STOP Early Journal Content on JSTOR, Free to Anyone in the World This article is one of nearly 500,000 scholarly works digitized and made freely available to everyone in the world by JSTOR. Known as the Early Journal Content, this set of works include research articles, news, letters, and other writings published in more than 200 of the oldest leading academic journals. The works date from the mid-seventeenth to the early twentieth centuries. We encourage people to read and share the Early Journal Content openly and to tell others that this resource exists. People may post this content online or redistribute in any way for non-commercial purposes. Read more about Early Journal Content at http://about.jstor.org/participate-jstor/individuals/early- journal-content . JSTOR is a digital library of academic journals, books, and primary source objects. JSTOR helps people discover, use, and build upon a wide range of content through a powerful research and teaching platform, and preserves this content for future generations. JSTOR is part of ITHAKA, a not-for-profit organization that also includes Ithaka S+R and Portico. For more information about JSTOR, please contact support@jstor.org. THE NEW CRITICAL EDITION OF OVID'S METAMORPHOSES By E. K. Rand About half a century ago, Ribbeck laid the foundations for the critical study of the text of Virgil by his monumental edition of that poet. It is surprising that nothing at all comparable to this under- taking has been done for Ovid until now. Hugo Magnus, who has devoted over twenty-five years to the text of the Metamorphoses, now edits this work with the prose paraphrases of "Lactantius Placidus" on the basis of a full collation of all the important manu- scripts. 1 The Introduction contains a summary description of the manuscripts, a discussion of their interrelations, a brief history of the text in the Middle Ages, an account of the chief editions, and a bib- liography of recent works which deal with the text. The apparatus includes both variant readings and, separately printed, citations of the text in later Latin authors, including some from the Middle Ages. The work concludes with an index of proper names and facsimiles of N, F, and M, the three oldest of the complete or almost complete manuscripts. The printer deserves a word of execration for not allowing a very plump volume of 766 pages to be bound in two parts without cutting into a book of the poem. The apparatus is extraordinarily full. It contains bibliographical references and comments on the reason for scribal errors; in one case we have the gist of a small critical article which leaves room for only five lines of text on one page and none at all on the other (1. 545). Nobody can object to this amount of detail in an edition which is intended, not for the reader, but for the student of the text. The same information, however, might now and then be presented more briefly; for instance in 2. 584: "alta, scripsi cum, N Igs," scripsi cum is otiose, and so recepi cum and the like elsewhere. This is a trifling matter; a vital objection to Magnus' method of citation in the apparatus is its ambiguity. Take for instance 3. 49: adflatu 1 Berlin: Weidmann, 1914. Classical Philology XI, January, 1916] 46 The New Edition of Ovid's "Metamorphoses" 47 funesti tabe veneni. The apparatus has: "afflatu funesti tabe vene- ni a afflatu funesta f} eO F s." Do these last-named manuscripts omit tabe veneni f Strictly we should infer that they do; somehow we know that they do not. But in 1. 133: "quaeque prius a," we cannot tell, without consulting the facsimile of the Bernensis, whether this reading replaces quaeque diu of the text or whether a has quaeque prius diu. There may not be many cases of such ambiguity; there should be none at all. If, now, we would examine the groups of manuscripts and turn, as one frequently does, to the summary description in the table of abbreviations, we find a peculiar method of classification (p. xxxv). The headings are: "I. Codices integri vel paene integri. 