...

primum maesta quieuerat. toro | ... (v.). tiam' nunc dormiens lacrimis ema] ... (vi.) . . . midat.2 & uelut quoddam torm | . . . (vii.) . . . (lu)ctu3 redintegrato prolixum heula . . . (viii.) . . . da5 decora bchia saeuientibus palmulis conuerberat.'

...

...

It is most important to observe how p's original scribe treated these seventeen short lacunae. He left gaps (not corresponding in length at all accurately to those in F), and he made a very few restorations, mostly good.

First for (a). Here he restored <pau>lo in (i.) and <et dies to>tos in (iv.): both restorations are obvious, suitable, and exactly the right length. In (viii.) he restored Tharsyllus for . . . rasyllus, another easy feat. On the other hand, he misread the nia which follows (ii.) as ma, dropped i and lis before and after (iii.), and also dropped / . . . before (v.), qu . . . before (vi.), s before (vii.), the c of solac . . . before (viii.), and the incomplete q of alioq before (ix.). He is also universally held to have dropped the whole of temerarius after (ix.) (the first four letters are not easy to read in F), and he certainly did not write it in its obvious position, immediately before priusquam: but the evidence here is less simple than editors suggest, and I shall return to this point.

In (b) the scribe attempted at most only two restorations, both somewhat wrong. He certainly wrote uulnera. sed lancea in (ii.), where measurement proves that F had uulnera. lancea. More obscure is the case of (i).. Here he has been written before lacrime, and van der Vliet is certainly right (against Helm) in differentiating this-he from the rest of the fourteenth-century supplementa. I think it is clearly earlier, but I doubt if van der Vliet is right in ascribing it to the original writing of the passage by the first hand. I suspect that it was added in an early revision. The gap in F undoubtedly requires not hae but tuae. The scribe dropped t after (iv.), . . . tiam after (v.), . . . midat after (vi.), (lu)ctu after (vii.), and . . . da decora bchia after (viii.). He also ignored interula in F's margin. For heula before (viii.) he wrote only heu, as van der Vliet observed: Helm wrongly ascribes this heu to the later hand.

These details are of great importance, as we shall see when we consider the supplementa added to in the fourteenth century. It is absolutely certain that no corrector who had nothing to guide him but unsupplemented could have produced the results which I am about to describe: to take only the most striking instance, no one could by guess-work, without even the faint light of da or la to help him, have restored correctly the rare word interula. It is therefore necessary for those who adopt the guess-work theory (though they have not, I think, observed the necessity) to assume that the corrector of 4 had also before him either F itself (in its torn state), or a most improbably accurate copy of it. The second alternative need scarcely be considered: and if we adopt the first, it is at least surprising that any scribe, while so laboriously inspecting F, should have refrained from writing his restorations into F as well as into . It would certainly have been obvious to him that F was here p's original. It is even, I think, unlikely that anyone would have restored interula from the inspection of F in its torn state. It is small, difficult to read, far out in the margin, and not exactly opposite the right line. Of this anyone can judge by looking at Helm's facsimile.

It is also desirable to observe the following mistakes made by p's original scribe in words and phrases not directly affected by the rent: they are important for the

1 Helm gives iā, but the ti ligature is clearly legible.

2 Or nidat (the m or n is imperfect).

3 The top of the and the end of the u are legible.

Helm 'heueu (uel heula).' I think heula is certain.

5 Or la. In the margin opposite la or da the first hand has written iterula. Helm is probably right in suspecting that F's text had interida.

6

regularly changes final ae to e. See Lowe in Class. Quart., l.c., p. 153. He also changes F's thrasyllus to tharsyllus, and his tlepolemus to alepolemus.

C

classification of the later MSS. VIII. 7 (182, 6 and 7) uerum religiosae necessitati F, uerum etiam necessitati religiose &: (182, 12) imagines F, imaginem & (not hitherto recorded): (182, 14) se F sese: VIII. 8 (183, 3) procellaque F, procellasque (not hitherto recorded) (183, 6) mugitus iterans F (with ru as a first-hand marginal variant), mugitus reiterans (probably by a misreading of F's ru, which clearly means rugitus : (183, 12) permanat F, permaneat (the ea is of unusual shape, perhaps a altered to ea by the first hand): (183, 13) mortis meae F, mee mortis : (183, 19) perluerunt F, proluerunt p. This word calls for some discussion. All editors give proluerunt as the reading of both F and 4. It is however clear, even from Helm's facsimile, that F has pluerunt altered to pluerunt, without even erasing the original cross-bar. But it is also clear that in F an upright letter by an early hand has been erased above the p. Professor Rostagno is convinced that this letter was : probably, as he suggests, the hand that altered p to p erased the letter above the line. The intention of the scribe who wrote / above the p is hard to guess. Professor Rostagno thinks that he intended pelluerunt.

