weakened, but it ought not to be allowed to destroy wholly this authority. Bechert, who had both G and M before him, very greatly overrated the comparative worth of the older MS. He is throughout blind to the patent weaknesses of G, and to all the virtues, whether patent or latent, of M. Mr. Housman, on the other hand, ever ready to find good in all MSS. and evil in all editors, will not allow that either MS. is one whit superior to the other. He stands judicially between two bundles of hay, and pronounces them both equally palatable (the figure is his own).1 M is as good as G and G as M. Very similar is the position of Breiter. But a violent reaction against G was bound to come before long. And recently a German scholar, P. Thielscher, has put forward the opinion that G has no independent value whatever. He wishes to banish its readings for ever from any Apparatus to Manilius. He regards it as nothing more than an interpolated copy of L. Though I notice with surprise that Thielscher seems already to have converted Kraemer,3 I do not fancy that his attack upon G will be regarded as successful by the majority of Manilian students. One important merit may be conceded to Thielscher. He has done good service in emphasizing the strength and sincerity of L. But his onslaught upon G is, I think, mis

calculated.

Mr. Housman, on p. xxv of his edition of Book I, has compiled a list of passages where the reading of G is clearly superior to that of the other MSS. This list might be longer than it is; but I will content myself by selecting a few examples from it :II. 15 Iouis et G: iuuisse LM. 19 notauit G: rogauit LM. 168 exterius mirantur GĽ2: exterminantur LM. 473 generant G: gerant LM. 584 lis GL2: leuis GLM. sortem LM. IV. 30 captis G: capitis LM. multum LM.

III. 69 sorte G:

221 Multo GL2:

243 Vesta tuos GL2: uastat uos L: restat uos M. 282 illuc agilem G: huc caliginem LM (caligine M). V. 46 ortus G: portus Z: portur M.

In all these places G, sometimes supported by the corrector of L, has the true reading where LM have a false reading. Now where LM agree in error, especially in senseless and unmetrical

1 Praef., p. xxxi.

2 Philol., 1907.

3 Bursian's Jahresbericht, 1908, pp. 234-51, Abt. ii.

4 Now and again L2 alone among our codices has the true reading: e. g. in Book II. 533 tergore cedit, 600 populo.

error, as e. g. IV. 282 huc caliginem, this agreement can only be explained by supposing that their reading is that of the archetype of all our MSS., A. Whence then has G acquired in all these places-and I could multiply the number of them infinitely-the true reading? In some cases no doubt its true reading may be due to conjecture-for it is, as Thielscher perceives, a deeply interpolated MS. But in many cases it is impossible to explain these true readings as conjectures. And the only other explanation is that where G alone has the true reading it has somehow or other gone behind A, behind the archetype of GLM. When at IV. 282 LM have huc caligine(m), the reading of A was undoubtedly huc caligine; and huc caliginē must also have been the reading of the common parent (a) of GL. The many places where G diverges from A and a, therefore, and offers instead a true reading, can only be explained by assuming that G, or the MS. from which it is immediately copied, has been corrected from some purer and probably older source than A. And that this has happened is most clearly seen when we observe that in many of its true readings G is supported by Z (as II. 584, IV. 221 in the readings given above). G and L, the former in countless places, the latter in a few, have been corrected from a MS. of Manilius better than any known to us.1

1 Thielscher's position is based upon two assumptions: (1) that L was written before G: but here the evidence does no more than at most to show that L is probably not later than G; (2) that, where G has in its text the readings of the corrector of L, it depends for those readings actually upon L. At the best this could be only a likely presumption. But this presumption is rendered very weak indeed by (1) the number of true readings peculiar to G; (2) the agreements between G and the Codex Venetus, of which I give an account in Introduction, II, pp. xlvii sqq. How will Thielscher explain the fact that, to take one example, at 820 G Ven. have cordibus, L has torridus? (3) by the frequent disagreement of G and the corrector of L. It is perhaps worth while to note here some places where G does not offer the correction which stands in L. I. 130 *sumptum L3 : summum GL'M. 140 *creentur L2: creantur GL1M. 163 strixerunt L2: *struxerunt GL'M. 265 nocantem L2 (uoluit nocentem): *uocantem GL1M. 365 *hunc L2: tunc GL'M. 470 *conditur L'M: ceditur GL1. 514 lustrarat L2: lustraret L': *lustrarit G: lustrari M. 520 *puncto L2M: ponto GL1. 582 cingens L2: tingens L1: tangens G: *timens M. 616 *uestigia L'M: fastigia GL1. 712 *findens L'M: fingens GL1. 726 fissuram L3: *fusuram