1. Antiquis- simi. 2. Recentiores"; "II. Codices mutilati et Fragmenta"; "III. Codices in libro XV — adhibiti." Now the first of the fragmentary codices mentioned is a, the set of excerpts in the famous Bernensis 363 s. IX, older by two centuries than M , N, and F, the manuscripts under I, 1. Magnus regards a, too enthusiastically, we shall see, as a source of extraordinary value, "quod a ceteris libris seiunctum sua via haud scio an paene ad aetatem Ovidianam redeat" (pp. x and xxvi) ; at any rate a is a fragment of distinct worth. Others of the fragmentary codices are as old, or nearly as old, and as valuable, or nearly as valuable. Class I should consist of the best repre- sentatives of the text, whether complete or nearly complete or sadly deficient in size. Quality of text decides the group and not quantity. This principle is naturally recognized by Magnus in his special studies and in the Introduction, but it should not be abandoned in the list of manuscripts to which the student turns again and again for the essential information about the chief families of texts. As the editor has conveniently indicated the oldest complete manu- scripts by heavy capital letters, the younger complete manuscripts by small letters, and fragments by Greek letters, the nature of the contents of any manuscript would be easily recognized, whatever the system of grouping. No useful purpose is served by putting all the fragmentary manuscripts under one heading. There are two chief families of manuscripts, according to Magnus, and X. The manuscripts of the former are M (Marc. Flor. 225 s. XI); N (Neapol. IV F. 3, "s. XI, Beneventan script," but of the 48 E. K. Rand Bari type and rather s. XII, as shown by Loew, Beneventan Script, p. 151); a (Bern. 363 s. IX), and other fragmentary codices ranging in date from the tenth to the twelfth century. This class, thinks Magnus, depends on an edition of the poem made in the later Empire by some careful scholar, who revised the text and added the prose versions of "Lactantius Placidus" with other explanatory notes. The manuscripts of class X, i.e., F (Marc. Flor. 223 s. XI ex -), and various later codices, among which the fragmentary r (Monac. 29007 s. XII) is the most significant, derive from the ancient Vul- gate text which had not received the scholarly revision that appears in 0. Though the readings of class X are often inferior, it has a special value in preserving the fifteenth book, which with the end of the fourteenth is lacking in 0. No attempt is made, for a reason that will later appear (see below, p. 51), to discover through what country the Metamorphoses came into the Middle Ages. Nothing in the material incidentally given by Magnus corroborates S. Tafel's unsuccessful endeavor to show that this country is Spain (Die Ueber- lieferungsgeschichte von Ovids Carmina Amatoria, 1910). In the case of both the Carmina Amatoria and the Metamorphoses, though our present information warrants no certain conclusion, the signs point rather to Ireland, a is in Irish script, /? (Brit. Mus. 11967 s. X ex ) betrays by its errors a descent from an Insular original, and the same may be true of N, if the peculiar mistake of eserat for egerat (2. 583) is a criterion. This subject calls for further investigation. The work of accumulating the entire material, revising the old and incorporating the new, is enormous and commands the gratitude of every reader of Ovid. With what success has the mass been sifted? What is the evidence for the classes as Magnus has dis- tinguished them ? Starting with the fact that Ovid burned a copy, but not all copies, of his Metamorphoses when he was forced into exile, and with the possibility that he worked over the poem and presented it to the emperor Augustus, Magnus considers the theory that our manuscripts contain in certain passages both the original and the revised form. This question of a double recension is delicate and perhaps insoluble. It has often been treated, by Magnus among others. He believes that all the passages concerned present a gen- uine and an interpolated reading; he finds no evidence of a double The New Edition of Ovid's "Metamorphoses" 49 version by the author. A typical case occurs in the story of Daphne (1. 544 ff.). As Apollo pursues her, the nymph calls for help: victa labore fugae, spectans Peneidas undas, "fer pater" inquit "opem, si flumina numen habetis! qua nimium placui, mutando perde figuram!" [victa labore fugae "Tellus," ait, "hisce, vel istam, quae facit ut laedar, mutando perde figuram."] Against the view of most scholars who have treated this passage, that here is a good instance of double recension, Magnus attributes the invocation of the river-god to Ovid and the invocation of the Earth to some interpolator of the fifth or the sixth century. I am inclined to remove the brackets from his text and assign the whole passage to Ovid. The nymph might have called both to her father and to Mother Earth. It is dangerous to infer that "Lactantius Placidus" who says nothing of the appeal to the Earth in his very brief account did not find that passage in his manuscript; or, sup- posing it was not there, its omission could have been accidental, the scribe being misled by the similar phrases, which have also occasioned various confusions in our extant manuscripts. Extensive repetition is quite in Ovid's manner; compare 1. 