cases its readings are those which if originally received into the text of Horace by Cruquius or others on its authority have proved to be also those of the MSS. which modern criticism has ranked highest-this proves it to have been a good MS.-but also in the fact that it had several readings which have every appearance of genuineness and yet of which there is no trace in any other MS. except in g, a Gotha MS. of the 15th century. Such is the famous 'campum lusumque trigonem' in Sat. 1. 6. 126, where every other MS. but g reads the incredible 'ratiosi tempora signi.' In other words it gives us access to a tradition of the text independent of that of the other MSS. That it was not free from interpolation, especially in the Odes, is allowed by those who rate it most highly; see e.g. Od. 4. 2. 6, where it reads 'cum.. saliere,' 4. 6. 21 'Alexus.'

2. The oldest Bernese MS.' (363 in the Public Library) was first used by Orelli in forming his text, and has since been recollated by Ritter for his edition, and by Usener for Keller and Holder. It is assigned by Ritter and by Usener to the 9th century. It forms part of a quarto volume, which contains also Servius' Commentary on Virgil, two Treatises on Rhetoric, Bede's History, and Ovid's Metamorphoses. It is imperfect, omitting all the Epistles, except the Ars Poetica, and the Satires, with the exception of the first two and part of the third of Book I, besides large portions of the Odes. The Odes are not arranged in their usual order, the copyist having apparently started with the intention of an

1 An interesting account of the history of this and of several other of the older MSS. of Horace is given by Ritter in the Prolegomena to his edition. This MS. was originally in the Abbey of St. Benoit-sur-Loire, at Fleury near Orleans. When the abbey was sacked by the Protestants in the Civil War in 1562, the MSS., in which it was rich, were saved, and found a home in the library of Pierre Daniel, an 'avocat' and literary man of Orleans, and bailli' of the abbey. At his death his books were divided between two friends and fellow-townsmen, Paul Petau and Jaques Bongars. The latter of the two died at Paris in 1612, and left his library, including his share of the Fleury MSS., to René de Graviset, a jeweller of Strasburg. De Gravisct migrated subsequently to Switzerland, and his son became a leading citizen of Bern, founded a public library there, and placed in it his books, amongst them this MS.

arrangement according to metres, for he begins with nineteen Sapphic Odes. These follow the common order, though the distinction of Books is not marked. They are succeeded by some of the Epodes, then by the Carm. Sec., then by the remaining Odes and Epodes, also in their usual order. The Ars Poet. 1-440 follows the Epodes, then Sat. 1. 1, 2 and part of 3. The omissions in the Odes and Epodes are as follows:-of whole Odes in Book I, Odes 3-7, 9, 11, 33 and 34; in Book III, Ode 3; and of Book IV, Odes 3 and 15, besides parts of twenty-one more Odes and Epodes, viz. Odes 1. 10. 14, I. 15. 20-32, 1. 16. 15–28, 1. 17. 15, 16, 1. 19. 11-13 and 15, 1. 29. 7–16, 2. 7. 19-28, 2. 17. 7-9, 3. 2. 2, 5-12, 17-32, 3. 4. 17-28, 39-52, 3. 6. 11-13, 15-48, 3. 16. 7-27, 29-44, 3. 22. 5-8, 3. 23. 12-20, 3. 24. 30-64, 4. 14. 5-52; Epodes 2. 37-70, 3. 9-22, 9. 13-38, 11. 13-28.

3. Of the 10th century the following MSS. have been collated for vol. i. of Keller's edition :

Paris, 7900. This has been held by Otto Jahn and others to belong to the century before. With it Holder closely (a) connects a MS. known as a, formerly belonging to Avignon, now in the Ambrosian Library at Milan (No. 136), which he has collated for the Satires and Epistles. I had the opportunity of inspecting it for a few hours, and have taken account of its readings in some of the more disputed passages in the Odes.

Paris, 7971, like the old Bernese, a relic of the Fleury Library.

