (fert) manu victosque deos, parvumque nepotem ipse (ducit) trahit (2. 320), scalae improviso (adportatae) subitusque (adlatus est) adparuit ignis (12. 576), ne tenues pluviae (diluant) rapidive potentia solis (exurat) acrior, aut Boreae penetrabile frigus (congelet) adurat (Geo. 1. 92-3). With this use of adurat for congelet we may compare: quam. . . canis urebat Luna pruinis (Val. Fl. 2. 287), and further note that etymologists seem to find a common root for the Latin calidus or caldus, and the German kalt, our cold.

This curious form of zeugma is that most favoured by Tacitus, who begins in: ita qui olim boni aequique Cherusci (laudabantur), nunc inertes et stulti (culpantur) vocantur (Germ. 36. 2) with an easy and simple example of it. I have noted the following in the first two books of the Annals: qui iudicium (habuit) et poenas in hunc modum 313 (sumpsit) exercuit (1. 44. 3) following Virgil’s: laeva malorum exercet poenas (6. 542), quia Romanis Germanisque idem conducere (putabam), et pacem quam bellum (malebam) probabam (1. 58. 2), cunctos adloquio et cura sibi (conciliabat) et proelio (con)firmabat (1. 71. 5), sed Maroboduum regis nomen invisum apud populares (reddebat), Arminium pro libertate bellantem favor (sequebatur) habebat (2. 44. 3).

Thus far we have been dealing with zeugma, as though it had to do only with the verb. But the present and past participles of the verb are often used as adjectives, as in: umentes oculos et pallida bracchia tendens (Ov. Met. 14. 734), or: arte laboratae vestes ostroque superbo (pictae) (Aen. 1. 639), an example of zeugma with the past participle. We have further examples of this in: pallam signis (variatam) auroque rigentem (1. 648), tumulum, viridi quem caespite inanem (exstructum) et geminas, causam lacrimis (exstructas), sacraverat aras (3. 304-5), ipse Quirinali lituo (praeditus), parvaque sedebat succinctus trabea, laevaque ancile gerebat Picus (7. 187-8), quem pellis ahenis in plumam squamis (composita et) auro conserta tegebat (11. 771). In: videre rates . . . inter opacum adlabi nemus et tacitos (remiges) incumbere remis (8. 108) from rates adlabi we must imply remiges (= remigantes) with incumbere.

We have already noticed how in poetry the noun is often substituted for the verb, and that in the prose of Tacitus this substitution is common. So in: fidem atque pericula polliceantur (Ann. 2. 40. 3) we should expect: se fidos fore atque audaces polliceantur. In: mederetur fessis, neu mortem in isdem laboribus, sed finem tam exercitae militiae neque inopem requiem orabant (1. 35. 2) mortem, finem, requiem are used as co-ordinate with mederetur. So in: praeda famaque onusti (12. 28. 1) and: ordinem agminis disiecti per iram ac tenebras (Hist. 3. 22. 2) we have examples of our first variety of zeugma not dependent on inflexion; and in: Inachus Acrisiusque patres mediaeque Mycenae (patria) (Aen. 7. 372) one dependent on likeness of meaning and derivation.

In: diversis animorum motibus pavebant terrebantque (Ann. 1. 25. 2), where for: metuendo et minando Tacitus substitutes a general term: diversis animorum motibus, we have a good example of our last class favoured by Tacitus. In:

vides quae maxima credis

Esse mala, exiguum censum turpemque repulsam

Quanto devites animi capitisque labore (Ep. 1. 1. 42-44),

for the general term labore we must substitute: dolore et periclo to 314 restore the fourfold union, balancing them with: exiguum censum turpemque repulsam. Here animi dolore must be related to exiguum censum, and is fully described in vv. 45-6, while capitis periclo is the risk run to avoid a disgraceful defeat in an election (turpem repulsam), having nothing to do with loss of life in Horace’s day, and referring merely to the loss of caput or civic status consequent on a conviction for bribery. Dolore and periclo both appear as readings for labore in the manuscripts of Acron, whose note: vis etiam mori propter honorem, seems to have misled some of our editors.



To: nube candentes umeros amictus augur Apollo (Od. 1. 2. 31) Acron’s note is: nube candentes. Melius candenti nube, quam candentes umeros, amictus. Candidis nubibus velatus, scilicet qui videri possis; irati enim di nequeunt conspici. Hauthal wishes, with the usual wisdom of the emendator, to change irati to velati; but it is easy to believe that when the gods, like Tennyson’s Hera, grow angry, they withdraw from human view ‘into the golden cloud’. Bentley thinks that Acron wishes to change Horace’s text here from candentes to candenti, and pleads that the resulting hiatus makes this impossible. But from Acron’s note it is perfectly plain that candentes was the reading he found in his manuscripts, as it is in the best we have. As regards the meaning, it is true that the gods wrap themselves in clouds to hide from mortal view, if we may believe the ancients. Homer represents this in: νεφέλῃ εἰλυμένος ὤμους (Il. 5. 186), and Virgil in: Venus obscuro faciem circumdata nimbo (Aen. 12. 416), But that, as the gods were hidden from mortals obscuro nimbo, so they could be revealed to them candenti nube,—the contradiction revolts Bentley, Hoc tamen inepte incommodeque, ut nihil supra. So he turns resolutely from the candenti nube of Acron to the candentes umeros of the Sungod, and seeks a parallel in other shining objects that the Roman poets present, the candidos lacertos of Tibullus’s Pholoe, and the candidos umeros of Horace’s Lydia. Shall we join in his refrain: hoc tam apte quam nihil supra?

For we may feel that Lydia’s shoulder-polish was a little different from the radiance that made the shoulders of the favouring god visible to his worshippers. And it is to show us the source of this radiance that Acron draws our attention to the shining cloud, which as it enwraps a modern saint or angel, our worshippers to-day, strangely enough, term a nimbus. Acron feels that from candentes umeros we must imply candenti with nube; for from the shining cloud comes the light that reveals to mortals the beneficent deity, who when angered veils himself in the nimbus or storm-cloud. So 316 the threefold: nube candentes umeros goes back to a fourfold: nube candenti candentes umeros; the epithet candens belongs rather to the nubes, the source of the radiance, than to the umeri, which the radiance lights up. This seems the meaning of Acron’s melius; candenti would be better in prose, and it is prose he is writing in his notes. The epithet candens is transferred by the poet from nubes to umeri; and by this reduction of a fourfold union to three we get the figure of the transferred epithet, the most usual form of hypallage.

