in the same collection in the 17th century affords the faintest presumption of a like connexion in the 10th, and in any case they are far enough removed from the New College MS. However, speculations of this kind furnish far less certain data than the analysis of the text, which we hope to prosecute with as little delay as possible. In order of merit there appears to be little to choose between O, and O,, or between O, and O, О, but there is a distinct interval between the two groups.

The Treatises. We have collated in full the Oxford MSS. with Hartel's text of the Ad Fortunatum, which they will help us to correct in several places, though Hartel has been more successful here than in the Testimonia. The general tendency of the Oxford MSS. is to agreement with W R V, edd. Among them O1 appears to be the best. In De Eccl. Unit. c. 4, O, alone, with n,, has escaped interpolation: O, and O, O, admit the shorter interpolations in M, and O, the greater part of the longer interpolation. In spite of this, O, has one or two readings of considerable interest in the earlier chapters, and appears to deserve closer examination. O̟, and O, have a peculiar text of some interest. On the whole it does not seem likely that the Oxford MSS. will reach the same degree of excellence in the other treatises as in the Testimonia. The different elements in them will need to be carefully separated.

2

4 5

2

The Epistles. Here O, and O̟ O̟, are the most important as containing the largest collections. The value of O, is, however, discounted at the outset by its exact agreement in the order of the letters with T; and there can be no doubt that it stands in close relation to that MS., though what is the precise nature of the relation still remains to be ascertained. It was natural to turn first to the group of letters, found in T alone of the older MSS., Nos. 8, 21-24, 27, 33-36, 41, 42, which includes, besides letters of Cyprian himself, others written not by, but to Cyprian, especially Epp. 21, 22, the correspondence between the African confessors, Celerinus and Lucianus, which bears more than anything of Cyprian's, the genuine traces of vernacular speech. Unfortunately both here and in other members of the same group, O, appears to contribute nothing of any value. It is also tantalising to find that the same epistles, 21 and 22, which the catalogue would lead us to suppose were contained in O̟, are not really there. The long letter of Firmilian (No. 75), which is contained in that MS., presents a worthless text. It would, however, be a mistake to infer that the text of all the epistles was of the same quality. They are broken up in the MS. into groups by the intrusion of treatises, and it will probably be found that each group stands by itself. The first and the last only have been tested; and though the last, as we have seen, is practically worthless, the first has every appearance of being of a very different

S

4

5

order. It is too soon to speak with confidence, but a single example may be given of a reading that is certainly of a kind to attract attention. In Ep. iv. § 1 (p. 473, ll. 7-10) Hartel's text has 'quominus fratribus et sororibus nostris constanter et fortiter consulatur et per omnes utilitatis et salutis uias ecclesiastica disciplina seruetur. For this O, and O, have 'quominus . . . consulatur per omnes utilitates et salutes ecclesiasticae disciplinae. No other MS. has precisely the same reading, but there are a number of approximations towards it: BLW omit the first 'et'; B has 'utilitates'; BQ W omit 'uias'; W reads 'ecclesiasticae'; and B W' disciplinae'; the same MSS. omit 'seruetur.' It is clear that Hartel's text and the reading of O are at opposite ends of the scale; either the one or the other represents the final result of a lengthened process of corruption, and the question is, on which side is the corruption? It is not only the character of a single MS., but of a class of MSS., that is at stake, and a comparative study of other data is desirable before coming to a decision. And yet there is surely a strong prima facie case in favour of O. Suppose that it has preserved the original reading, and it is easy to see how by a series of easy and natural changes it might pass into the reading of Hartel's text. The expression per omnes utilitates et salutes,' though very possible, is rather unusual: it might naturally be paraphrased by salutis uias,' which would lie especially near at hand if, by a common chance, the 'e' of 'salutes' had got converted into an 'i': then the genitive 'ecclesiasticae disciplinae' hangs in mid air, and must needs receive the obvious correction ecclesiastica disciplina seruetur'; and after this change, or simultaneously with it, it was a small thing to add the first 'et': so, by degrees, the abrupt edges are filed down and polished to their present smoothness. On the other hand, the converse process does not seem to be nearly so natural: it involves a double omission, both of 'uias' and of 'seruetur'; yet if either of these words had been left out it would seem more obvious to replace it by conjecture than to go on to the series of other changes which end by leaving the condensed form of text as O, has it. It would be a help to the future criticism of Cyprian if other scholars would give an opinion on this point.

In another aspect the large number of letters contained in O, is important. It will contribute data to a problem of much interest, viz. that which is raised by the varying, and yet in different degrees recurrent order of the epistles as we have them collected in the different MSS. It must be possible by the comparison of these to work back through the archetypes of existing MSS. to the primitive groups of Cyprian's letters, and determine in what form the collections of them circulated at a date not far removed from their original composition. A problem of the same kind, though simpler, is presented by the order, also varying, and also to some extent recurrent, assumed by the treatises. Some steps have

already been taken for working at these allied problems, not without promise of results.

