N2, V7, B2, L3, H, 8. They are much influenced by 4, but this influence is secondary. B2 is more influenced by Class II. than are N2 or V7.

Class IV. MSS. derived in the main from 4. They all possess the long supplementa, but in a form which suggests, as I shall show, that their principal ancestor was derived from unsupplemented, and that the Class obtained the long supplementa, not from supplemented, but direct from Class I. I shall also show that a supplemented MS. of Class IV., and not either supplemented or a pure MS. of Class I., probably conveyed the long supplementa to Class II. (b), from which they passed, for the first time, to ø in the fourteenth century.

L5, L7, V3, V5, L4 (in Met. VIII., but in Met. I. it is probably of Class II.: it is much defaced, and I have not investigated it deeply). L7 is a copy of L5.

Class V.-A special class may be given to the very curious L2. We know from1 the subscription to Met. XI. that it was originally copied from two MSS., and then corrected throughout by the original scribe from an excellent and most ancient one' in the year 1425. In Met. VIII. the original source was almost certainly L6 of Class II. (a) in Met. I. probably L5 of Class IV., but I have not gone deeply into this question. The 'excellent and ancient MS.' was undoubtedly 4, whose readings, good and bad, have been transferred to the first text with almost incredible care. Had F and both perished since the fifteenth century, this late MS. would be incomparably the most valuable in existence: as it is, it might easily prove valuable in cases where is defaced. I have not collated it thoroughly, and I shall only mention it occasionally.

I must now state the chief features of the various Classes and Divisions (other than Class I.) in the rent passages: I ignore spelling variations, and most of the obvious blunders of individual MSS.

=

Class II (a).-182, 6 and 7: uerum religiosae necessitati, with F and I., against . So also II. (b) and III. 182, 7: subcubuit (subcumbens F & I., IV.) with II. (b) and III. 182, 7 quidem omitted after uultu non, with II. (b), and H of III., but I think no others: B3 of II. (b) has quidem, but misplaced (between uultu and non). 182, 7: hilaro, with F II. (b): the rest have hilari, except N2 (yllare) and L3 (hylaro) both of III. The reading subcubuit and the omission of quidem are both consequences of a long omission which immediately follows in II (a), but in no other Class: their function was to make a mutilated sentence grammatical: this omission is from uerum paulo (VIII. 8=182, 8) to solacio cruciabat (ib. 182, 14) inclusive. They open VIII. 9 (182,14) with Thrasyllus uero praeceps et temerarius (lacking the sed, and the alioquin et de ipso nomine, of the long supplementa, and inserting uero). So II. (b), but no others: except that 's supplementer, and o's ape, the revised L2, have borrowed this uero. 183, 6: mugitus iterans (with F's text, against 4). So II. (b), except V4, which follows . 183, 12 permaneat (with 4, against F). So II. (b) and IV. 183, 18: tuae, with I. and III. (no trace of the -he which now stands in ). 183, 19: proluerunt (except O protulerunt). 183, 19: sed inserted before lancea, with II. (b), III., IV., and p's original scribe. This may be borrowed from 4, but it is an obvious insertion, and makes the sentence more commonplace: I believe that it is independent. 183, 20 and 21: the words from et addidit to illuminat inclusive omitted. This is peculiar to II. (a), except for three MSS. of II. (b), V4 and its copy N1, and B3. This omission is followed by the words quod audiens Charite decoras genas saeuientibus palmulis uerberabat, of which the first three words are an invented substitute for the first four lines of VIII. 9 (184, 1-4), and the last five are a careless reproduction of F's decora bchia saeuientibus palmulis conuerberat. This reading is intimately connected with the previous omission, and is again peculiar to II. (a), except for V4 and B3 of II. (b). V4's copy Nı is 1 'quem transcripsi cum duobus incorruptissimis exemplis (he must mean "most corrupt ") me inscio

neque cognito, correptus deinde et emendatus per me ipsum cum optimo atque uetustissimo exemplari, 1425.'

here affected by a later marginal variant in V4, of which I shall speak when I discuss II. (b). It is also certain that this was the original reading of the chief ancestor of V6 and N3 of II. (b), as I shall show in the same place. The reading has affected all MSS. of III., but none of I. or IV. G of II. (a) reads uerberat. 184, 6 and 7: (with slight variations) nec cum quoquam nocturnarum imaginum secreta participabat, instead of F & nec tamen cum quoquam participatis nocturnis imaginibus. This is also the reading of V6 and N3 of II. (b), and the rest of II. (b) and III. are strongly affected by it. It has had no effect whatever on I. or IV.