Of the nature and origin of this transference Bentley has no idea, nor does he always recognize it when he meets it. In: iam tibi lividos distinguet Auctumnus racemos purpureo varius colore (Od. 2. 5. 10-12) he rejects the manuscript reading just given, as well as Lambinus’s emendation to: purpureus vario, and gives us in purpureo varios the plain prose to which his nature directs him. Here the poet gives Auctumnus a purple complexion, because he changes the grapes to purple, just as with the poet atra mors passes to pallida mors because of the pallor of the dead. It is the fourfold union: varius Auctumnus racemos varios, with the omission of varius in its apter and more obvious application. And yet Bentley understood why Horace in: inaequali tonsore (Ep. 1. 1. 94) calls a barber ‘uneven’: inaequalis est tonsor, qui tam prave capillum secat, ut una parte brevior sit, altera longior.

But again in: timet . . . miles sagittas et celerem fugam Parthi (Od. 2. 13. 17) against all his manuscripts he wants to change celerem to reducem. The soldier has no dread of the swift flight of the Parthian; it is the feigned flight he dreads and the swift arrows with which, as he turns back, he transfixes the soldier. Celerem is the epithet transferred from sagittas to its concomitant fugam; and it is because the feigned flight of the Parthian is swift in these speeding shafts that the soldier dreads it. On the next page I find his note on: pugnas et exactos tyrannos densum umeris bibit aure volgus (2. 13. 31-2), where following, as he imagines, Acron’s note: bibit, avide audit, he changes umeris to avida. Densum volgus, he says, is sufficient without further qualification, but we want an epithet with aure. So he writes avida aure, failing to see that, when Acron makes bibit equivalent to avide audit, he shows how Horace by his striking oxymoron: bibit aure really supplies this lack. When Ovid writes: densum trabibus nemus (Met. 14. 360) it is short for: nemus densum densis trabibus; and Horace writes here: densum 317 umeris volgus for: densum densis umeris volgus ‘a crowd packed with thronging shoulders’.

Let us begin with an example where the fourfold union has not been reduced to three. We read: illum absens absentem auditque videtque (Aen. 4. 83), where, though absens could be omitted and is usually omitted in prose, its expression adds greatly to the poetic force of the verse. The usual prose construction is that shunned in verse; and the omission of absentem would have made the line weaker still. But in: pulverulenta fuga Rutuli dant terga per agros (12. 463), where fuga dant terga is for: conversi fugiunt, of the repeated terms we have that retained which is the less obvious and goes with the word which gives the verse its distinctively poetic character. In: procella velum (adversum) adversa ferit (1. 103), serpens . . . quem obliquum rota (obliqua) transiit (5. 274), mixtoque insania (mixta) luctu (10. 871), furens (media) mediisque in milibus ardet (1. 491), of the several pairs the terms omitted are so clearly implied in those expressed that their omission would not be felt, were it not that the term omitted in each case is the more obvious, and that which would have found expression in prose. In: saxum (spumans) spumantia contra litora (5. 124) the very omission of spumans tends to call up in the mind of the reader the picture Stevenson gives us in the Merry Men; and the same seems true of the omissions in: postes auro spoliisque superbi (2. 504), or: Troiae renascens alite lugubri fortuna tristi clade iterabitur (Od. 3. 3. 62), or: miramur, facilis ut (lyram) premat arte manus (Prop. 2. 1. 10).

The old grammarians, starting from the most puzzling form of hypallage, defined it as: mutua casuum permutatio, as we have it in: dare classibus Austros (Aen. 3. 61) for: dare classes Austris, the obvious prose form. The form of hypallage of which we have just cited some examples they styled hypallage imperfecta; for hypallage perfecta they required that both cases should be exchanged with each other, as in: et cum frigida mors anima seduxerit artus (4. 385) for: animam seduxerit artubus.

Let us cite first a few simpler cases of such double hypallages; for double they seem to be. In:

Ibant obscuri sola sub nocte per umbram (6. 268),

Servius says we have a hypallage for: sub obscura nocte soli ibant. This seems to be the result of a union by distribution of two threefold clauses reduced from ibant soli sub sola nocte, and ibant obscuri sub obscura nocte, in each of which the more obvious term of the 318 like pairs is omitted as sure to be suggested by the expression of the less obvious term. We have another example in: vides ut alta stet nive candidum Soracte (Od. 1. 9. 1) ‘you see how Soracte stands white with high snow’. Of course it is Soracte that is high and the snow that is white. So Shakspere tells of ‘books in the running brooks, sermons in stones’. In: perpetui numquam moritura volumina Sili (Mart. 7. 63. 1) it is Silius who is to be read from age to age, and his volumes which will never die;A and in: si quis procurrat ad oras ultimus extremas (Lucr. 1. 970) it is he who has run up to the outer edge, who is beyond it; and he does pass beyond it by hurling his spear. In: aurea quam molli tergore vexit ovis (Prop. 2. 26. 6) it is the sheep that is golden and the hide that is soft.

We have other examples in: tacitae per amica silentia lunae (Aen. 2. 255), where but for the hypallage tacitae would be superfluous, at subitae horrifico lapsu de montibus adsunt (3. 225), omnem cursum mihi prospera dixit religio (3. 362), madidaque fluens in veste (5. 179), seraque terrifici cecinerunt omina vates (5. 524), tepidaque recentem caede locum (9. 455), sic flammas aditura pias aeterna sacerdos (Ov. Am. 3. 7. 21). I have spoken of: auras suspiciens hausit caelum (Aen. 10. 899) as a double metonymy; the older grammarians would have called it a perfect hypallage; but there is no visible exchange of cases discernible. I suspect that metonymy comes to pass much in the same way as hypallage, though I have not traced it fully.