In 184, 3 the scribe of has been wrongly accused by Helm of not reproducing F's reading the point is most important, for Helm has built on this misreading a very impressive argument against the authenticity of the supplementa. Helm gives the readings of F and correctly: they are qỡ dã F, q dā : but Helm wrongly interprets p's reading as quodam. In fact, both1 mean quoddam. I shall return to the significance of this point: it really provides a strong argument in favour of the supplementa.

I now pass to the work of 's fourteenth-century supplementer, who treated the remaining gaps as follows (making no attempt to fill the vacant spaces neatly).

First for (a). Here in (ii.) he wrote munia, and underlined the following ma in (iii.) he wrote in medullis in the gap between penitus and luctu: in (v.) he wrote luctuoso desi in the gap between insumebat and derio: in (vi.) he wrote quas ad habitum dei in the gap between defuncti and liberi, without emending to formauerat the following forauerat which had faithfully copied from F: in (vii.) he wrote seruitio diuinis percolens between adfixo and honoribus: in (viii.) he wrote tio cruciabat— between sola and tharsyllus, and added uero2 after tharsyllus (182, 14), the false reading of a large class of inferior MSS. In (ix.) the facts are more obscure. Helm prints this supplementum as 'quin-et de ipo noie tem'ari'': but this ignores the existence of an erasure, under quin and the horizontal stroke which follows it. Professor Rostagno, who kindly examined this for me, writes that the original writing has been utterly erased, leaving no legible trace. Possibly an early corrector added temerarius, after a fresh inspection of F (where the whole word is still, with difficulty, legible): and this may have been erased when the full reading was inserted in the fourteenth century. I fancied at one time that quin- looked later than the rest of the supplementa, but Professor Rostagno thinks not.

Next for (b). Here the supplementer left untouched in (ii.) the original scribe's wrong restoration of uulnera. sed, and also the -he of (i.). In (iii.) he wrote Et addi between alienum and dit : in (iv.) he wrote nat. At between inlumi and illa: in (v.) he wrote faciem impressa between thoro and nunc (etiam being thus entirely lost): in (vi.) he wrote nantibus genas cohumidat between ema and et uelut: in (vii.) he wrote to inquieta quieti excussa luctu between torm and redintegrato, in (viii.) he wrote heu eiulat. Discissaque interula decora brachia between prolixum and seuientibus. The same supplementer, as van der Vliet noted, wrote in pectore over inspectore in 182, 9: Helm wrongly attributes this emendation to the first hand of p.

1 See Lowe, The Beneventan Script, p. 191. In VIII. 24 (178, 22) reproduces go by q: there Helm (as we shall see later) prints

quid without comment.

2 Professor Rostagno thinks that this uero is contemporary with the rest of the supplementa.

It is obvious that these supplementa, if they are (to quote Lowe) 'the result of clever conjecturing are remarkably good and no serious objection has been put forward against them on internal grounds. Helm, indeed, objects1 to de ipso nomine temerarius, on the ground that Apuleius uses many other names of obvious significance, without troubling to explain them. I do not feel the weight of this argument, and it is noteworthy that in 1913 Helm still printed de ipso nomine in his text, as the most plausible restoration of the passage. A much more serious argument is that which Helm bases on the assumption that p's original scribe had accidentally substituted quodam for F's quoddam in 184, 3, and that this had provoked the insertion, by o's supplementer, of tormento in the ablative. Were this true, it would be most important, but, as I have pointed out, it is false. We have indeed, if we accept the supplementa, to suppose that F's quoddam is a blunder (so, in Florida 2, Helm 2, 9, F has quod dā modo for quodam modo): but, since also has quoddam, the passage is really a strong argument against the guess-work theory. Had p's corrector 'cleverly conjectured tormento, he would presumably have emended quoddam to quodam. The passage is, in fact, a perfect parallel to 182, 12, where & carelessly wrote imaginem for F's imagines. Here p's supplementer has inserted quas ad habitum, which implies F's imagines, but he has not corrected p's imaginem: nor has he altered to formauerat the unintelligible forauerat in the same sentence.

=

The crucial question, however, as recent editors have recognized, is this: granted their intrinsic merits, are these supplementa of the right lengths to fill the gaps in F? Both van der Vliet and Helm asked the question: yet both were strangely content to answer it with guesses.