Almost as I write there comes to my hands, through the kindness of Professor Mommsen, a notice, which will, I believe, appear in the next number of Hermes, of a discovery made by him in the Phillipps library at Cheltenham last autumn, which is full of interest and suggestiveness for the student of Cyprian. This is nothing less than a list of the writings of Cyprian with the contents of each in oriXO, reckoned after the measure of the Vergilian hexameter of 16 syllables, and dated by a scribe's note in the consulship of Eusebius and Hypatius=359 A. D., or just a century after Cyprian's death. The list occurs amongst a quantity of miscellaneous matter, and follows immediately a similar list, also with stichometry, of the canonical books of the Old and New Testaments. The MS. is numbered 12266 in the Phillipps collection. It had been described by Zangemeister (Durchforschung d. Bibliotheken England's, p. 99, Wien, 1877), who had however failed to notice the list of the writings of Cyprian. Many questions are raised by this discovery, which I hope soon to have an opportunity of discussing. The point that has the most direct bearing upon our present enquiry is a suspicion of large interpolation which it seems to cast upon Book III of the Testimonia. Comparing, as Professor Mommsen has done, the number of orixo assigned to each book, with the number of lines in Hartel's edition, the estimates for Books I and II correspond nearly enough, but for Book III there is a wide discrepancy.

It is true that Hartel has admitted within brackets (pp. 134-138) a long insertion from W, which is wanting both in the rest of his MSS. and in all that we have examined. This would account for 108 lines; and the MSS. testify to other smaller interpolations, but nothing that would make up a difference of more than 1000 lines. It is true also that a process of which there are still traces in the MSS. might not unnaturally be thought liable to be extended. And yet I hesitate to believe that there has been any such serious interpolation. If there had been, it must have been almost infallibly detected by a difference in the Biblical text. But there is no difference. Book III, like the rest, so far as we have examined it, is strictly homogeneous; and it has all the features of the Cyprianic text elsewhere. This will appear from our analysis of a number of readings taken from the book in pp. xlv-lxii of the Introduction. It therefore seems to be by far the simpler hypothesis to suppose that I or M has dropped out from the sticho

metric notation. The stichometry of the books of the New Testament must be corrupt in several places.

There is a passage in Jerome which has a bearing upon the point. Writing against the Pelagians, Jerome refers to the claim of Pelagius to 'imitate or rather to supplement the work of the blessed martyr Cyprian.' In doing so, Jerome says, he does not see that he contradicts himself, and he proceeds to quote in full the three texts of the 54th head in Book III, appealing to it by name: 'Ille in quinquagesimo quarto titulo tertii libri, ponit neminem sine sorde et sine peccato esse, statimque jungit testimonia,' &c. (Dial. c. Pelag., i. 32). This is precisely as it stands in our present MSS. It might at the first glance be thought that Pelagius had actually interpolated the Testimonia, and that some of his interpolations had come down to us. If that were the case, they would be easily separable, because Pelagius would most probably use an Italian text, as he appears to have done in his Commentary on Romans. But in any case, the evidence of Jerome is proof that the number or 'tituli,' at least up to No. 54, was the same in his day that it is now; and the particular 'titulus' has come down to us precisely as he read it. But this titulus' is nearly two-thirds of the way through the book, and has already passed the 1000th line in Hartel's edition (not counting preface or 'capitula'). At the most, therefore, if there has been any interpolation, it did not consist in adding new tituli,' but only in adding more texts under old tituli'; and for the reason given above it seems improbable that even this was done to any considerable

extent.

These remarks are not made with any idea that they close the question. On the contrary, they are intended rather to open it. It is sufficiently evident on all sides that much still remains to be done for the criticism of Cyprian1.

1 A paper on the Cheltenham list was read before the Oxford Society for Biblical Archaeology and Criticism on Feb. 22nd. The order of the contents of Hartel's and the Oxford MSS. was tabulated and compared with that of the list, and some interesting relations seemed to be suggested,

APPENDIX III.

The Relation of n to Ambrose: renderings of тà vпáруоνTA.

I had for some time discovered the inaccuracy of the indices in Ballerini's edition of Ambrose, but it was not until too late that I became aware of their utter incompleteness. Not above a tithe of the real references are given; not above a tithe even of those which the editor himself has noted in the margin. If these had been transferred to the index and allowed to take up the space which has been occupied by printing the passages referred to at length, or partly at length, the utility of the edition would have been largely increased. Fresh indices have now been made for me, and I have been enabled to test more closely the relation of n to the text used by Ambrose, about which the fact mentioned on p. ccxxviii was calculated to excite curiosity. If the text of ʼn stands in a close relation to that of a2, and the text of a, presents a remarkable coincidence with that of Ambrose, it is natural to ask if there are many like coincidences between Ambrose and n.

I cannot claim to have examined closely more than the readings that we have seen reason to regard as specially characteristic of a n. A large harvest of references soon dwindles down when it comes to be applied to a particular point like this. The instances in which characteristic readings of a n or of n are capable of comparison with quotations in Ambrose are as follows. The volume and page of Ballerini's edition are given up to vol. v.; the concluding volume I do not possess.