We seem to have it transferred to verbs, where exchange of cases is out of the question, in: vos . . . coetum . . . celebrate faventes (1. 735) for: vos coetui favete celebrantes. In: vix primos inopina quies laxaverat artus (5. 857), while primos is transferred from quies to artus, inopina is transferred to quies not from artus, but from the suppressed Palinuri. So: omnem cursum mihi prospera dixit religio (3. 362) is for: totum cursum prosperum futurum praedixerunt omina prospera omnia, a combination impossible for poetry. But in: rudentem contorsit laevas proram . . . ad undas (3. 562), postera . . . dies primo surgebat Eoo (3. 588), candenti perfecta nitens elephanto (6. 895), Idaeae sacro de vertice pinus (10. 230) the double hypallage seems to have come from the regular interchange of epithets.

The easiest transference here seems that from a noun to its dependent genitive, or from a genitive to its governing noun. The two form a complex of meaning which facilitates the transfer, and inclines me to think that it was from this starting-point that the figure 319 developed. In: incredibilis rerum fama (Aen. 3. 294) or: innumerabilis annorum series (Od. 3. 30. 4) the transference seems easy and natural; and in: ruentis imperi rebus (1. 2. 26) or: miseri post fata Sychaei (Aen. 4. 20) one sees the omission of a term rather than a transfer, which is of course the truth. Of the transference from the genitive we may note these examples: vastum maris aequor (2. 780), imo barathri . . . gurgite (3. 421), Euboicis Cumarum . . . oris (6. 2), Tyrrhenus tubae . . . clangor (8. 526), arma dei . . . Volcania (12. 739), superbos Tarquini fasces (Od. 1. 12. 34), iratos . . . regum apices (3. 21. 19), Tyrrhena regum progenies (3. 29. 1), regum tumidas . . . minas (4. 3. 8); and of transference to the genitive: molem hanc inmanis equi (Aen. 2. 150), summi fastigia tecti (2. 302), templum . . . desertae Cereris (2. 714), iugis summae . . . Idae (2. 801), variarum monstra ferarum (6. 285), confusae stragis acervum (6. 504), Stygiique per flumina fratris (10. 113), coetu variantis acervi (Lucr. 1. 775).

We find also the epithet transferred from the subject to an adverbial clause, as in: vertimus . . . certantibus aequora remis (Aen. 3. 668), haec . . . canit divino ex ore sacerdos (3. 373), primisque elabitur undis (5. 151), quadrupedante . . . sonitu quatit ungula campum (8. 596). In: mediaeque per Elidis urbem ibat ovans (6. 588), iuvat . . . mediosque fugam tenuisse per hostes (3. 283), tres . . . incertos caeca caligine soles erramus pelago (3. 203) we feel the omission with the subject of medius, incertus, caecus, and feel the change in meaning of the last two when used with adverbial phrases. Martial’s verse: nulla magis toto ianua poste patet (1. 70. 14) seems for: nulla ianua magis patet totis cum postibus tota.

In return we have many cases where the epithet has been transferred from the adverbial phrase to the subject, as: quibus ibat in armis aureus (Aen. 9. 269), eludit gyro interior (11. 695), nocturnusque vocat clamore Cithaeron (4. 303), creber utraque manu pulsat (5. 460), tot pullulat atra colubris (7. 329), imaque sedit inguine (10. 785), cui pineus ardor acervo pascitur (11. 786), manu trepidae iaciunt (11. 893), ubi plurima fuso sanguine terra madet (12. 690), hesternisque rubens deiecta est herba coronis (Mart. 9. 61. 17). In: purpureis ales oloribus comissabere (Od. 4. 1. 10) purpureis ales oloribus is for Venus purpureorum olorum alis vecta, where alis vecta has been shortened to ales, a striking extension of this transfer.

In like fashion in: quod adest memento componere aequus (Od. 3. 29. 33) aequus is transferred to the subject from the adverbial 320 phrase aequo animo, which has disappeared as a result of the transfer. So in: hunc primo levis hasta Themillae strinxerat (Aen. 9. 576) levis is for a vanished leviter, and so in: ubi prima fides pelago (3. 69) and: lacrimae volvuntur inanes (4. 449).

In: fessos opibus solatur amicis (5. 41) we have in: amicis an epithet transferred from the object to an adverbial clause, and so in: diversa per aequora vectos (1. 376). We have the opposite in: Herculeis sopitas ignibus aras excitat (8. 542), and in: tribus aut novem miscentur cyathis pocula commodis (Od. 3. 19. 12) it is from the subject of a passive that the epithet is thus transferred. In: alitis in parvae subitam collecta figuram (Aen. 12. 862) the subitam is transferred to the object from a vanished adverb subito, and in return in: hanc primum ad litora classem conspexi venientem (3. 651) primum is plainly for primam.

In such transfers we have many examples of the double or perfect hypallage, but they are of nouns, not adjectives. Perhaps the failure of the ancients to draw our distinction between nouns and adjectives here may have to do with this difficulty, but it does not solve it. We read: (navem) rapidus vorat aequo re vortex (1. 117), where we expect: vorat aequor rapido vortice. So: (navis) excussa magistro (6. 353) for: nave excussus magister. In: celebramus litora ludis (3. 280) the old meaning of celebramus, ‘we throng’, obviates the hypallage. But in: sanieque aspersa natarent limina (3. 625) for: saniesque aspersa fluebat in liminibus, and: excisum Euboicae latus ingens rupis in antrum (6. 42) for: antrum incisum in latus rupis, this exchange is the only account I have to give of the irregularity. In: magno clamore morantur (5. 207) for: magna voce morati clamant, the verb is involved in the interchange. In: horrere videns iam colla colubris (6. 419) we have a transference to the subject from the dative which points to the interchange of constructions with dono and circumdo. The transference is still to the subject in: respiciunt totumque adlabi classibus aequor (10. 269), but it is now from an ablative. It is to the object that the transference takes place from the ablative in: spem fronte serenat (4. 477) and: Lavinia . . . visa . . . longis comprendere crinibus ignem (7. 73).

