^{page 200 note 1} Two examples, taken at random from Book I, will show what I mean. At 328 d, 5 Mr. Adam, with most recent editors, writes (so A) with no hint of a variant in his critical note. In his commentary he says: ‘There is no sufficient reason for reading νεαν⋯σκοις with Π and other MSS.,’ which implies that the one word is as good as the other, and that the use of νεαν⋯αις is quite normal. As it is, however, we require some proof that the word νεαν⋯ας was ever used by any Attic writer simply for ‘young man.’ So far as I know, it never occurs except with the shade of meaning which comes out explicitly in νεανικ⋯ς and νεανιε⋯ομαι. That being so, it is surely important for us to know that, in this case, Π(D) is confirmed by two quite independent witneses, F and Stobaeus. When Π(D) goes over to the enemy like this, the easiest assumption surely is that the other reading is due solely to the archetype of AM. The same thing applies to the omission of γε after κα⋯ μ⋯ν χα⋯ρω two lines below. Mr. Adam says only : ‘γε need not be added (with Π and other MSS.) after χα⋯ρω.’ Certainly not, if we look merely at what is ‘possible’ Greek ; but, if we look at the evidence, it certainly must. For here again Π(D) goes over to the enemy, Stobaeus having γε, and F having τε, which means that its original had γε. Is it likely that νεαν⋯αις was changed to νεαν⋯σκοις by at least three different persons at very different dates, or that the idiomatic γε was similarly interpolated at a time when the Attic κα⋯μ⋯ν…γε had long been obsolete (cf. Lucian, Lexiph. c. 10)? Surely it makes less demand upon our faith to believe that the archetype of AM is alone responsible for readings of which there is no trace in any MS. independent of it.

^{page 200 note 2} Bessarion's codex (Ξ) is really a recension and draws upon many sources of information (Ο. Immisch, Philol. Stud. p. 13, n. 3), but it is not probable that he had access to any tradition unknown to us. The argument from the supposed agreement of certain peculiar readings of Ξ in the Phaedo with the Petrie papyrus is quite illusory. When the papyrus was published, Ven. T and Vind. W had not been collated in the Phaedo, and it now appears that these readings are not peculiar to Ξ, but come from the regular tradition. Ven. T is also the chief source of Ξ in the Republic as far as 389 d, where the old part of the MS. comes to an end, and the text of T, though not directly derived from A, as will be shown later, is practically the same. From 389d onwards, the groundwork of the text of Ξ is derived from a MS. akin to Vind. F, as will appear. The relation of q, through certain intermediaries, to Π is still clearer, and it is far more probable that its peculiar readings are all due to conjecture than that they represent an otherwise unknown tradition. I say nothing of Ang. v, a MS. of the sixteenth century, which Mr. Adam found most ‘useful’ next to ΠΞq; for he has already admitted that Vind. F should have been quoted instead of it.

^{page 201 note 1} In the Introduction to his text of 1897 (p. x, n. 1), Mr. Adam gives a list of twenty-one places where Ξ and q must, in his opinion, represent a tradition ‘independent of both A and Π.’ In the great majority of these cases they certainly do, but we know the independent tradition from other sources not mentioned by Mr. Adam. Some of them are in F (e.g. 388 e ⋯φ⋯η, 532a περα⋯νει, 522d δι⋯λθωμεν, 604c α), some are in M (e.g. 425d λ⋯ξεως), some are in the testimonia (e.g. 540c διαφ⋯ρονσα Aristides, 544c διαφ⋯ρουσα Stobaeus, 590e βο⋯λεται Iamblichus, Stobaeus). If we deduct these and also the places where the reading of Ξq is not universally accepted, we are practically left with the following: 407c τιν⋯ς for τινος Ξq, 411d γευ⋯μενον for γευομ⋯νου q, 431b οὖ for οὔνΞ, 580d om. λογιστικ⋯ν Ξ, 606c ⋯ν⋯ης for ἂν εῐης q. We cannot think so meanly of Plethon and Bessarion as to suppose them incapable of making these collections, and, till further evidence is forthcoming, that is the hypothesis on which we should proceed.