Paris, 7974. The common readings of this and the last MS. are marked by Keller and Holder by the letter F.

Paris, 7972. This is one of the MSS. which contains the Mavortian inscription; see next page. With it Holder (1) connects the Leyden MS., which was one of which Bentley made use. This is marked /.

Paris, 7973, belonged, like B and y, to Pierre Daniel. b

To these Keller and Holder add a l'aris MS. (7975) of the 11th century, to which they both attribute high value (y).

Of the 10th century also is the MS. in Queen's College, Oxford, a collation of which was given at the end of this volume in the first and second editions (Regin.).

It has been already suggested, that the only method by which conclusive evidence upon questions of disputed reading could be gathered from MSS. of such late date would be by grouping them in families or 'recensions,' each family being made by its common readings to bear witness to some archetype of much greater antiquity than the copies which we now possess. This has been attempted by Keller and Holder'. A and a, λ and 1, and y, as we have seen, have been supposed to be severally related in this way, and the readings of their imagined archetypes are indicated by Holder by the signs A', X', F. With his fuller 'apparatus criticus' on the Satires and Epistles, the process is carried still further, and the result is, to the extent that uniform readings can be obtained, a certain number of conjecturally restored MSS. of the 7th and 8th centuries. These and the other solitary representatives of older MSS. are again grouped into three classes.

'Their general conclusions were given to the world first, as regards Keller, who is responsible for the Odes and Epodes, in the Rhein. Mus. vol. xix, p. 211-227; as regards Holder, who undertakes the Satires and Epistles, in the Preface to vol. ii. They have been set out in full detail, and with some reconsideration in particular points, by Keller, who published the Epilegomena to this edition in 1879, 1880.

The first of these is distinguished as containing, with many faults of carelessness, and with a certain number of grammatical corrections (e.g. 'videri,' against the metre in Epod. 16. 14), little or no proof of alteration on rhetorical or general grounds. In this class Holder places, of the MSS. named above, A, a (in respect of most readings), D, y, and, though not on an equality with these, T.

The second class is supposed to show the corrections of an early and intelligent emender. That such dinpowraí existed in carly times is stated by the Pseudo-Acron on Ars Poct. 345. And the name of one is found in the inscription which appears, in slightly different terms, at the end of the Epodes in A, 7, Reg, g. 'Vettius Agorius Basilius Mavortius V. C. et inl. (vir clarissimus et inlustris) Excom. Dom. excons. ord. (Excomite Domestico, exconsule ordinario) legi et ut potui emendavi, conferente mihi magistro Felice oratore urbis Romae.' Felix is not known, but Mavortius was consul in the year A.D. 527. Asterius, consul 494, is similarly connected with the history of the text of Virgil. Bentley had treated this Mavortian recension as the ultimate point to which the oldest MSS. might be expected to take us back. Holder treats it rather as the first, though not the worst, source of systematic corruption. In the class which he thinks owe their peculiarities to this editing, he places B, V, with g, and in respect of some of their readings, A, a, and λ, the first two of these being in respect of much of their text affiliated to class one, the third to class three.

The third class (in which rank F (=$↓), λ'( −λl,) u, π, σ, and others) is held to be of less value, both as having been derived, in the first instance, from more faulty originals, and as bearing signs of later and less intelligent correction.

It may be observed that the tie which is imagined between the MSS. which are classed together is confessedly a loose one. Room is left for considerable influence of MSS. of one class upon those of another. As a fact, there are very few cases indeed of doubtful reading where the division of testimony corresponds at all exactly to the three classes. Doubtless there

are a few crucial passages where the value which is set on such MSS. as V or B goes far to decide the choice of a reading, and there must be more cases still where the instinctive feeling of the trustworthiness of particular MSS. on particular points which comes from long intercourse with them, but which no 'apparatus criticus' will convey to a reader, has guided Keller or Holder to a conclusion which the mere inspection of the list of MSS. on one side and the other hardly explains. But a survey of the places where Keller's readings of the Odes and Epodes differ from those of other recent editors will show, I think, that his theory of the relation of the MSS. has not had any very revolutionary effect upon his text, that he admits conjectures at least as largely as others, and that his conclusions, however they were formed, are supported generally by evidence from the Scholia or quotations, or by considerations of style, metre, and the like'.