361-62: namque ego (crede mihi) si te quoque pontus haberet, te sequerer, coniunx, et me quoque pontus haberet and many other passages. This trick of style should make us care- ful not to apply the knife too vigorously to Ovid's exuberances. Who would think of emending saepe pater dixit "generum mihi, filia, debes," saepe pater dixit "debes mihi, nata, nepotes" (1. 481-82) or imagine that Ovid is responsible for the son-in-law, but some interpolator for the grandsons? Similarly, 8. 286; 8. 697-98; and surely 6. 282 had better stand. In other cases we must admit either a double recension or an interpolation. A crucial instance is 8. 597-610, which describes the metamorphosis of Perimele. The second of the two added passages, which occurs in h e g n F 2 s, ends with 607-8 (I follow Magnus in printing it in italics) : dumque ea contrecto, totum durescere sensi corpus et indvcta condi praecordia terra. 50 E. K. Rand The following lines, dum loquor, amplexa est artus nova terra natantis, et gravis increvit mutatis insula membris, are clearly a variation on the same theme. They are as neatly con- nected with the speech immediately before the added passage as they are inappropriate here. Both additions — 597-600* and 603-8 — have a distinctly Ovidian flavor and have been defended by modern scholars. One Ovidian trait is obscured by Magnus, who omits a line (I call it 600 b ) in the first of the added passages. It connects with the briefer text thus: Cui quondam tellus dausa est feritate paterna [600 b ] adfer opem mersaeque, precor, feritate paterna da, Neptune, locum; vel sit locus ipsa, licebit [602]. Verse 600 b , in h $ n Planudes, etc., is a genuine part of the additional text; the omission of 601 in e is natural enough — one of the manu- scripts of Planudes trips on the Greek homoioteleuta in the same fashion. The rhetorical repetition here and, in the second passage, the picture of the river-god fingering the nymph's quivering breasts and finding them turn to terra firma either are Ovid's own or would please him immensely. In the eleventh and twelfth centuries there were poets besides Hildebert of Tours who wrote very decent, or indecent, verse in Ovid's style, as the pseudo-Ovidian poems of the period show. As the manuscripts containing the longer version in our present passage are none earlier than the later twelfth century and as it does not appear in F and I of class X, Magnus may be right in calling it an interpolation. However, in the next story, Philemon and Baucis, similar passages occur (652-656 a , 693 a -693 b , 697 a -698 a ) which are found in more of the X manuscripts, including F; further, the hand that supplies 693 a -693 b and 697 a -698 a in M is said by Magnus to be not much later than the text itself. For aught I see, there is still room for the hypothesis that the parent manuscript of X, and perhaps of O as well, contained in the margin these longer versions, which are the work of Ovid himself, being his earlier rather than his later recension; these marginalia would then be included, with varying degrees of completeness, in different copies and their descendants. To discuss the subject profitably, we need first a The New Edition op Ovid's "Metamorphoses" 51 more careful account than has appeared of the nature of the mediaeval commentaries which accompanied the Metamorphoses and from which glosses and variants filtered into the text; such an affair might have contained clever improvisations on Ovidian themes and these would make their way into the text also. For the moment we had better suspend judgment; perhaps the question can never be settled. Another matter that will not win the approval of all scholars is Magnus's rejection of the old-fashioned theory of an archetype, which he formerly held but which he now declares has wrought great havoc in criticism (p. ix). He had formerly collected numerous instances of agreement in significant error to prove the existence of a single manuscript from which all extant codices must derive. He now asserts that Ovid must always have been read from the Empire through the Merovingian period into the Caroline, and that not one copy, but many copies, were