'Sed aliis etiam atque etiam diffido,' writes Helm,2' quia mihi quidem spatio quod in F exstat non satis videntur convenire; nam v. 13" seruitio diuinis percolens honoribus" ut aptum est Apuleiano usui dicendi. . ., ita plures nescio an habeat litteras quam quibus lacuna sufficiat ; v. 14 vero duo verba "solacio cruciabat" nullo modo spatium vacuum supplere possunt, nisi conicias verbum in F falso repetitum et deletum fuisse; et sententiarum tenor cum coniunctionem exigeret, in & alia manus "uero" post "Thrasyllus" addidit, vulgata rectius in lacuna ipsa "sed," ego, quoniam ne hoc quidem satis magna est amplitudine, uerum." It is obvious that such problems are not matters for conjecture but for measurement, especially in a hand so unfamiliar and tricky as Beneventan, with its ligatures and its wide a's and t's. Accordingly I have repeatedly measured all F's gaps with a small ivory ruler marked in millimetres and I have found elsewhere in F, and measured with the same ruler, all the combinations of letters found in p's supplementa. The results made it obvious that Helm had not done the same: in almost every case the supplements fitted the space with the minutest exactitude. A large proportion can be checked by anyone in Helm's facsimile: but the gaps there cannot be quite exactly measured, owing to the wrinkling of the parchment, which can only be judged in the original. It must be remembered that in (a) (i.)-(vii.) inclusive, where the gaps include ends of lines, the space is somewhat elastic, as the lines vary in length: the rest are more rigid. I have kept a record of the particular groups of letters which I measured as standards, but here I need only speak of points which call for some special comment, and particularly of the cases where p's supplementa do not exactly fit F's gaps.

(a) (vii.) p's supplement tio cruciabat. Tharsyllus is too short, as Helm justly observes. The addition of either Sed or Verum exactly fills the gap: in F's hand the two words are identical in length. The uero added after Tharsyllus is in direct conflict with F: it is a valuable indication of the immediate source of p's supplementa, as I shall show later on. (ix.) quin et de ipso nomine fits, if we make the necessary allow. ance for the wrinkling of the parchment, and assume that the words were rather

1 Florida, 1910, pp. xxxii sqq.

2 Florida, p. xxxii.

closely written, as they often are in F: for instance on the next sheet (fol. 161b) the words sed necquicquam frustra timorem (190, 4 and 5) are written continuously, with no break between the words. (b) (ii.) (uul)nera exactly fits the space, but there is no room for the sed which the original scribe inserted. (iii.) Et addi (dit), with a capital E (common with et in F, and given here by p's supplementa) exactly fits. (iv.) (inlumi)nat. A(t) does not fit. F clearly had inluminauit. At, with rather a long space after the full stop (exactly as long as that between inducor. At in III. 7 (57, 14= fol. 1356 of F). (v.) faciem impressa exactly fits the space in F, if the required proportion of the mutilated etiam is also inserted. This is especially striking, since p's original scribe dropped the remains of this etiam, and his supplementer has not included it. (viii.) Here it will be remembered that F reads heula . . . of which p's original scribe kept only heu: the corrector inserted a second heu, followed by eiulat. discissaque interula decora brachia (p's scribe having omitted the da [or la] decora bchia which still stands after the gap in F). What fits F's gap exactly is simply t. discissaque interi (or interu)-the t being the last letter of heulat, which F no doubt wrote by mistake for its normal form heiulat. This again is a very odd coincidence, if p's corrector was merely guessing. But if he was adapting to p's imperfect text the reading eiulat, he may well have had the happy thought of making sense of p's heu by expanding it to heu heu.

It thus appears that except in six cases (I include the -he of [b] [i.]), p's supplementa fit F's gaps exactly: of these six exceptions, two are due to the mistakes of the original scribe, and one (—he) probably to an earlier corrector. All can be remedied by very slight changes: but, if p's supplementa were the only evidence, such correction would perhaps be unscientific. In fact, however, in several of the later MSS., and especially in the class which I call Class I., which includes three MSS. of the fourteenth century, we find a continuous text, showing no trace whatever of any alteration or insertion. This continuous text is almost, but not quite, identical with that of o's supplementa, and in every case but one it fits F's gaps exactly. The one exception is (b) (iv.), where these MSS. agree with p's supplementa in reading inluminat. At, instead of the inluminauit. At which the length of F's gap demands.

In the face of this evidence, we are surely justified in concluding that p's supplementa are not 'the result of clever conjecturing,' but the result of an attempt to insert in p's imperfect text the readings of a genuine tradition, derived from a copy of F untorn. It is obvious that, if this conclusion is accepted, the later MSS. deserve careful investigation. I will therefore proceed at once to the classification of the later MSS. this classification is based primarily on their readings in the neighbourhood of the rent, but it is fully confirmed by the evidence of the other books.

Class I.-MSS. which have all the long supplementa (I use this phrase to describe supplements roughly identical with those found in 4), in a form which exactly fits F's gaps, except for the reading illuminat, but which show no trace whatever of the 'short supplementa' (which I shall describe in connexion with the other classes). give below an exact statement of the readings of Class I. in the neighbourhood of the rent my second and concluding article will be devoted to wider aspects of its character and to proofs of its general value. The Class shows indisputable traces of o's influence, but its principal source is undoubtedly a copy of F untorn: a copy, therefore, made not later than the end of the twelfth or the beginning of the thirteenth century.