*The asterisk denotes in each case the true reading or the reading nearest to truth.

And now we can see at once both how much and how little justice there is in Thielscher's onslaught upon G. G, in so far as it is a descendant of a, is, as Thielscher accounts it, a good deal inferior to L. L is the sincerer, less interpolated, representative of a. G is in itself an inferior Z; but it is an inferior Z which has been worked through (or its parent had, more probably) by a fool with a better MS. of Manilius than the world will ever see again-till Herculaneum gives up its dead. I say a 'fool' worked through G with such a MS.; for the sparsity of the corrections which he has introduced from it argues an extreme tenuity of intelligence. Yet we owe this fool, even so, a deep debt. If it is levity to say with Housman that G and M are equal', it is equal levity to say with Thielscher that G is a MS. which may be altogether set aside. This much, however, we may say: G's blunders deserve no attention. M's blunders are infinitely important. Its strangest corruptions are again and again nearer the truth than the specious correctness of the other MSS. But G where it is wrong is useless. When its readings give sense we have, if they differ materially from those of M, to consider whether their appearance of correctness is the result of interpolation or of 'correction' in the true sense. But the examples of true correction from a pure and ancient source are so many that G remains an indispensable guide in the construction of any sane text. I may sum up the relative merits of G and M by saying that they are two MSS., of which the one, M, is good all through and bad all through, the other, G, is bad all through but good in places.

GL1M1: fixuram M2. 756 contexit L'M: *conuexit GL1. 776 *que L2: qui GL'M. 802 *candet L2: candit GL1M. 836 *capillos L'M: capillis L1: capillus G. 843 utero L2: *uteros GL1M. 844 *paruos L'M: paruis GL1. 863 *cum L'M: ne GL1. 882 labentesq. L: *labentisque GL1M. 884 erictonios L2: erectecos L'M: ericteos G.

I will not pursue examples of the same phenomenon through the other books. Nor must I be supposed to think that these or other examples prove that G is not derived from L. They prove nothing, but, in conjunction with other facts, they materially weaken the presumption of G's dependence upon L (which is, after all, a mere presumption). It is particularly difficult to conceive how G, which, according to Thielscher, contains so many true readings reached by conjecture, should so often reject the true readings offered to it by the corrector of its original. Nor can I explain, upon Thielscher's hypothesis, the number of places where GM agree against L : e. g. in III. 1-400 alone, at 3, 45, 70, 73, 213, 236, 265, 295, 333, 360, 367, 377, 387.

APPENDIX I

NOTE ON THE MADRID MS.

The object of this Note is twofold. I desire firstly to notice a difficulty-to which proper attention has not, I think, been called-in connexion with the received dogma of the Poggian origin of M. Secondly, I wish to give prominence to a recent and important discovery made by P. Thielscher in his examination of M.

M (or M 31, to give it its full title) has to be considered in connexion with another MS. in the Madrid National Library, X 81. M 31 contains the Siluae of Statius and the Astronomica of Manilius. X 81 contains Valerius Flaccus, Asconius, and Sigebert's Chronicon. I will call M 31 by its usual designation M. I will call X 81, so far as it contains Valerius and Asconius, P— following Clark in his edition of Asconius. That portion of it which contains the Chronicon I will call π.

