University of Virginia Library

A Possible New Source for Servius Danielis on Aeneid III-V
by
Arthur Frederick Stocker

THE TEXT FOR THAT FORM OF THE SERVIAN commentaries on Vergil which is known from the name of its first editor[1] as Servius Danielis rests on an exceedingly slim manuscript foundation. Moreover different manuscripts constitute the basis for the text on the Eclogues and Georgics, Aeneid I-II, and Aeneid III-XII, respectively, from which it may perhaps be inferred that there occurred at some point in the tradition a division into what became archetypal volumes, whence the text descended along separate lines. A careful description of these manuscripts has been published by Professor J. J. H. Savage.[2]

The extent of Servius Danielis is considerably greater than that of the vulgate Servius. The relation between the two can best be seen by examining the pages of the single published volume of the so-called "Harvard Servius,"[3] the only edition purporting to do full justice to both forms of the commentary without prejudging their respective claims to venerability. It will be found that they have a great many scholia in common. These the Harvard editors print the full width of the page. Others are peculiar to one or the other form of the commentary—many (accounting for its greater total length) to Servius Danielis, a few to the vulgate Servius. These the Harvard editors print in columns three-quarters the

width of the page, Servius Danielis flush with the left margin, vulgate Servius flush with the right. Still others embody what appears to be parallel comment on the same passage in Vergil, and are printed side by side in half-width columns, Servius Danielis again on the left and vulgate on the right.

From the moment of the appearance of Servius Danielis there has been the greatest diversity of opinion concerning the origin and value of the additional scholia. Daniel himself thought that his Servius auctior was, if not the genuine commentary of Servius, at least a better representative of it than the vulgate, which he viewed as an abridgment, like Paulus' of Festus, that had driven the original almost or entirely out of circulation. Thilo, on the other hand, whose late nineteenth-century edition of Servius[4] has hitherto been standard, saw in the Danieline additions merely interpolations gleaned from various sources and more or less artfully grafted upon Servius. From this understanding of the matter he took license to butcher Servius Danielis sometimes beyond recognition in tailoring it to fit the convenience of a vulgate-centered edition.

It may be taken as established by modern scholarship[5] that the new material presented by Servius Danielis is not a miscellany of ancient lore, but derives from a single source, the antiquity of which is attested by, among other things, the large number of citations of authors whose works were quite certainly no longer extant even in the early Middle Ages. Further, this source seems quite clearly to have been an entirely different commentary on Vergil, some characteristics of which, as distinguished from those of Servius', can still be discerned. To mention only an obvious one, cross-references within the additamenta show that the unidentified commentary treated the works of Vergil in the order Eclogues, Georgics, and Aeneid, whereas Servius dealt with them in the sequence Aeneid, Eclogues, and Georgics. Servius Danielis, then, represents a contaminatio, or fusion, of the genuine (essentially the vulgate) commentary of Servius with some other ancient commentary, of which we may say that it is probably more ancient than Servius' own, probably one from which Servius himself drew extensively

enough so that there were long passages of coincidence suggesting the possibility of combining the two in a composite edition, and probably not the lost commentary of Aelius Donatus on Vergil, as Professor Albert H. Travis has shown by a careful comparison of the style of Servius Danielis with that of Aelius Donatus' extant commentary on Terence.[6] Adopting the terminology of Barwick and Rand,[7] we may call the non-Servian forebear of Servius Danielis D, the vulgate Servius S, and Servius Danielis, as a composite of the two, DS.

If what has been said about the nature of DS is true, and if DS possesses, as it were, an integrity of its own, it becomes manifestly important that its text be constituted with all possible accuracy, a thing that Thilo did not even aspire to do. The attempt to accomplish this, however, runs immediately a-foul of the difficulty inherent in the small body of manuscript evidence. For Aeneid III-V, to which I confine myself in this study, there is only one complete DS manuscript, the Bernensis, bibl. publ. 172 (saec. IX/X), which Thilo designates by the siglum F. Others provide only occasional help: a Leidensis, bibl. publ. Voss. F 79 (saec. IX ex.), which contains an exceedingly meagre abridgment of DS text; a "Virgil of Tours" (Bernensis, bibl. publ. 165, saec. IX),[8] in the margins of which, along with a great deal of miscellaneous matter, are found bits of DS; and a puzzling manuscript of Cassel (bibl. publ. ms. poet. fol. 6, saec. IX/X), which merits an article by itself for the DS affinities which it exhibits in what, by "Harvard" usage described above, would be common or parallel text, but for the most part does not include the DS additions to the Servian vulgate.