at the disposal of the Caroline scribes. That may be, but how does he know it ? From the evidence of the common errors in our manuscripts, we infer that they all descend from one copy at least as early as the ninth century, as a is of that date. Quoting with approval the dissertation of Tafel mentioned above, Magnus holds that the common errors do not come from an archetype but arose "im Laufe der Ueberlieferung." They were found, he declares (p. ix), "et in eis libris, quos legebant veteres, et in illis, qui veniebant ex antiquo aevo in medium, nee tamen in omnibus omnes, sed paulatim hinc illinc illati .... hos igitur libros aetati Carolingicae, pignoris depositi fidae illi sanctae- que custodi, traditos cum suis erroribus, non scribae alicuius insip- ientis sed saeculorum culpa commissis, verius dicas archetypum codicum nostrorum, verius signes siglo A." Let us see how this theory works. Supppose there are thirty such errors, all present in each of fifty extant manuscripts of different dates. What right-have we to imagine, let us say, the existence of ten manuscripts in the ninth century, each with a different set of three of the thirty errors, which groups, after various cross-breedings, resulted in fifty manuscripts each containing the thirty errors? Perhaps a mathematician might calculate the necessary permutations and combinations to produce this condition. Incidentally, it does 52 E. K. Rand seem as if in the process some of the different sets of errors might be cleared away with nine sources in existence that knew them not. None would deny the gradual accumulation of error or be disposed to lay to one scribe the blame for all the faults of an archetype, but the history of the errors lies behind the archetype and is past finding out. The single text to which we can argue back on the basis of the fifty manuscripts contained the thirty errors which all succeeding scribes perpetuated; the question of who had previously occasioned them is beyond our power to raise. " Etenim si duo libri C et D ex eodem exemplari B descripti eosdem errores habeant," asks Magnus, "num recte concludes eos omnes fuisse in A, ex quo B fluxerit?" But whoever does so conclude, or what bearing has this case on the present matter ? We can argue from C and D as to the status of B, and find paleographical reasons, now and then, for the errors of B, but what proportion of them was committed and what inherited by B we cannot tell. It may well be, as Magnus says, that two different errors of omission or transposition present in all our manuscripts arose in two different centuries. Sed haec prius fuere. All that we can prove in our hypothetical case by reasoning from the data before us is the existence of one copy containing the thirty errors. Nor is this a surprising conclusion to one who considers the history of the early Middle Ages. The best connoisseurs of this period, like Traube (Vorlesungen, II, 133) and Manitius (Geschichte der lateinischen Literatur im Mittelalter, I, 250 ff .) are aware of a sharp break in the transmission of ancient literature after the sixth century and its resumption in the last part of the eighth. Charlemagne and his fellow-workers in the revival of learning would send to Italy or Ire- land or Spain for a copy of some work known to exist in one of these centres. The copy with its errors thus became the unique source of the manuscripts known to us. The existence of a number of copies each containing the thirty errors is of course possible, but it must be demonstrated by something more than the assertion that the works of Ovid never ceased to be read. This assertion finds little support in the very scanty citations of Ovid in early mediaeval literature. The same disposition to exceed the actual evidence appears in the discussion of 1. 304. In his fable of Deucalion, Ovid describes various animals swimming helplessly in the flood : The New Edition of Ovid's "Metamorphoses" 53 nat lupus inter oves, fulvos vehit unda leones unda vehit tigres, nee vires fulminis apro, crura nee ablato prosunt velocia cervo. The first line is quoted by Seneca (N.Q. iii. 27, 13 — not 3 as Magnus has it), and certain manuscripts of the X class and Planudes give the passage correctly. But (i.e., M and N) and e and r of class X omit the two hemistichs fulvos .... leones and unda .... tigres, later hands in and r making good the omission in the margin or between the lines. Manuscript a presents the curious text: nat iupus inter oves nee vires fulminis apro unda vehit tigris