Class II. (b).—I can be briefer here, as I have already mentioned several points. It will be best to discuss the two rent passages quite separately, as in the second rent passage V4 and B3 of II. (b) have left the II. (a) readings unexpanded.

In the first rent passage, the II. (b) readings are identical (apart from trivial spellings and individual corruptions) with F's text+p's supplementa (including the et dies totos of p's first hand in 182, 10 and 11) except for the following points. 182, 8: V4 (alone) reads with I. and III. uidebatur for F & iubebatur. 182, 9 rursus (for F pro& rursus). So III., but I. has prorsus: of IV., L4 reads prorursus, the rest prorsus. 182, 12: imaginem quam (omitting F p's defuncti). The singular imaginem (4, against F's imagines) is read by all MSS. of III. and IV.; and L3 of III., and V5 and L4 of IV. all agree with supplemented in the ungrammatical combination imaginem defuncti quas. The true imagines defuncti quas is peculiar to I. 182, 13: V4 (with its copy N1) and N2 of III. have ipse for ipso. 182, 13: sese (Þ, against F's se). So all III. and V3, V5 of IV., but none of I. 182, 14: as already mentioned, they leave untouched II. (a)'s Thrasyllus uero praeceps et temerarius. In the second rent passage V4 and B3, as I have said, have left the II. (a) readings unexpanded, except that V4 does not omit tamen in 184, 6, and perpetrates the egregious emendation nec tamen cum quoquam matronarum uirginum secreta participabat (for nocturnarum imaginum): this is copied by N1.

The other II. (b) MSS. all embody some of the long supplementa. V6 (dated 1345) presents the simplest case of obvious expansion. It had best be quoted in full. 184, 1-5 At illa ut primum mesta quesierat thoro faciem impressa etiam nunc dormiens lacrimis emanantibus genas comuidat et uelut quodam tormento inquieta quieti excussa prolixum eiulat discissaque interrula decora brachia seuientibus et cetera. Quod audiens charite decoras genas seuientibus palmulis uerberabat. nec cum coquam nocturnarum ymaginum secreta participabat. It is obvious that At illa. . . seuientibus et cetera was a marginal supplement in a MS. from which V6 was directly or indirectly copied, and that the intention of its writer was that At illa . . . seuientibus inclusive should be substituted for Quod audiens . . . seuientibus inclusive. The scribe of N3, faced with the same problem, showed rather more intelligence. He substituted the same marginal version for Quod audiens . . . seuientibus, but he kept et cetera, and expanded it, with perverse ingenuity, to et cetera membra: his passage ends decora brachia seuientibus et cetera membra palmulis uerberabat. Nec cum quoquam nocturnarum imaginum secreta participabat. VI (Petrarch's MS.) and P1, confronted with the same problem, solved it properly, by substituting the same marginal version for the Quod audiens passage, and dropping et cetera altogether. Their version ends decora brachia seuientibus palmulis uerberabat. Nec

tamen....

V4's copy Ni illustrates the extraordinarily complicated history of some of these supplementa. In V4 a later hand has added, throughout the Metamorphoses, many readings obviously drawn from V1, including a number of very individual marginal notes; and, among other corrections, this hand has written /. against V4's Quod audiens, and has added (in V4's margin) Vi's version of the whole passage: itself, as we have seen the result of a successful fusion of the two versions which V4 had failed to fuse, and betraying its mixed origin by various small points, especially by

uerberabat instead of conuerberat, and by the form of the following sentence. Nor (as we shall see) was even the early marginal version, embodied, with varying success, by both VI and V4, itself (in all probability) a simple tradition, but the result of a still earlier fusion (in a MS. of Class IV.) of a MS. of Class I. with a copy of unsupplemented. The scribe of Ni has embodied this later marginal variant of V4's in his text: but, not realizing that it should be substituted for the Quod audiens passage, he has merely inserted it before Quod audiens, so that in Ni seuientibus palmulis occurs twice, with only five words between, and Quod audiens Charite becomes wholly pointless. Finally, a late corrector of NI, probably using an early printed text, based on a MS. of Class I., has attempted, by careful underlinings and marginalia, to clear up the mess: but, even had his directions been intelligently followed by a fresh scribe copying N1, much unhealed corruption would still have remained. How strongly all this contrasts with the lucid simplicity of Class I.'s text! All the MSS. of II. (b), except B3, V4, and N1, have remedied the omission in II. (a) of the last words of VIII. 8 et addidit . . . illuminat.