We have the epithet transferred from the subject to the object in: caeca regens filo vestigia (6. 30), maestasque sacravimus aras (5. 48), laetum cuncti celebremus honorem (5. 58), rorantia vidimus astra (3. 567), where the transference indicates that the foam rose so high as not merely to drench the sailors, but as to seem to drench the 321 heavens. So too in: corpora . . . ignibus aegra dedere (2. 566), fessum quotiens mutet latus (3. 581), frena ferox spumantia mandit (4. 135), tacitumque obsedit limen Amatae (7. 343), dum terga dabant palantia Teucri (12. 738), per te immaturum mortis adimus iter (Prop. 3. 7. 2), medias fallit permixta sorores (Stat. Silv. 1. 2. 10). In: sese medium iniecit periturus in agmen (Aen. 2. 408) medium may be joined either with sese or with agmen, giving a threefold transfer.

In return we find the epithet transferred from the object to the subject in: tectusque tenet se (Aen. 10. 802), conversique oculos inter se atque ora tenebant (11. 121), dum sacra secundus haruspex nuntiet (11. 739), qua . . . (fortuna) ostendit se dextra (2. 388), verus mihi nuntius adfers (te) (3. 310), lumenque obscura vicissim luna premit (4. 80), tu secreta pyram . . . erige (4. 494), navis se tarda movebat (5. 280), sonitum dat stridula cornus (12. 267). In: miratur nemus insuetum fulgentia longe scuta . . . pictasque . . . carinas (8. 92) we at once supply: insueta spectacula. In: convolvit . . . terga arduus ad solem (2. 475) we see that arduus is for: ardua adtollens; cf.: sibila colla arduus adtollens (5. 278). In: ille dies . . . primusque malorum causa fuit (4. 169) primus is transferred to the subject from causa, a predicate nominative, which is for an old accusative. In: tectusque recusat (2. 126), where tectus may be taken as for caute, it seems to stand for: dum se tegit.

Again from these two transferences, from the subject to the object and the reverse, we have a double or perfect hypallage, the strongest of all, by which the living subject is made the object, and the object, often lifeless, becomes the subject. In: me . . . infelix habuit thalamus (6. 521) this double hypallage seems less strange to us than it would to the Roman, to whom habet here is the older use for its frequentative habitat, as we see it in: ille geminus qui Syracusis habet (Pl. Men. 69). To: cui plurimus ignem subiecit rubor (Aen. 12. 66) Servius’s note is: hypallage pro ‘cui ignis animi subiecit ruborem’. We read: quae te, genitor, sententia vertit? (1. 237) and: neque me sententia vertit (1. 260). Servius recognizes the hypallage in both, and in the former he suggests that quae is for cuius; for Venus does not, he thinks, venture to assail Juno directly. Other examples are: apes . . . exercet sub sole labor (1. 431), supremo cum lumine vita reliquit (quemquam) (6. 735), sin nostrum adnuerit nobis victoria Martem (12. 187).

The relation of the dative with the accusative seems often close, and there is in both Latin and Greek a constant transition from the 322 use of the dative with transitive verbs to that of the accusative, a transition which has led to the loss of the dative in modern Greek, and which in Latin leads Terence to the use of the accusative with inmineo, and classical writers to join either the accusative or the dative with temperare and moderari, culminating in Apuleius’s use of the accusative with suadere for the person persuaded. To the analogy promoted by this is perhaps due the transference of the epithet from the dative to the nominative we find in: miserandaque venit arboribus satisque lues (Aen. 3. 138) and: mentis vacat hic tibi solus fortunaeque locus (11. 179). With these must be placed such hypallages as: ingens adgeritur tumulo tellus (3. 63) and: parietibus textum caecis iter (5. 589). In: omnibus idem animus . . . dare classibus Austros (3. 61) Servius tells us we have a hypallage for: dare classes Austris; and in: arma parate animis (11. 18) he suggests that we have a hypallage for: armis parate animos, where armis will be a metonymy for bello (gerendo). To: ne tanta animis adsuescite bella (6. 832) Servius’s note is: mire dictum: ab ipsis enim quasi consuetudinem facit populus Romanus bellorum civilium. This may be the true account of this puzzling hypallage; the wars, the object primarily indirect to which they were accustoming themselves, became in time their habit itself, which would be a cognate accusative with adsuefacere, and as such were expressed as the direct object of adsuescere here, while the old direct object passed by analogy to the dative.

Closely similar to this is the course of transference with dono and circumdo. Plautus has: modo qui hanc mihi donavit (Poen. 469), but: hoc donavisti dono tuom servom Stichum (Stich. 656). Livy has: eorum contioni satellites armatos circumdedit (34. 27. 5), but Caesar: Octavius quinis castris oppidum circumdedit (B. C. 3. 9. 4), and Virgil: arma . . . circumdat nequidquam umeris (Aen. 2. 510), but: canibus circumdare saltus (Buc. 10. 57). With a compound of do we may assume that the dative is the older construction, and that presently the object surrounded is felt as an object of cognate notion and is transferred to the accusative; when the old accusative, following the analogy of: donare aliquem aliqua re, is in turn transferred to the ablative. For this seems the older construction with dono, a derivative of donum and a verb of later formation, which is later led by the analogy of do to take the construction: donare alicui aliquam rem as well. This double construction is extended to many other verbs in poetry; we note on the analogy of circumdo with the ablative: 323 plurima salute Parmenonem . . . impertit Gnatho (Ter. Eun. 270), me divom pater . . . fulminis adflavit ventis (Aen. 2. 649), fama est . . . Trinacriam caelum subtexere fumo (3. 582), duroque intendere bracchia tergo (5. 403), paribus palmas amborum innexuit armis (5. 425), idem ter socios pura circumtulit unda (6. 229), spumantis equi (in)foderet calcaribus armos (6. 881). The following are swayed by the analogy of donare with the dative: vina bonus qui deinde cadis onerarat Acestes (1. 195), stipatque carinis ingens argentum (3. 465), inmeritam saevae natam mactare Dianae (Ov. Met. 13. 185); and Horace has donare thus constructed in: donarem . . . meis aera sodalibus (Od. 4. 8. 2).