Clark regards P as a transcript made by Poggio himself of the MS. (or MSS.) of Valerius and Asconius which he unearthed at St. Gall in 1416 (Epp., Tonelli, pp. 25–9). π, again, he holds to be Poggio's copy, in his own hand, of a MS. of the Chronicon which he found some years later in Britain. M, lastly, is, according to Clark, a copy made for Poggio by a German scribe at Constance of the MS. (or MSS.) of Statius' Siluae and Manilius discovered in, or near, Constance about the same time as Valerius and Asconius. It is, in fact, the very copy spoken of by Poggio in his now well-known letter to Francesco Barbaro.1

It is supposed that M and P once formed a single MS. Both have the same binding; both bear the name of the same owner, 'Del Sor Conde de Miranda'; and the titulus to M runs thus: 'Manilii Astronomica-Statii Papinii Siluae-Asconii Pediani in Ciceronem-Valerii Flacci nonnulla.' Through the last seven words a line has been drawn (when the two MSS. were separated ?). The titulus, it will be noticed, says not a word of Sigebert's Chronicon. Clark's view is that M and P were put together in one volume while Poggio was in Germany or England: that Poggio brought π back with him from England, and, disjoining the previously united M and P, added π to P.

Now this view is, I think, open at certain points to some rather serious objections. Of these the most powerful is that the titulus to M is written in a hand which cannot be earlier than the end of the sixteenth century. Why no one has called attention to this I do not know. Klotz (Siluae, p. ix) calls the hand recentior

1 C. R., xiii, 1899, pp. 119 sqq.

2 I now find it noted by Loewe apud Breiter, Manilius I, Praef., p. v.

vaguely, and says, what did not need saying, that the titulus must have been written after M had lost its first folio (M begins at Manilius, I. 83). Wishing to be rather more accurately informed, I showed to Mr. Madan, of the Bodleian Library, a photograph of the page of M containing the titulus. Mr. Madan, with his wide palaeographical knowledge, was able to say confidently that the titulus was written in the late sixteenth, or early seventeenth, century. It follows that cannot have been added to P before that date.

On the other hand, P and # can hardly have been conjoined after the sixteenth century; for it is reasonable to suppose that their conjunction followed upon the separation of P from M; and the bindings of P and M belong to the sixteenth century. This piece of knowledge I owe to the courtesy of the Sub-Librarian of the Madrid National Library, who very kindly settled for me one or two doubtful points in connexion with these MSS. On the question of bindings he writes: 'Membranaceum uero inuolucrum quo ambo Msti teguntur saeculo xvio additum, ut ex litteris in umbilico utriusque uoluminis appositis inferri potest.'

About 1600, then, M and P were disjoined, and P was bound up with π. Poggio therefore had nothing to do with connecting P and ; and, indeed, why should he conjoin authors so unrelated as Sigebert and Valerius-Asconius ? 2

π

That P is throughout written in Poggio's hand there is no reason to doubt. Clark holds that also is a Poggian autograph. The hands of P and π are admittedly very dissimilar; but it is thought that this difference may be accounted for by supposing that Pis Poggio's more careful hand, while π is swiftly and rather carelessly written. This explanation, however, is scarcely tenable. I showed photographs of both hands to Mr. Madan, who was not able to think them identical. He pointed out to me that even if it were possible to explain the dissimilarity of the minuscule characters by the hypothesis of a bella manus and a uelox manus, yet in both MSS. the capitals (initials and headings) are written with great care and are yet totally different, e.g. P writes Ah, AH. Moreover, I understand Clark to identify P with the actual MS. sent by Poggio to Nicolaus. Yet, speaking of that MS. to Guarino (Tonelli, p. 29), he says that he wrote it uelociter. But P and cannot both be Poggio's uelox manus.

π

As a matter of fact, it is improbable that P is the actual MS. sent by Poggio to Nicolaus; for between 1423-9 Poggio is constantly asking Nicolaus to return to him this codex: e.g. Tonelli, pp. 207, 294, 303. The most instructive of the passages where he makes this request is p. 207: 'Expecto Valerium Flaccum, Pedianum, et Varronem, quae forsan transcribam, ni

1 This photograph, with others I shall mention, was very kindly lent to me by Mr. Clark.

2 Valerius and Asconius are conjoined because they were found together.