fulvos trahit unda leonis crura nee ablato prosunt velocia cervo. Here the two hemistichs have been replaced but strangely misfitted. The obvious fact, seen by Gercke and Ehwald, is that 0, a, e, and t all depend on a single copy in which the omission was first made and then supplied; the scribe, after writing oves, caught at tigres in the next line, continued with nee vires, and later discovered and made good his mistake. Magnus, after objecting to Gercke and Ehwald in his apparatus, admits in the Introduction (p. xii), that "in aliquo nobili exemplo, ex quo multa apographa postea venerunt, hemistichia primo omissa, post suppleta sunt." On the principle of "licet hario- lari in re incerta," he imagines this to be the copy sent by Ovid to the Emperor. Thereafter, he continues, scribes either omitted the added verses (as in 0) or inserted them in the wrong place (as in a) or in the right place (as in Seneca's copy). This last suggestion, no less than the idea that the error was made in Ovid's own copy, is hariolatio rather than proof. Our facts take us only to some nobile exemplar in which the omission and its rectification were made. Of its date we can say only that it was older than a. Seneca read the right text, but there is nothing to show that the error had already been committed and emended. Whatever the age of the manu- script which contained the corrected error, it is the source of both and some of the manuscripts of X (e and r). Since certain manu- scripts of X have the right reading, this highly significant case is one of the criteria of class 0; we should inquire whether e and r, which Magnus calls X, are not really contaminated with X, as he shows is true of k s. XII (p. xv) and ju s. XIII (p. xix). How such 54 E. K. Rand an inquiry would terminate I cannot presume to say, but at least we need further pondering on "vanum istud et saepe male intellectum nomen archetypi," and a further examination of manuscripts e and r. Indeed, after testing the new apparatus in a number of important passages, the more doubtful I am that Magnus has really established his groups. Instances like 7. 144-46: tu quoque victorem conplecti, barbara, velles: obstitit incepto pudor, at conplexa fuisses, sed te ne faceres, tenuit reverentia famae, where not only but e s. XII/XIII and h s. XIII transpose vs. 145 after vs. 146, and still better, for the complicating phenomenon of similar endings is not present, 6. 496-98, Hanc ego, care gener, quoniam pia causa coegit ut voluere ambae (voluisti tu quoque, Tereu), do tibi perque fidem cognataque pectora supplex where in h e p vs. 496 is transposed after 498 and falsely replaced in g and I, suggest that manuscripts ehp g and I, called X, have also some relation to 0. Similarly, 1. 91-93 are omitted by a e ir N h e, manuscripts representing both classes, and in 2. 191-94 there is the same transposition in e M Fl. Here are six manuscripts put in the X family by Magnus and yet agreeing in important cases with the errors of 0. An interesting matter to which Magnus refers (p. vii), but the evidence of which for the classification of the manuscripts needs fuller treatment, is the existence of a mediaeval commentary on the Metamorphoses, written as early as the eleventh century. We may owe to this, I believe, and not to any recension of the later Empire the obvious incorporation of glosses in both and X. Here are a few examples. In 0: 14. 533, Mulciber urebat perque alta ad carbasa malum) malum] caelum M , malum (ma in ras. m. 2) N; 1. 475, silvarum latebris captivarumque ferarum) latebris] tenebris M; 2. 620, vidit et arsuros supremis ignibus artus) supremis] sup- postis JV. In X:2. 582, at illa/pluma erat inque cutem radices egerat imas) egerat] fixerat ehls; 2. 681, onusque fuit baculum sil- vestre sinistrae) baculum silvestre sinistrae] baculus silvestris olivae X. Further, the classes and X are in some cases infected with the The New Edition op Ovid's "Metamorphoses" 55 same gloss, as the following show: 1. 641, pertimuit seque exster- nata refugit) seque exsternata] seseque exterrita] e N eh I; 2. 587, mox alta per auras/evehor) alta] Nigs acta cett.; 2. 658, vocisque meae praecluditur usus) praecluditur] praevertitur F; 2. 688, Battum vicinia tota vocabat) tota] rura Net. Such evidence leads us to associate these two groups still more closely, and not to regard one as an ancient scholarly recension and the other as an ancient vulgate. Now exactly the same kind of infiltration of glosses is obvious in the Berne fragment a. Its original was provided with explanations and variants, as in 1. 