Before we leave the MSS. it may be in place to say a few words upon two subjects which present themselves very early to the reader of any notes upon Horace.

1. The first has been glanced at already. I do not know that conjectural emendation has really been exceptionally busy

'Deducting questions of orthography, punctuation, etc., there are thirty-seven places in the Odes and Epodes where Keller's text differs from that of both Ritter and Mr. Munro. Of these nine are due to his introduction into the text of conjectures Od. 1. 7. 8 honore,' 1. 15. 30 * Pergameas,' 1. 16. 5 ‘adyti,' 1. 23. 5. 6 * vepris . . ad ventum,' 3. 5. 15 ‘trahenti,' 4. 4. 17 'Ractis,' 4. 10. 5 ‘Ligurine,' Epod. 2. 27 'frondes,' g. 87 humana invicem). Eleven are cases where readings resting on very slight MS. authority, or even upon none, have been received on other grounds, such as the authority of the Scholia, etc. Keller, Od. 1. 3- 37 ‘arduum,' 1. 16. 8 ‘si,' 1. 20. 10 ‘tum,' 1. 22. 2 ‘Mauri,' 1. 22. 11 * expeditus.' 2. 3. 11 ‘quo et,' 2. 6. 19 'fertilis,' 3. 3. 55 ' debacchantur '; Kitter and Munro, 1. 17. 14 ‘hic,' 2. 13. 23 ‘discretas,' 3. 4. 10 ‘altricis '». There are very few of the remainder where the effect of his view of the MSS. can be distinctly detected in the absence of other arguments from the Scholia, or from internal evidence; such are perhaps 1. 8. 2 hoc,' 3. 21. 10 necgleget,' 4. 9. 31 'sileri,' Epod. 16. 33 'flavos,' 17. 60 *proderit.'

upon Horace's text. That the two are specially associated in the minds of general readers is due doubtless, in a great measure, to their greater familiarity with the author, to the bril liancy of the conjectures themselves, the contributions of a long series of the greatest scholars from the Renaissance onwards, particularly perhaps to the unrivalled power, learning, and cloquence with which our greatest English scholar recommended the method and its results in his edition of the poct. There was something however in the nature of the critical evidence on which Horace's text rests which made conjecturai emendation, if not specially necessary, at least specially tempting. Necessary of course it was not in the sense in which it is necessary in the text of Aeschylus or of Lucretius, to restore sense or metre in a chaos; but in a way the multiplicity of MSS. tempts us to do for one author what the poverty of MSS. almost compels us to do for another. A variety of readings, all consistent with metre and intelligible, and all resting on fairly equal MS. testimony, must imply the hand of one or more emenders of the text at an early period. It is a natural interpretation to assume in such a case that all alike are attempts, more or less skilful, to fill up a gap in the original authority; and this once believed, a scholar of the 16th or 17th century may not unreasonably think himself as competent to guess the riddle as a scholar of the 4th or 5th. It is manifest that no impassable barrier separates cases where the MSS. are divided from those where they are consentient. Division is only a sign of the disease. We have already seen that it is confessedly possible for the same blunder to infest every MS. A modern editor will probably set aside, as a rule, purely conjectural emendations; at any rate, he will hesitate to give them the reality which is implied by printing them in the text; not because he denies the possibility of corruptions, or does not feel the plausibility of many conjectures, but only because experience has taught us that there is no necessary limit even to the cleverest and most piausible guessing, and because it cannot be proved that in such a text as that of Horace guessing on a large scale is necessary. One more remark may be