Despite small variations, it is obvious that in the second rent passage, as in the first, the version of the supplementa which has made its uneasy way into the MSS. o II. (b) is a single version, with certain peculiar readings: especially the false order redintegrato luctu in 184, 4. It is probable, though not certain, that the supplementa of the first rent passage and those of the second have entered II. (b) from one and the same source: if so, the indications point plainly to a MS. of Class IV. as their principal origin. This is especially suggested by the readings et dies totos (182, 9 and 10), imaginem (182, 12), sese (182, 13); and by the presence in 183, 18 of hae (combined with tuae) in VI and P. These readings, indeed, are common to Class IV. and 4, but the II. (b) MSS. lack (as we have seen) several of the distinctive readings of supplemented, and in 184, 2 they have (with mutilated F, I. and IV.) etiam, which and its faithful follower, the revised L2 and also III. omit. I shall argue later that the readings of IV. in the rent passages are derived, not from supplemented, but from a copy of unsupplemented, revised with a MS. of I.

Before leaving II. (b) I should add that the insertion, by p's supplementer, of uero after Tharsyllus in 182, 14 makes it very probable that a supplemented MS. of II. (b) was in fact the immediate source of p's supplementa. The supplementer's retention of the false sed in 183, 19 (uulnera. sed lancea) fits well with this theory.

Class III. As I have said, this is the most difficult class. I have already indicated some of its peculiarities: I will here recapitulate and complete the list. 182, 7: subcubuit with II. (F and the rest subcumbens): but quidem omitted (with II.) only by H of III. 182, 7: hilari with I. (F and the rest-and L3 of III.-hilaro). 182, 8 uidebatur with most of I. and V4 of II. (b) (F and the rest iubebatur). 182, 9: rursus with II. (b) (F L4 of IV. prorursus, I. and rest of IV. prorsus). 182, 10 and II: diesque totos with I. (ø's original scribe, II. [b] and IV. et dies totos). 182, 12 imaginem with 6, II. (b) and IV. (F imagines, and so I.): L3 of III. agrees with 4 supplemented and V5 L4 of IV. in the ungrammatical imaginem defuncti quas. 182, 13: sese with o and II. (b) (F and I. se). 182, 14: Sed Thrasyllus with I. and IV. ( supplemented and II. [b] Th. uero). 182, 14: praeceps alioquin et de ipso nomine temerarius. 183, 3 procellasque, with 4, L5 of IV., and L2 corrected (procellaque F and the rest, with H of III.) 183, 6: rugitus (with F's margin and I.), but reiterans (with 4). 183, 12: permanet with I. (permanat F, permaneat p, II. [b] and IV.). 183, 18: tuae (without hae). 183, 19 pluerunt (? pluerunt N2)-a peculiar reading, combined, in some MSS. at least with sanguine (also read by B1 and L1 of I.). 183, 19: uulnera. sed lancea with 4, II., and IV. 183, 20 and 21: et addidit . . . illuminat. 184, 1: audiens (obviously derived from II. [a]) inserted between illa and ut : peculiar to III. 184, 2: non (for F [e]tiam

nunc). The reading non for nunc they share with A1, B1, LI, V2, but the omission of etiam they share only with 4, and L2 as corrected.

In the rest of the second rent passage, III. agree with I. against II. (b) (where I. and II. [b] differ) in the form of their supplementa, but have the unique feature of inserting, in 184, 5, decorasque genas after decora brachia: a reading, like audiens, in 184, 1, obviously derived from II. (a): but they read conuerberat, not (with II. [b]) uerberabat. In the first sentence unaffected by the rent (184, 5 and 6), however, they agree with II. (b) in the main features of their reading, but not in the order of words, which is that of F, 4, I. and IV: they have nec tamen cum quoquam participabat nocturnarum imaginum secreta. The variations are unimportant, except that N2 omits secreta. This sentence, especially as it appears in N2, is one of the grounds on which I consider III. to be fundamentally akin to I.

III. has other readings peculiar to itself, such as exhorruit for et horruit in 183, 2 (a reading shared only by O of II. [a]), and licebat for licebit in 183, 11, but space forbids their enumeration. Some of its members have individual peculiarities, especially B2: which omits noctes in 182, 11, reads atque alioquin for alioquin et in 182, 14, and decoras genas (not decorasque genas) in 184, 5. This last is the reading of II. (a), and may represent the original reading of III.