21. litem] uel limitem ss.; 1. 25, ligavit] locavit (uel ligavit ss.); 1. 63, vesper et] uel uesperet uerbum primae coniug ss.; 1. 105, in duris haerentia mora rubetis) rubetis] id est rubus ss.; 1. 193, faunique satyrique et monticolae silvani) silvani] u hie soluitur in uocal ut sit dactil ss. Some of the glosses, moreover, worked into the text, as 1. 2, nam vos mutastis) vos] di; 1. 59, cum sua quisque regant diverso flamina tractu) regant .... flamina] rotat .... fulmina; 1. 196, an satis, o superi) superi] pueri (superi ss.); 3. 53, splendenti lancea ferro) ferro] telo (ferro ss.). With this information before us, we should be exceedingly chary of accepting with Magnus readings of a not supported by other manuscripts; for instance 1. 33, redegit] vulgo coegit a Magn.; 1. 56, frigora] fulgura; 1. 99, erat] erant (here he admits that erat may be right). I may add that a manuscript in Insular script which con- stantly confuses i and e (e.g., inse for ense, 1. 191 and secure for securi 3. 28) is not much of an authority for the spelling -is in the accusative plural. Now in one case the incorporated gloss agrees with N: 1. 70, cum quae pressa diu massa latuere sub ilia] pressa diu f uerant caligine ceca (multa N) N a, the line being metrically reformed after caligine had displaced massa. Magnus, in his various articles, recognizes clearly enough the presence of incorporated glosses in our texts of the Metamorphoses. We need, however, a minute investi- gation, if possible a reconstruction, of that early mediaeval commentary, the effects of which are felt, if I am right, in all our manuscripts. In opposition, then, to Magnus' theory of two ancient recensions and a multiplicity of copies in the early Middle Ages, I incline to the following. In the eighth or the ninth century, per- haps in Ireland, the single manuscript (A) from which all our copies 56 E. K. Rand derive, was provided, or various copies of it were independently pro- vided, with a set of glosses and variants of the kind abundantly familiar in the history of the early Middle Ages. There may be two main families descending from A, i.e., and X, but their criteria and their subclasses need to be more accurately determined. Into all the descendents of A, glosses and "emendations" from the early mediaeval commentary have infiltrated in varying degrees. The critic must have a lively sense of this situation and the ability to tell a mediaeval gloss from a word of Ovid's. Of course judgments will differ. Here are a few of the places in which, it seems to me, Magnus has substituted the mediaeval gloss for Ovid, while Heinsius, with inferior material and no knowledge of wissenschaftliche Methode, has now and then divined the truth. 1. 55, pigris radicibus haeret, /ora cacumen obit] F e 2 s Heins., vulgo abit t habent e h I habet Magn.; 1. 710, "hoc mihi concilium tecum" dixisse "manebit!") concilium] cod. Lovan., Heins., vulgo consilium e F N g I s conscilium h colloquium M s {harper}) Plan., Magn.; 2. 275, sustulit omni- paros collo tenus arida voltus) omniparos] coni. Heins. omps e omni (feros in ras. m. 2) N omniferos X vulgo oppressos (omniferos ss feros m. 2) M oppressos Merkel, Ehw., Magn.; 2. 730, vertit iter caeloque petit diversa relicto) diversa] terrena M s Merk., Ehw., Magn. It were rash for one who has not, like Magnus, worked through the whole array of evidence to attempt to draw up classes of manu- scripts. One may, however, after only a partial examination conclude that the method of classification adopted by Magnus is unsatis- factory. Let us test it further in the case of v (Vat. Urb. 32 s. XI). This fragment consists of two leaves in different hands but from the same book. The first contains 5. 483 — 6. 45, the second 7. 731 — 8. 104. This is assigned by Magnus to class X, on the strength of certain readings and the fact that the summaries of "Lactantius Placidus" are lacking. But the omission of the summaries is not a sure test; they might have stood in an manuscript from which v was copied. The readings cited by Magnus are also of dubious significance. They are: 5. 558, alarum insistere remis) remis] v pennis $. This instance shows that v is not affected by a gloss incorporated in M and N; it is not, however, necessarily of class X The New Edition of Ovid's "Metamorphoses" 57 because agreeing with that class in a right reading. It might have got this right reading by another line of descent from the archetype. 7. 