^{page 201 note 2} Here it becomes very important to notice that the oldest and most numerous marginal corrections in A were written by the same hand that wrote the text and scholia. This can be shown from the way in which the minuscule hand of the corrections and supplements is, in certain places, made to alternate with the semi-uncial writing of the scholia, and especially by a comparison of Par. 1962, a MS. of Maximus Tyrius written by the same hand (cf. T. W. Allen in Journal of Philology, vol. xxi). These readings, then, represent the deliberate intention of the scribe, and should not be confused under the common symbol A² with corrections of a later date. They form a valuable collection of instances of the sort of error to which A was liable, and M is useful as showing that many of these were made by himself, and do not come from the common source of AM.

^{page 201 note 3} This can be shown of Π(D) by such errors as the following : 394c, 1 δ δ⋯ for λλη; 398d, 5 διδομ⋯νου for αἰδομ⋯νου; 399c, 5 ν⋯ν δν for ν⋯ν δ⋯; 401a, 7 κακονο⋯ας for κακοηθε⋯ας; 401c, 8 λ⋯ρα for αὕρα; 559d, 10 πουσ⋯ου for που οἴου; 581a, 10 δεῖ for ⋯ε⋯; 581e, 1 δπηι ⋯σχει for δπηι ⋯χει. I have proved the same thing for F (C.R. vol. xvi. 98 sq., xvii. 12 sq.) to the satisfaction of Professor Hermann Diels (Deutsche Litteraturzeitung, Dec. 5, 1903, p. 3006), and need not repeat my argument here. It is to be observed that we cau only get a direct proof of the value of a MS. by studying its errors. The occurrence of correct readings proves nothing; for they may have got into the MS. in all sorts of ways,— from marginal notes, scholia, commentaries, and mere conjecture. Errors, on the other hand, can often be made to tell their own tale.

^{page 201 note 4} It is important to remember that the MS. of Proclus's Commentary on the Republic was written by the same scribe as Par. A and must, therefore, be credited with similar excellences. We have also a MS. of Eusebius written, like the Clarkianus, for Arethas, though by a different scribe. These at least can hardly have been interpolated from inferior MSS. of Plato, nor is such interpolation at all probable in the case of Iamblichus or Stobaeus.

^{page 202 note 1} Phd. 71c . Symp. 214c . Riddell, Dig. § 288.

^{page 202 note 2} Only one MS. on my view, but Mr. Adam does not, I understand, admit the affiliation of Par. DK, Mon. q, Flor. β, and Vind, D to Ven. Π (=D), and he has, therefore, a more difficult problem to solve.

^{page 203 note 1} Cf. the following omissions in A : 358a (17 letters), 360a (17 letters), 364a τε κα⋯ δικαιοσ⋯νη (15 letters), 366a (14 letters), 373a (15 letters). These all occur in one part of the Republic, but the same remark will be found to apply elsewhere. It appears that the archetype of A and Ven. T was written in double columns like those MSS. themselves, while Π, like the Clarkianus, goes back to an original in which the line went right across the page and contained about 45 letters.

^{page 203 note 2} Cf. O. Schwab, Hist. Synt. der griechischen Comparation (Schanz, Beitr. Heft 11), pp. 23 sqq. and pp. 96–99.

^{page 203 note 3} Schneider (ad loc.), speaking of Bremi and Faesi, ‘qui, si pro participiis optativos non cum particula εἰ, sed cum relativo δς supposuissent: (sic) εἰδ⋯ναι, minus, credo, offendissent in mutations structurae pene legitima.’ It is true that εἰ μ⋯ εἰδε⋯η has been allowed to stand in Jowett's note ad loc., but the official defence of the text of J. and C. must be taken to be Campbell's Essay on Syntax, § 56, where we read