A1, B1, L1, V2, E, S, N4, a (=the editio princeps). BI is a copy of A1.

The following are the most notable readings of Class I. in the neighbourhood of the rent: where nothing is said, it is to be understood that these MSS. agree with F+p's supplementa, except that a few easy corrections-common to all such MSS. (except F and 4), as have not omitted blocks of the text-are assumed throughout as normal: namely, 182, 9 in pectore, 182, 12 formauerat, 184, 3 quodam. I ignore unimportant differences of spelling.

182, 6 and 7: uerum religiosae necessitati, with F, against : but V2 (an impure MS.) has inserted, after religiosae, p's etiam, though keeping F's order. 182, 7: hilari (hilaro F 6). 182, 8: uidebatur (iubebatur F 6). So Class III. and V4 of Class II. (b): possibly an original variant in F, which has here lost its margin: a has iubebatur. 182, 9: prorsus (prorursus F 4). 182, 10 (Gap [a] [iv.]) diesque totos. So Class III. This fits F's gap exactly, as does also the guess of p's original scribe, et dies totos. 182, 12 (Gap [a] [vi.]): imagines defuncti quas ad habitum dei. No other Class agrees with F in the plural imagines. 182, 13: se (with F: sese). 182, 14: (Gap [a] [viii.]) cruciabat. Sed Thrasyllus. F certainly had either Sed or Verum before Thrasyllus. 183, 6: rugitus iterans (adopting F's variant ru: mugitus iterans F's text, mugitus reiterans : Class III. have rugitus, but combined with o's reiterans). 183, 12: permanet (permanat F, permaneat ): so Class III.: an emendation, not a good one. 183, 13: mortis meae (= F, against 4). 183, 18 (Gap [b][i.]): tuae (with no trace of p's he). 183, 19: proluerunt, except A1, which had ? ptulerunt or ? plulerunt, altered to pluerunt by a very early hand (probably earlier than BI, which has pluerunt): BI (followed by L1), missing a faint stroke in A1, has sanguine for sanguinem earlier in the sentence. 183, 19 (Gap [b] [ii.]): uulnera. Lancea. No other Class has this reading, which was certainly F's. 183, 21 and 184, 1 (Gap [b][iv.]): illuminat. At illa. 184, I (Gap [b] [v.]): etiam nunc (4 omits etiam, even after being supplemented). A1, B1, LI, V2 have non instead of nunc. 184, 4 and 5 (Gap [b][viii.]): prolixum eiulat. Discissaque interula decora brachia: BI characteristically misread Ar's abbreviated discissaque as discissam, which LI, not less characteristically, emended to discissa.

It will be observed that there is little in all this to suggest the influence of 4, and abundant evidence for the direct use of F, in contradistinction to . It is true that A1, besides thrasillus, often has, for F's thrasyllus, tharsillus, which suggests o's tharsyllus, but it never (I think) has alepolemus, as regularly has, but tlepolemus (with F) or thepolemus. E has usually transillus and lepolemus, most deceptively altered by another hand to trasillus and prepolemus: a, like E, has Lepolemus. B1 (though a copy of A1) often writes depolemus. I do not believe that Ar's tharsillus has anything to do with p.

I now pass to the remaining Classes: I will briefly characterize them all, before entering into details about their readings in and near the rent. I usually ignore variants added by later hands.

Class II. Division (a).-MSS. which show no trace (in their texts) of the long supplementa, but which contain certain short supplementa, unconnected with the long ones, and designed, not to fill the gaps in F, but to bridge them by makeshift phrasing.

L6, M1, M2, G, g, D, O, V8. In Met. VIII., L2 (here placed in a separate Class V.) was originally of this class: so too is probably L4 in Met. I., but in Met. VIII. L4 is of Class IV.

Division (b). MSS. of the same Class as II. (a), but containing in their texts some or all of the long supplementa, with plain traces that these have been inserted at some period, in a text originally lacking them. They all show in addition traces of the short supplementa. I shall show that a supplemented MS. of this Class probably conveyed to its famous supplementa.

V6, V4, B3, N1, N3, VI, P. NI is a copy of V4. The lost Fuxensis (if Hildebrand's reports of it are correct) undoubtedly belonged to Class II.

Class III. The most puzzling group, owing to their very mixed character. I have little doubt that they are descendants of the stock of Class I., profoundly affected by Class II., but it might be argued that Class II. is the original stock and Class I. the later influence. Extant MSS. furnish examples of scribes practising both types of V2 of Class I., in particular, has been so thoroughly revised with the help of some MS. of Class II. type, that a copy of it would almost defy classification.

revision.