731, desiderioque calebat/coniugis abrepti) calebat] v dolebat M s calebat m. 2 N. Dolebat is a gloss on calebat, as Magnus formerly thought (Jahrb. fur class. Philol, XCIII [1893], 628). Ehwald is the only editor who reads dolebat. This case, then, is like the first. 7. 756, dat simul et iaculum, manibus quod (cernis) habemus. Here M shows the glosser's hand; its reading, iaculum quos nos (manibus ss m. 2) ut cernis habemus seems to have been that of N as well. In h I s habere replaces habemus, while the rest of X are apparently correct, v has iaculum manibus quod cernis ha, which is right as far as it goes. We should treat this case, then, like the readings already given. 8. 85, vitali nata [i.e., Scylla] parentem/crine suum spoliat) uitali] v e s Merkel fatali F, alii X, vulgo. Vitali strikes me as too good for a gloss. Magnus should have noted that Riese as well as Merkel reads it. If it is right, this case does not differ from those preceding. If fatali should be read, it is curious to class v with X because agreeing in error with 0. The analysis given by Magnus fails to recognize the elementary truth that class-relationship among manuscripts is determined solely by agreement in errors; agreement in correct readings between two manuscripts means merely that they are both true to the original text, from which they may have descended by entirely different routes. Surely there is nothing in the readings cited by Magnus to lead us to associate v with X. Indeed, if he has chosen the right reading in the last instance, v belongs with 0, or at least is affected by the same gloss that has ousted the text in 0. Let us turn to Magnus' apparatus for further light on v. To begin with, in the following passages he does not, either by inference or by direct statement, let us see what the reading of vis: 5.576, Elei; 598, propiori; 608, Elim; 669, minantia; 6. 18 Turn; 7. 762 (om. ?); 805, iuvenaliter; 8. 5, ante exspectatum; 15, Letoia; 59, tenentibus; 64, non metuam; 65, durus. On 7. 771, pugnae is cited for v, but it is apparently a variant for pugnat in 772. There are thus in 480 lines twelve instances, highly important for the question of classification, in which Magnus does not indicate the readings of v. Perhaps the collation by Hosius which he used was defective in these places; but 58 E. K. Rand it would have been easy to secure photographs of the two leaves. For twelve of the twenty-one manuscripts on which he bases his text of books i-xiv, Magnus depends on collations made by others. Though every last bit of evidence should be in before one pro- nounces a final judgment on v, some facts are obvious enough to be stated now. Although in the cases cited by Magnus v agrees with correct readings of X and although in three cases it agrees with one of the manuscripts of X in insignificant errors that different scribes could easily commit independently (5. 529, sorte] forte Fv, 561, ne tamen] nee tamen F v, 675, mota] nota F v), there is no certain bond of agreement between v and X. The case is quite different with 0. We may approach this fact by considering first another feature of v, namely, that like M and N and F it is infected with glosses. Some were copied faithfully in their proper places as inter- linear notes, e.g., 5. 587, vertice] (origine ss.), 644, aera (aethera ss.). Others have worked their way into the text, as 5. 588, ad humum] usque ad humum, 8. 34, regebat] premebat (regebat ss.). Such glosses, explanatory in the original intent, were imagined by the scribe of v, or some ancestor, to be variant readings and were sometimes substituted for the text. Besides these, genuine variants had been selected from some other manuscript source and written between the lines in some ancestor of v. A study of these will readily indicate how one should proceed to classify v. 5. 616, sed certe sonitusque pedum terrebat et ingens/ .... anhelitus) terrebat] tegebat (terrebar ss.). Here tegebat is a gloss substituted for the original terrebat, the reading of A, terrebar being an "emendation" of inferior codices used by Heinsius and now shown to be as early as the eleventh century. 5. 641, advehor Ortygiam, quae me cognomine divae/Grata meae superas eduxit prima sub auras.) Grata meae] s Heins., vulgo Ortamiae /3 M Grta (a ss. m. 2) m/ (i ss. m. 2) N (Grata in ras. m. 2) meae F Grata mihi e h I s Orta (Grata ss.) v. The con- flate character of v is apparent; the curious error of 0, Ortamiae, was its original text, which then was supplemented with Grata mihi from X. The peculiar condition of F and the mihi inehls suggest that the archetype of both and F had committed the error and also corrected it. The New Edition of Ovid's "Metamorphoses" 59 Very similar is 5. 