Class IV. does not call for much discussion. I have already said that its readings in Met. VIII., and wherever else I have tested them (except for L4), make it certain that the whole group is derived mainly from 4. A good instance is 182, 6 and 7, where they all follow against F. Whether, on the other hand, the supplementa which they all possess were derived from & as supplemented, or were inserted, from another source, in a copy of unsupplemented, is less easy to say. They show no trace of the shorter supplementa of II. (a), but there is no reason, in any case, why they should. I believe that they are derived from unsupplemented, because their supplementa are definitely, though slightly, superior to 's in the following points. 182, 14: Sed Thrasyllus, with I. and III., against and II. 183, 18: tuae hae. I have pointed out that hae probably stood in before the long supplementa were inserted, and that 's supplementer did not trouble to alter it, or even to add tuae to it. 184, 2: etiam nunc, though the carelessness of p's original scribe, in ignoring the surviving tail of F's etiam, was not remedied by o's supplementer. 184, 4: prolixum eiulat, with all other MSS. that contain the phrase, instead of p's prolixum heu heu eiulat (in which, as I have shown, the first two words are original, the second two due to the supplementer). These facts might, of course, be explained by the hypothesis of a careful revision of a copy of supplemented with the help of a MS. of Class I.: but it is simpler to suppose that the text so revised was a copy of unsupplemented: for such a copy, being obviously defective, would have been more likely than a copy of supplemented to provoke such revision. I should add that this conclusion is confirmed by the fact that, though the spurcum additamentum was added to in the thirteenth century, yet, of the MSS. of IV., neither the two of the fourteenth century (L5 and V3), nor L5's late fifteenth century copy, L7, show any trace of it: and, though it appears in L4 (dated 1422) in the hand of the original scribe, it is there, as I have shown, borrowed not from direct, but from LI: it occurs also in V5 of the late fifteenth century, but borrowed from L4. These facts suggest that the stock of IV. diverged from a century or so before it was supplemented in the rent passages. We have thus seen that the faithful L2, in its corrected form, is the only certain or even probable example of a MS. indebted in these passages to the hand which filled up the gaps in : that hand which we are asked to consider the sole and irresponsible source of the whole of the long supplementa.

It seems, therefore, that the examination of the whole body of MSS. confirms the opinion which I expressed at an earlier point: that in all probability the supple

42

MANUSCRIPTS OF THE METAMORPHOSES OF APVLEIVS

menta' of Class I. are not supplementa' at all, but the original readings of F, copied from that MS. before its mutilation: and that it is consequently reasonable to hope that the detailed examination of Class I.1 will throw light on dark places of F which even is powerless to illuminate.

In my second and concluding article I shall prove that this hope has in fact been fulfilled.

Tempore

LVCRETIVS II. 263 'nonne uides etiam patefactos tempore puncto.' puncto' occurs only here in Lucretius and in no other author; but 'puncto tempore' is read in II. 456, 1006, IV. 214; 'puncto in tempore et,' VI. 230. Temporis puncto' is found at I. 1109, and 'temporis in puncto' at IV. 164, 193. Puncto . . . diei' occurs in IV. 201. 'Punctum' as a noun corresponds to aroμos, for a point has no dimensions; St. August. Ep. 205, 14, atomo temporis, inquit, hoc est in puncto temporis quod diuidi non potest.' Inquit' refers to St. Paul I. Ad Corinth. XV. 52, where the N.T. has év áróμ and the present Vulgate in momento.'

Ep. 190, 15, St. Austin says 'summa celeritate atque atomo temporis'; and St. Jerome, Ep. 119, 2, 'in atomo et in puncto temporis atque momento,' 5' atomus . . . punctum temporis est'; J. Cassian, Inst. II. 7, 2, 'puncto breuissimo'; Lact. Inst. VII. 12 'uno temporis puncto.' 'Punctum temporis' or 'atomus temporis ' means an instant of time that cannot be divided.

'Puncto tempore' must have the same meaning; and Lindsay, Class. Quart. XIII. 19, and Diels, Lukrezstud. V. 50, finding Munro's explanation unsatisfactory, would have 'tempore' written tempori,' with apocope of the letter s. This proposal is refuted by Lucr. VI. 230, where 'tempore' is followed by 'et.' There is no instance in Lucretius (and probably in no other author) of the dropping of s before a vowel (Randall, A.P.A. 34, lxvi).

In Claudianus Mamertus, De Statu Animae III. 16, the rare adverb 'punctatim' occurs in the meaning condensed into a point': 'collectim strictimque et ueluti punctatim . . . redegi'; and this passage leads to the interpretation of 'puncto' in Lucretius as 'made into a point' or 'condensed into a point'; time is 'pointed'— that is, reduced to a point or moment-and 'tempore puncto' means 'atomized time.' 'Puncto' is a regular participle, and means pointed in the sense of made into a point or atom.

Lucretius had no imitators in this peculiar use of the passive of the verb, and he would not have used it had it not been for the atomistic philosophy. If pungere' can mean 'make a point,' then 'pungi' can mean 'to be made into a point'; but the genius of the language preferred the noun 'punctum' to the participle, and to make a point is punctum facere' and not 'pungere.'

UNIVERSITY OF CALIFORNIA.

W. A. MERRILL.