There is a fifth manuscript, another Bernensis, bibl. publ. 167, approximately contemporary with F (saec. IX/X), which has been found to contain a little over half of the DS scholia found in F. Its value Thilo specifically denies, stating (without proof) that it is, in a very unusual way hereinafter to be described, a copy of F,[9] and in this he is supported by Savage, at least so far as the Aeneid is

concerned.[10] A contrary view, however, has been held by Hagen[11] and Funaioli,[12] working with the commentary on the Eclogues and Georgics, and, in view of the new importance assumed by DS in the light of the discoveries of Barwick and his successors,[13] it seems worth while to re-examine the whole question of the relationship between these two manuscripts with the hope of vindicating the authority of Bern. 167 (which may be called G) as a new witness to the DS text for the Aeneid.

For purposes of this inquiry, I have in my possession photographs of both manuscripts for Aeneid III, 1-707, and Aeneid V, 9-826,[14] within which limits I have personally re-collated their text. I have also been aided by having the photographs of all the other important Servius manuscripts for Aeneid III and V, both the DS and the S, so that I have been able to weigh the testimony of F and G in the light of the full tradition. Finally, loan has been made to me of some unpublished notes of the late Professor E. K. Rand, of Harvard University, and Mr. George B. Waldrop, of Shady Side Academy, Pittsburgh, who once undertook but did not complete a similar study, using a much smaller section of text[15] as their sample. To these and to conversations with both of the gentlemen mentioned I am indebted for many stimulating suggestions.

F and G have both been described by Savage.[16] F has text in three columns. The center column is occupied by the text of Vergil, with ordinarily twenty-four lines to a page. The outer columns contain the scholia of Servius Danielis. The scribe followed no principle that I have been able to discover in assigning scholia to one column or the other. Since the order of the scholia was in most cases clearly fixed by the order of occurrence in Vergil

of the words that were being commented upon, it was not important that he should. His sole consideration seems to have been to transcribe each note as near as possible to the line of Vergil in which the lemma occurred, and he used whichever margin afforded the requisite amount of space. G, on the other hand, has text in two columns, the inner one (as the book was bound) containing the text of Vergil, usually with thirty to thirty-three lines on a page, and the outer one something more than half of the DS scholia ad loc. At first glance, the selection seems quite arbitrary, but as long ago as 1867 it was noted by Hermann Hagen, cataloguer of the manuscripts at Bern,[17] in his edition of the so-called Bern scholia to the Eclogues and Georgics,[18] that, with very few exceptions,[19] the scholia in G are the ones found in the left margin of F. So scrupulous was the scribe of G to reproduce all of these that on occasion, when space failed him, he went so far as to transcribe on scraps of parchment, now interlarded with the gatherings of G, notes for which otherwise he would have had no room,[20] and we may theorize that on other occasions, when there is a break in G in what would be the sequence of scholia from the left margin of F, he did the same thing, but the scraps have since become lost.[21] Scholia found in the right margin of F, however, he disregarded so systematically that I have found only three of them in G within the entire compass of Aeneid III and V.[22]

Over the intriguing question why and under what circumstances anyone should so undiscerningly have perpetuated about half of the DS commentary and ignored the rest it is not necessary to linger. Perhaps the scribe had in mind some other manuscript, in which there really was a distinction between scholia of

one sort in one margin, and of another in the other. It is, of course, true that the present state of G would be explained if the scribe who made this unfortunate selection had been copying directly from F. The distribution of scholia between the left and right margins of F, however, is quite as likely to derive from F's source as to have been original with F itself, and the coincidence which we observe proves nothing more than that the two manuscripts are in the same line of descent. The same may be said of the large number of demonstrably corrupt readings which they share. In Aeneid III and Aeneid V alone, I have noted several hundred, of which I cite only a few as examples:

• A. III, 2 (Th. 333, 10) visum] usum F G
• A. III, 20 (Th. 339, 12) qui arcis] quartis F G
• A. III, 104 (Th. 359, 12) Electræ] electrace F G
• A. III, 113 (Th. 363, 21) strictis] stratis F G
• A. III, 140 (Th. 369, 10) animas dulces linquebant] animæ eos linquebam et F G
• A. III, 167 (Th. 373, 11) finitimas] finiamas F G
• A. III, 202 (Th. 378, 1) qui Aeneas] quinea F G
• A. III, 209 (Th. 379, 14) symplegadas petras] simpligiadas petros F G
• A. III, 246 (Th. 385, 3) de diris] dederis F G
• A. III, 246 (Th. 385, 9) verum cum] acrum F G
• A. III, 319 (Th. 397, 10) cur] cui F G
• A. III, 330 (Th. 399, 24) Pyrrho] porro F G
• A. III, 332 (Th. 401, 17) profectos a Creta] profecto sacrata F G
• A. III, 370 (Th. 408, 3) virgo] uiro F G
• A. III, 402 (Th. 414, 20) eum] enim F G
• A. III, 450 (Th. 421, 23) pro non] ponam F G
• A. III, 544 (Th. 434, 19) belli omen] bellionem F G
• A. III, 590 (Th. 441, 26) moratum] montium F G
• A. III, 657 (Th. 448, 29) meminit] memirit F G
• A. III, 697 (Th. 455, 14) oraculis] oculis F G
• A. V, 28 (Th. 592, 17) maritima nata declamatio] mira nata declinatio F G
• A. V, 30 (Th. 593, 12) a Criniso] carinoso F G
• A. V, 45 (Th. 595, 6) Anchisen] andiisen F G
• A. V, 49 (Th. 596, 6) decem] autem F G
• A. V, 81 (Th. 602, 9) infra omnes] intra homines F G
• A. V, 122 (Th. 608, 13) in tergo] interrogo F G
• A. V, 287 (Th. 618, 9) de hastis] beatis F G

---

135




• A. V, 467 (Th. 628, 8) et Terentianus] oratius F G
• A. V, 553 (Th. 633, 19) quosdam] quos dixit F G
• A. V, 613 (Th. 637, 19) amœna] acta mœna F G
• A. V, 810 (Th. 650, 21) magnoque] manoque F G

They establish beyond peradventure the existence of a close family relationship between F and G, but they afford no support for the contention of Thilo and Savage, that G is a copy of F.

It is, as every textual critic knows, exceedingly difficult to prove that one text was taken from another, in the absence of direct evidence on the point. There are at least two minimum requirements, one positive, the other negative. Positively, there must be a substantial number of false readings in the supposed copy which could most readily be explained from some ambiguity or peculiarity that might plausibly be considered unique in the supposed original. Negatively, there must be an absence in the alleged copy of correct readings not found in the suspected source (except such as might be the product of scribal emendation). Even if these two conditions are met, the hypothesis remains vulnerable to attack from many quarters, as, for example, if it can be shown that the "copy" contains any considerable number of divagations that would not normally result from copying this "original," but would be likely if the scribe had been looking at another kind of script.

Of "positive" indications that the scribe of G was copying from F, a few can be cited:

• A. III, 67 (Th. 349, 20) possessio] F possessi G. There is a dot under the 'o' in F, probably accidental, which might easily be mistaken by a copyist for the sign of deletion.
• A, III, 67 (Th. 349, 21) nisi . . . nutriatur (23). SVPREMVM (350, 15) non . . . clamat] F om. G. This is an exact line in F.
• A. III, 68 (Th. 350, 16) CIEMVS] F Genus G. The 'C' and the 'i' run together in F so as to resemble a 'G' in square capitals.
• A. III, 93 (Th. 358, 1) SVMMISSI] Summis si F Summissi si G. F's only error is in word division, but a copyist, seeing "Summis," might easily have corrected it from Vergil into "Summissi," and then retained the detached "si."
• A. III, 135 (Th. 368, 12) SVBDVCTAE] F subducitæ G. F uses the familiar 'oc' form of 'a', which might have deceived a copyist into thinking that the preceding letter, actually a 't', was an 'i', while the cross stroke went with the first loop of the 'a' to form a 't'.
• A. III, 229 (Th. 382, 26) veni . . . rursum prius (27)] F om. G. The omission

  ---

  136

  
  
  
  is easy, by homœoteleuton, but particularly if one is copying from F, where "rursum ueni interdum" occurs at the beginning of one line and "rursum ueni etiam" at the beginning of the next.
• A. III, 272 (Th. 388, 22) LAERTIA] F Tertia G. F uses a rustic capital 'L', topped with what might easily be mistaken for the horizontal stroke of a rustic capital 'T'. It also represents the 'AE' (here, of course, not a diphthong) by 'e' with a cedilla, which mediæval scribes, notoriously careless about diphthongs, were prone to overlook.
• A. III, 337 (Th. 403, 1) prosperi . . . præstitit (3)] F om G. This again is an exact line of F.
• A. III, 351 (Th. 404, 15) aut prius] F autem G. Over the 't' of F, in the line above, happens to stand a 'p' with the cross stroke through its shaft, the regular symbol for "per." A copyist might take this for a horizontal stroke over the 't' of "aut," which would be the regular suspension for "autem."
• A. III, 420 (Th. 417, 22) imitatur] F mitatur G. Over the 'i' of F, in the line above, is a 'q' with so long a shaft that the 'i' looks not unlike a continuation of this shaft.
• A. III, 500 (Th. 428, 25) Roma] F romam G. A loop in the symbol for "pro," occurring in F directly over the 'a' of "Roma," might have been mistaken by the copyist of G for a stroke over the 'a', which would be the normal suspension for "Romam."
• A, V, 245 (Th. 615, 26) navali] F naualim G. F again has been guilty only of wrong word division. It has "naualim" at the end of one line and "nesteus" at the beginning of the next, where clearly the division should be "nauali mnesteus." G has "naualim mnesteus." He might have taken "naualim" from F, just as he saw it, and then made the easy emendation of "nesteus" to "mnesteus."