The uses of induo deserve separate mention; it is the opposite of exuo, and is compounded of indu (= into) and ovo ‘I dress’; cf. subucula and exuviae. It seems to have been primarily constructed with three regimens, as in: (1) sibi et torquem et cognomen induit (Cic. Fin. 2. 73. 22); (2) soccos quibus indutus esset (De Orat. 3. 127. 32); (3) tu te in laqueum induas (Pl. Cas. 113). For in with the accusative Virgil has substituted the ablative in: an sese mucrone ob tantum dedecus amens induat (Aen. 10. 682), as have Livy in: induissent se hastis (44. 41. 9) and Caesar in: quo qui intraverant se ipsi acutissimis vallis induebant (B. G. 7. 73. 4).

The transfer of an epithet to a different object will usually lead to a change of meaning in the epithet itself. Often this change is so slight that we do not feel the need of a different epithet in the new connexion; as in: placidam per membra quietem inrigat (Aen. 1. 691). But with a change of element, though the word transferred may still seem adequate, the change of meaning is quite plain, as in tremulo in: splendet tremulo sub lumine pontus (7. 9), where the ripple of the water is transferred to the sunlight and the lustre of the sun to the depth of tawny Tiber; or in: late ferreus hastis horret ager (11. 601), where the iron of the spears is transferred to the field now bristling with them; or in: pontus . . . scopulos . . . superiacit unda spumeus (11. 626), where we feel that spumeus belongs rather to the wave or to the cliffs than to the deep, to which it is here transferred. In: adversis rerum inmersabilis undis (Ep. 1. 2. 22) adversis has, besides the force of ‘fronting’ with undis, the force of ‘adverse’ with rerum. So in: positus . . . carbo in caespite vivo (Od. 3. 8. 3) ‘a live coal placed on the green turf’. In: duris dolor ossibus ardet (Aen. 9. 66) the change from cruel to hard involved in the transfer of durus from dolor to ossibus is very plain. So in: male barbaras regum libidines ulta est 324 (Od. 4. 12. 7) the change involved in the transfer of barbarus from reges to libidines is clear.

Just as we transfer our speed in a railway carriage to the homes and trees that seem speeding past, so Virgil transfers the rush of a flying missile to the air through which it passes, as in: sagitta . . . volucres diverberat auras (Aen. 5. 503). We have a like transference, but from sound, in: celeres defer mea dicta per auras (4. 226), and from the messenger god in: ferre iubet celeres mandata per auras (4. 270). We have a complete conversion of this transfer to the moral sphere in: (malos) ad impia Tartara mittit (6. 543), where, however, Virgil, to avoid omitting the necessary malos, has substituted impia for it with Tartara with a fine enhancement of force from the personification. In: tali Cyllenius ore locutus mortales visus medio sermone reliquit (4. 277) medio is transferred from Mercury, who has not given his thought full expression, to his speech thus untimely interrupted, giving it the double meaning of ‘incomplete’, a force not inconsistent with medius—and ‘ended’, the opposite of medius. It recalls: Rumone secundo (8. 90) with the double meaning in: secundo of ‘favouring their seeming speed’, a meaning not inconsistent with secundus—and ‘upstream’, a meaning the opposite of secundus. There are more of such contradictions in Virgil, as we shall see. In: vescitur Aeneas . . . perpetui tergo bovis (8. 183) the epithet perpetui properly belongs to tergo ‘the unbroken chine’; but when transferred to bovis gets an entirely new meaning, as Servius tells us. The meat of the ox sacrificed to Hercules, when sold, brought such a price that with it the worshipper was able to buy a new ox, which he called perpetuus ‘that continued to him by the gods’; in other words, he was able to eat his cake and have it.

But the change is, as a rule, more marked when the transference is from a living to a lifeless object, or the opposite, as in: apricis statio gratissima mergis (5. 128), where apricus is transferred from the sunny nesting-place to the gulls delighting in the sunshine. To turn to men, in: inutile ferrum cingitur (2. 511) the transition is from Priam disabled by old age to his sword, useless because of the old age of its wearer. And so in: ni frustra augurium vani docuere parentes (1. 392) and: continuo in montes sese avius abdidit altos (11. 810). In: agmine remorum celeri ventisque vocatis prona petit maria (5. 211-12) agmine has three meanings, and we may translate ‘by the push of the speeding rowers’, or ‘by the swift sweep of the oars’, or ‘by the swift array of oarsmen’. In: prona maria we have also 325 a transference from proni remiges, implied in agmine remorum, when it means the seas down which they swiftly glide; for when applied to the waters pronus has the idea of sloping down to the land; ‘I sail into port’ is: (in portum) deferor (Ep. 1. 1. 15); compare ἀναπλεῖν and καταπλεῖν. In: manesque vocabat Hectoreum ad tumulum (Aen. 3. 304) the epithet is properly Hectoreos with manes, and is transferred to tumulum only by prolepsis; the tumulus here is a cenotaph, and only when Andromache can by the blood of her victims attract to it the manes of Hector, will it become ‘Hectoreum tumulum’.

Virgil’s use of tacitus in such transference is interesting. In: quis te, magne Cato, tacitum . . . relinquat? (6. 841), though one may remember Cato’s skill in oratory, and the record that, though he was prosecutor in more cases than any man of his day, he never lost a case, the obvious meaning is of course: nullis laudibus elatum. In: tacitumque obsedit limen Amatae (7. 343) Servius sees that the threshold is to be used as an ambush, from which Allecto may surprise the queen; and so tacitum is here for: ubi reginam occulte excipiat. In: quae sublegi tacitus tibi carmina nuper (Buc. 9. 21) tacitus seems for: nesciente te, words that Servius thinks should be supplied. In: totumque pererrat luminibus tacitis (Aen. 4. 364), Servius explains: luminibus tacitis pro ipsa tacita, which is far from satisfactory; no doubt her very sighs were eloquent. But Servius knew of others who had a better grasp of the meaning here, and recognized that tears constituted a language of the eyes; for they explained tacitis as for siccis. In: tacitis regnavit Amyclis (10. 564) the meaning is more involved. Both the Spartan Amyclae and the Italian town of that name, so called by its founders who came from Amyclae in Sparta with Castor and Pollux, both perished because of their silence; the Spartan Amyclae, because the townsmen, tired of false alarms, imposed silence on all announcements of the coming of a foe. But though silence was a marked characteristic of the Pythagorean discipline that prevailed in the Italian Amyclae, it was not this part of their creed that destroyed the townsmen, but their rule, founded on their belief in metempsychosis, that forbade them to destroy living creatures; so that they tolerated the serpents that bred in the neighbouring marshes, till they multiplied so as to destroy the town.