670, conantesque loqui), for which /3 M had originally by a curious error conantesque oculis, while X replaces conantesque by conataeque. The reading of v, Conataeque oculis (loqui ss.), shows again an original modified by contamination with X. As N is here correct, v should be set on the M and not the N branch of 0. 8. 62, melius sine caede moraque/inpensaque sui poterit superare cruoris.) inpensaque] incessaque v (inpensa ss. m. 2) M ; sui] suum (sui ss. m. 2) M su (i in ras.) v; cruoris] cruorem (cruoris ss. m. 2) M cruorem (cruoris ss.) v. Here in one line are three plain cases of an original M reading in v corrected from some other manuscript. 8. 77, nee in hoc] nee hoc (in ss. m. 2) M nee (h m. 2) uc (ad uel o ss. to. 2) N neque ex nee (in ss.) hoc v. Here v starts with nee hoc the original reading. Neque is found in s Heins. and editions before that of Merkel. 8. 90, proles ego regia Nisi/Scylla) regia] gloria (regia m. 2) M regia in ras. N 2 gloria (uel regia ss.) v. To the foregoing certain cases of connection between v and 0, we may now add the following. 5. 509, ad]ut v 0; 7. 824, ficti] facti v M N 1 e I; 7. 760, solverat] s Heins. solvitur v M N 1 Solu(unt in ras.) N soluunt X (F incert.); 8. 59, causaque] causasque v M and especially the verse 8. 87] Priscian. X Plan., om. v 0. v therefore is a mixed manuscript. Its basic text is O and specifi- cally M. In some ancestor of v, readings from another manuscript were entered between the lines. We may infer from 5. 616; 8. 77, and 5. 645, Atque] uel inde ss. v Inde s, that this other manuscript was not one of the representative X manuscripts as selected by Magnus, but rather some ancestor of the inferior codices cited by Heinsius, the readings of which are now pushed back several cen- turies. Now Magnus, in his previous discussion of u (loc. tit., pp. 627 ff.), had correctly associated the fragment with the M branch of O, as I found after working out my own conclusions from his apparatus. He notes agreements with s and X but pertinently remarks that these are agreements in correct readings, implying thereby that not correct readings but errors determine groups among manuscripts. In short, here is another generally accepted principle of criticism which he formerly followed but which he now thinks 60 E. K. Rand best to abandon (see Introduction, p. xvii). In so doing, he has darkened counsel. So far, then, as these tests allow us to see, I believe that a great deal is yet to be done with the text of the Metamorphoses, in tracing its history in the Middle Ages, in distinguishing the classes of manu- scripts, and in estimating the effect of mediaeval glossaries. A con- spicuous and disheartening fact is that there is not a single important manuscript of the poem which has the quality of R of the Ars Ama- toria and Remedia or P of the Heroides and the Amores. These codices luckily preceded the period of the eleventh and twelfth centuries, well called by Traube the Aetas Ovidiana, when Ovid's intense popularity resulted in the elaborate glossing and collating and contaminating of his text. But as we have seen, this process had already begun in the early Caroline period, if not before in Ire- land. The Berne fragment, a, which Magnus regards as sui generis among the manuscripts of the Metamorphoses, is by no means in the same category with R and P of the Carmina Amatoria. It should be compared rather with of the Ars Amatoria, which is also Insular and also permeated with glosses. This leveling influence of glosses in all our manuscripts of the Metamorphoses should lead us to use more freely the readings of X and even s; M and N cannot be treated with the veneration appropriate for a text the only errors in which are due to scribal incapacity. Magnus has reformed pro- cedure considerably in this respect, but I believe we should go farther still. We need first a more thorough scouring of the selva oscura of glosses and classes. Then will be the time for some scholar with the genius and fine taste of Heinsius to make on the basis of ample and rationally sifted material a really critical edition of the Metamorphoses. For the present, the thesaurus of textual material which Magnus has accumulated after years of patient toil will be an indispensable source of information for the critic and should stimulate research. Habvakd University