We have seen how: totumque pererrat luminibus tacitis is short for: totumque pererrat ipsa tacita luminibus siccis, where we have the epithet, proper in its primary connexion, taking the place of another that better fitted its new position. So we read: alto a 326 sanguine divom (Aen. 5. 45) for: progenie clara altorum deorum, ipse deum manifesto in lumine vidi (4. 358) for: manifestum claro in lumine, primam qui legibus urbem fundabit (6. 810) for: legibus qui primus priscam urbem fundabit, frontem obscenis rugis arat (7. 417) for: frontem deformem obscenis rugis, udae vocis iter (7. 533) for: vocis liquidae udum iter, insani Martis amore (7. 550) for: Martis furentis insano amore, foedati . . . ora Galaesi (7. 575) for: ora lacerata foede interfecti Galaesi, quos illi bello profugos egere superbo (8. 118) for: illi superbi profugos egere iniusto bello (Serv.), caede viri tanta (10. 426) for: caede tam atroci viri tam magni, extremam . . . sinu perfundit harenam (11. 626) for: sinu extremo proximam harenam perfundit, formidolosis dum latent silvis ferae (Epod. 5. 55) for: dum latent in silvis tenebricosis formidolosae ferae, molem propinquam nubibus arduis (Od. 3. 29. 10) for: molem arduam et propinquam altis nubibus, iracunda Iovem ponere fulmina (1. 3. 40) for: Iovem iracundum perfervida fulmina deponere, regina dementes ruinas parabat (1. 37. 7) for: regina demens praecipites ruinas parabat.

Easier and more simple transferences for different words are: delecta . . . corpora . . . includunt caeco lateri (Aen. 2. 19) for: includunt occulta obscure lateri, creber ad aures visus adesse (multorum) pedum sonitus (2. 731), vocemque (resonantem) inclusa volutant litora (5. 149), armatum peditem gravis adtulit (gravida) alvo (6. 516), utque pedum primis infans (incerta) vestigia plantis institerat (11. 573), obliti ignoto camporum in pulvere (neglectum) relinquunt (11. 866), nemus (umbrosum) uvidique Tiburis ripas (Od. 4. 2. 30), de te splendida Minos (iudex illustris) fecerit arbitria (4. 7. 21), prensare manu (leni) lentissima bracchia (Sat. 1. 9. 64), aliae spem gentis adultos educunt fetus (Geo. 4. 163) for: educendo educant; Servius explains: educendo faciunt adultos.

Very difficult once seemed to me such constructions as: bacchatamque iugis Naxum (Aen. 3. 125), Neritus ardua saxis (3. 271), Averna sonantia silvis (3. 442); I expected what I find in prose, the ablative of description, as in: Naxum insulam iugis bacchatis, Neritus insula arduis saxis. But it will be clear, from the examples which we have just reviewed, that Virgil has here transferred the epithet, e.g. ardua, from the descriptive ablative to the object described, giving him a new form of description with which he can vary the descriptive ablative, a form which, involving as it does the poetic transference of the epithet, he favours as against the prose idiom of the ablative of description. We have here the epithet in agreement with the object described 327 completed by an ablative of specification, as we see it in: Arcentis filius . . . insignis facie (9. 583), totidemque pares aetate ministri (1. 705), socios . . . praestantes virtute (8. 548), progeniem virtute . . . egregiam (7. 258), pictas abiete puppes (5. 663), urbes litore diductas (3. 419). When we compare: forma pulcerrima Dido (1. 496) or: ipse acerrimus armis (12. 226), where the ablatives may be classed as the old ablatives of specification, with: praedurum viribus Orsen (10. 748) or: egregium forma iuvenem (12. 275), we see how closely the new Virgilian ablatives of specification resemble the old.

But in Aen. 12. 275: egregium forma iuvenem et fulgentibus armis is the full verse, where we find this new substitute for the ablative of description placed beside an example of that ablative to vary the diction. We have the same union in: turbidus hic caeno vastaque voragine gurges (6. 296) with which compare 9. 105, Euryalus forma insignis viridique iuventa (5. 295), (hastam) rudem nodis et cortice crudo (9. 743), (telum) solidum nodis et robore cocto (11. 553), mille . . . densos acie et horrentibus hastis (10. 178), dona . . . auro gravia sectoque elephanto (3. 464), to take a few examples I have noted. We have the order reversed in: cervus . . . forma praestanti et cornibus ingens (7. 483) and: pari ferocia et velut aucti numero (Ann. 2. 25. 5). Tacitus gives us the opposite of this construction in: Augustum fessa aetate (Ann. 1. 46. 3) for: Augustum fessum (= confectum) aetate. In prose the ablative of specification is often joined with the preposition a or ab, as in: nisi qui a philosophia, a iure civili, ab historia fuisset instructior (Cic. Brut. 161. 43). So in this Virgilian form we find at times a or ab, as in: recens a volnere (Aen. 6. 450), a stirpe coniunctus (8. 130), Alpheae ab origine Pisae (10. 179).

But of like origin and meaning with the ablative of description (qualitatis) is the genitive of description, an idiom peculiar to Latin. Though it is frequent in Horace, e.g.: parvula . . . magni formica laboris (Sat. 1. 1. 33), Virgil seems to shun this idiom. I have noted only six examples of it in the Aeneid: alter ab Arcadio Tegeaeae sanguine gentis (5. 299), Dardanius divinae stirpis Acestes (5. 711), atri velleris agnam (6. 249), Assaraci quem sanguinis (6. 778), Ascanius clari condet cognominis Albam (8. 48), nodum informis leti (12. 603). To: geminos inmani pondere caestus (5. 401) Servius’s note is: inmani pondere pro ‘inmanis ponderis’ more suo, ut ‘hamis auroque trilicem’ (3. 467). We have here proof positive that Servius saw the close relation between the ablative of description and the Virgilian ablative of specification, of which: hamis auroque trilicem is an 328 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 opportune 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 329 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: mediae 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 perfect hypallages such as we have repeatedly examined in this chapter.

A In: perpetui numquam moritura volumina Sili (Mart. 7. 63. 1) it is Silius who is to be read from age to age, and his volumes which will never die;
Text printed as shown; hand-written correction reads “it is his books which are to be read ... and he who will never die”. This is probably what the author meant.



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. 1).

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) 331 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 is 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 332 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, ἐστι δὲ τερπομένοισιν ἀκούειν (Od. 15. 392) is parallel, but does not account for the 333 Latin idiom. Interesting here is Propertius’s use of potest for licet in: pauper, at in terra, nil ubi flere potest (3. 7. 46), evidently for: ubi sine lacrimis vitam agere licet. Compare: non sic futurumst: non potest (Ter. Phorm. 303), or: dic exeunti, potes non reverti: dic redeunti, potes non exire (Sen. Ep. 49. 9), or: possum scire ego istuc ex te quid negoti est? (Pl. Cas. 654). The use of can for may, usually so sternly censured by pedagogues with us, was usual in colloquial Latin.

In: ubi non Hymetto mella decedunt (Od. 2. 6. 15) the addition of de leaves the meaning of cedunt unchanged; just as its omission in: socii cesserunt aequore iusso (Aen. 10. 444) leaves the verb with the same meaning it has in: e pastu decedens agmine magno (Geo. 1. 381); and in prose we see the like in: inde cessero: in Africam transcendes (Liv. 21. 44. 7) and: de provincia decessit (Cic. Verr. 2. 2. 48. 20). We find pellere for depellere in: patrio . . . pellere regno (Aen. 3. 249), Europa atque Asia pulsus (1. 385), and in prose in: uti omnes ex Galliae finibus pellerentur (B. G. 1. 31. 11) and: ut possessores pellantur suis sedibus (Off. 2. 78. 22). Virgil has: praeceps se . . . ad undas misit (Aen. 4. 254), but: demissa . . . nubibus Iris (10. 73). Virgil has always vastare, but Livy has: ad devastandos fines discessere (4. 59. 2) as well as: omnia ferro ignique vastata (7. 30. 15). Following the analogy thus established, we are not surprised to find: agros Remorum depopulati (B. G. 2. 7. 3) and: qui cum agros maximos et feracissimos per se ipsum, . . . depopularetur (Verr. 2. 3. 84. 36), but: eorum agros populaturum (B. G. 5. 56 fin.) and: noctu populabatur agros (Off. 1. 33. 10). From depopulari ‘to strip a land of its people’ by the omission of the prefix we get the form populari, just as from cohortari ‘to encourage the cohorts’ we get hortari. Populari and hortari look like the older and simpler forms, but the reverse is the case. With depopulari as regards formation we may compare expectorare and degenerare; degener is later and is first formed in Virgil. So in later Latin we get verbs like adripare, the French arriver, and aboculare, the French aveugler. Woelfflin mistook the secondary populari for the primary form, of which he thought depopulari a derivative, and explained it as from populus ‘the army’, and as meaning ‘to traverse the country with an army’. But we have little evidence that to the Roman populus ever meant an army, though in the days of Prussian lordship the idea seemed natural enough to Woelfflin.

But perfectly parallel to the use of populari for depopulari is the use 334 of texit for detexit in Virgil without change of meaning, asserted by Servius for Aen. 10. 424-5:

dum texit Imaona Halaesus

Arcadio infelix telo dat pectus inermum.

All the moderns translate: ‘while Halaesus covered (i.e. guarded) Imaon with his shield, the unfortunate man left his breast bare to the shaft of the Arcadian chief’. Dum here takes the perfect of tego, and is joined with dat, the present of dedit, a construction exactly opposite to that prevailing for dum in prose and poetry; for dedit is not lasting but momentary. Virgil might use this exact opposite as a surprise to his readers, but it occurs nowhere else; and it is far more likely that he used texit for detexit after the analogy of populatur. Servius’s note is: dum texit, dum spoliat; nam tempus praesens est, non praeteritum ab eo quod est tego. Sic Plautus: ego hunc hominem hodie texam pallio. Winter (Fr. Pl. v. 294) places the quotation among the fragments of unknown Plautine plays; but it is hardly conceivable that Servius’s citation is not genuine. Thilo and Hagen refer me to: illic homo hodie hoc denuo volt pallium detexere (Amph. 294) ‘yon man wants to purloin this cloak anew to-day’, giving me detexere the form parallel with depopulari. If I follow Servius, I return at once to the usual syntax with dum: while Halaesus is stripping Imaon of his arms, he exposes for a moment his own breast. In the context I find Halaesus engaged in a course of ruthless slaughter; there is no hint that he was likely to turn from it and expose his life to guard a fellow-warrior.

Texit has here a meaning opposite to its usual sense, just as: Rumone secundo (Aen. 8. 90) meant ‘up the Tiber’. But we have an example of the same verb once expressed giving us two opposite senses in:

Tum virgam capit: hac animas ille evocat Orco

Pallentes, alias sub Tartara tristia mittit,

Dat somnos adimitque, et lumina morte resignat (Aen. 4. 242-4).

The difficulty is in the meaning of resignat, which may mean either to duly seal, cf. recipere, or to unseal, cf. recludere. Mr. Jas. Henry tells me: ‘Virgil could hardly have chosen a more clear, proper, and forcible word to express the unclosing (unsealing) of the eyes of the sleeper, the metaphorically dead, than resignare’. I turn to Servius and read: resignat, claudit. I cannot but admire the assurance of Mr. Henry, when he tells me that he cannot entertain Servius’s interpretation for a moment. Who was Servius? and what qualifications had he for interpreting Virgil? But: lumina morte resignat 335 explains: adimit somnos; does it not also explain: dat somnos? Was Mercury more usually engaged in restoring the dead to life than in conducting the shades to Hades? In the last verse the second colon corresponds to the double idea in the first, and in resignat the prefix has two meanings related to: evocat animas and to: alias sub Tartara mittit. Morte is for either: in morte or: ex morte, and the meaning is: ‘he bestows slumbers and takes them away, and closes men’s eyes in death, and opens them to life’; opening and closing eyes in sleep constitute a pair.

Of course this is a very strong and interesting example of zeugma. Let us look at other examples of zeugma to which this ellipsis gives rise. In: quos dura premit custodia matrum (Ep. 1. 1. 22) premit from the mother’s standpoint seems for comprimit, from the boy’s for opprimit. In: quo res sponsore et quo causae teste tenentur (1. 16. 43) tenentur fits with res, but is for obtinentur with causae. In: Iliacas igitur classes atque ultima Teucrum iussa sequar (Aen. 4. 538) sequar fits with classes but is for exsequar with iussa. In: hinc omne principium, huc refer exitum (Od. 3. 6. 6) with hinc we must imply defer, used for deduce, from the refer expressed with huc. In: regna Neoptolemi referam versosque penates Idomenei (Aen. 11. 264) versosque seems short for: eversa eversosque, though the form versosque implies also his flight to Italy; for to go into exile is: vertere solum. In: te (latum) circum omnes alias irata puellas differet (Prop. 1. 4. 21), where te is for: tuum nomen, latum is to be implied from differet.

In: vim duram et vincula capto tende (Geo. 4. 400) tende with vim seems for contende, and with vincula for intende. In: luce demum postquam dux et miles et facta noscebantur (Ann. 1. 39. 8) noscebantur with dux et miles is for internoscebantur, and with facta for cognoscebantur, and in:

Romulus et Liber pater et cum Castore Pollux,

Post ingentia facta deorum in templa recepti,

Dum terras hominumque colunt genus, aspera bella

Componunt, agios adsignant, oppida condunt,

Ploravere (Ep. 2. 1. 5-9),

when we think of the comparison of these demigods with Augustus, who is still on earth, it seems right to take colunt with terras for incolunt, and with hominum genus for excolunt, translating: ‘while they dwelt on earth and refined the race of men’.

In: discedam, explebo numerum, reddarque tenebris (Aen. 6. 545) 336 our modern scholars follow Heyne’s view that explere means ‘to fill up’. They translate: explebo numerum ‘I will fill up the number’, i.e. ‘I will go back to the rest of the shades’. But Servius’s view is the opposite of this: explebo est minuam: nam ait Ennius: navibus explebant sese terrasque replebant (Ann. 561, M.) . . . sensus ergo est: minuam vestrum numerum et reddar tenebris ‘I shall withdraw from your number and be restored to the dark’. He adds that others take explebo here wrongly for complebo, as do our modern scholars. There is a third view, of which Servius knows: finiam tempus statutum purgationi et in corpus recurram (= reddar tenebris), for we have: clausae tenebris et carcere caeco (6. 734) of life in the body on earth. Servius prefers to take explebo as minuam, as that signifies Deiphobus’s compliance with the Sibyl’s wish for his departure. This would no doubt be the meaning Virgil would first wish to convey to his reader.

If so, we find in explebo the prefix ex conveying to the Roman reader the two opposite ideas conveyed by complebo and deplebo and to us by ‘fill up’ and ‘empty’, ideas which constitute a pair. The meaning found in Ennius’s explebant is also found in: exple animum eis teque hoc crimine expedi (Ter. Hec. 755, cf. vv. 785 and 787 and And. 188), to which Donatus’s note is: explere inanire, which brings exple into line with expedi. So Nonius (I. 471, M.): explere minuere. In Pliny N. H. 11. 19 the codices Vaticanus and Parisinus give: sunt et operis morbi, cum favos explent; and Hardouin seems right in preferring this to the non explent of the other MSS. as presenting a more difficult reading, and one certain to be changed by later scribes. In: haec tu cum istis tuis auctoribus excogitasti, ut vetera vectigalia venderetis, et expleretis nova (Leg. Agr. 2. 98. 36) Ennius’s sense of explere fits so well that the only escape from it is to emend to expilaretis. In:

Fleque meos casus: est quaedam flere voluptas;

Expletur lacrimis egeriturque dolor (Ov. Trist. 4. 3. 37-8),

expletur is best taken in this sense: ‘pain is relieved and removed by tears’. In: oblivioso levia Massico ciboria exple (Od. 2. 7. 22) ‘fill up the tankards and drain them’, Horace must have been delighted with the double meaning explere supplies; it is the rarer meaning of ‘empty’ that best suits with the nec parce of v. 20, and with funde (v. 22). The determination of our editors to give one meaning and one only to each word in Horace and Virgil is, I must assume, a feeling for the intellectual limitations of the primary classes in our schools.


Regarding the three interpretations of explebo presented by Servius, while I agree with him in accepting minuam as that to be taken by the average reader, and that for the reason he suggests—its agreement with the context—I feel that Virgil intended that the thoughtful reader should accept the third, which assumes a familiarity with vv. 723-751. In the verse: discedam, explebo numerum, reddarque tenebris, we have three cola, which should offer three distinct meanings. If we take explebo as deplebo, the archaic use, we have a repetition of discedam. If we take it as complebo numerum, we have a repetition of it in: reddar tenebris. True, it is rather more closely connected with discedam in the scansion of the verse, and that is the rarer meaning. But probably explebo numerum reddarque tenebris is a poetical inversion for reddar tenebris et explebo numerum: ‘I will return to the shades and fulfil my purification for another life on earth’; and Virgil hopes that after perusing vv. 723-751 his thoughtful reader will see this, and restore to the verse its due order in thought. The threefold use seems parallel to that of superum (v. 780).

How does the verb expleo come to have the meaning properly given by compleo or repleo? for replebant in Ennius’s verse cited above has the meaning of complebant. Of the two uses of expleo, that equivalent to depleo ‘I empty’ seems the older and more genuine, when we compare ex here with that in: exeo, expendo, emitto, expugno, eripio. In: navibus explebant sese terrasque replebant (Enn. Ann. 561, M.) we have the union: explere et replere ‘to empty and fill again’, which imply complere ‘to fill completely’, a meaning usually conveyed by the second verb replere, but also later by the first verb explere. So the meaning ‘to fill up’ for explere seems the result of synecdoche. Replere too is primarily and properly ‘to fill again’, as we see in: veteremque exire cruorem passa, replet sucis (Ov. Met. 7. 287). The simple verb pleo i