None of these is conclusive. The two most significant are the cases (A. III, 67; A. III, 337) where G omits, for no discernible palæographical reason, exact lines of the text as it appears in F. The first of these is an uncommonly long line at the foot of the page, extending the width not only of the "left margin" but also of the center column used for Vergil. The second is a normal line, consisting of 46 letters. These two, however, are not the only cases of the apparently inadvertent omission in G of short bits of text that stand in F.[23] Disregarding some that result from homoeoteleuton (which might deceive any scribe, regardless of the text from which he was copying)[24] and a few that may have been deliberate (on account of the scribe's reluctance to attempt the transcription of Greek,[25] I have found four that closely parallel the ones which have been cited, and a fifth that partially does so:

• A. III, 46 (Th. 344, 18) seges . . . puerum] F om. (mg. suppl.) G. This is not a line of F. As spelled out in F, the omission is one of 47 letters.
• A. III, 48 (Th. 345, 2) victa . . . obrutum] F om. (mg. suppl.) G. Again, not a line of F and an omission of about the same length (46 letters in F).
• A. III, 141 (Th. 369, 21) consentiat . . . tantum] F om. (mg. suppl.) G. These words come from the middle of a long line in F. In F they total 46 letters.
• A. III, 477 (Th. 425, 18) Adelphis . . . dicit] F om. (mg. suppl.) G. Not a line in F. 52 letters.
• A. III, 241 (Th. 384, 21) quod . . . geritur] F om. (mg. suppl.) G. 20 letters at the end of a scholion, divided in F between two lines.

The first four of these are about of a length (47, 46, 46, 52 letters). They can be explained most easily on the assumption that the writer of G missed a whole line of his original. If that is so, his original was not F. If his original was not F, however, it must have been a manuscript very much like it, to account for the regularity with which he reproduces only those scholia that appear in the left margin of F. It must have been a manuscript which, like F, had space in the scholia column for about 46 letters to the line. It may even have been the source of F. We may theorize that in seven instances the eye of the writer of G skipped over a whole line in his exemplar. Two of these happen also to be whole lines in F, as would be natural enough if G's exemplar was the parent of F also. By the theory, the passage from A. III, 241, would be an incomplete line in the original of G. The evidence of omissions, then, would, if my reasoning is correct, serve to refute rather than to support the contention that G was taken directly from F.

The rest of the evidence comprising "positive" support for this

thesis carries little more conviction. "Possessi" (A. III, 67) may have been the reading that was seen by both F and G; G adopted it, F (copying with an eye for the sense) emended almost unconsciously to "possessio," but then perhaps caught himself and deleted the 'o'. Haplographies (A. III, 420; A. V, 245) and dittographies (A. III, 93) are common enough in all conditions of text, as are omissions by homoeoteleuton (A. III, 229). "Aut" and "autem" (A. III, 351) are very frequently confused in text transmission.[26] G's "subducitæ" (A. III, 135) suggests compellingly only that G was copied from some text using the 'oc' or 'cc' form of 'a', or perhaps the Beneventan form of 't', which we shall see later is likely on other grounds.[27] Only G's "Genus" (A. III, 68), "Tertia" (A. III, 272), and "romam" (A. III, 500) remain to point the finger at F, and on these not much of a case can be built.

Negatively, the postulate of direct descent fares no better. It is true that G makes few substantial improvements upon the readings of F. In view of the closeness of the relationship clearly existing between them, it should not be expected that it would. It is true, too, that the individual errors of G are much more numerous than the individual errors of F. F is the better manuscript. Apart, however, from what may be taken as easy scribal emendations, G is in some instances undoubtedly correct where F is in error:

• A. III, 10 (Th. 336, 3) Ilium] G illum F
• A, III, 16 (Th. 338, 11) felicitate] (i supra alt. e ss.) G felicitati F
• A. III, 246 (Th. 385, 11) volatu] G uoluto in uolutu (in ras.) F
• A. III, 274 (Th. 389, 4) Leucate] G leutate F
• A. III, 330 (Th. 399, 23) desponsatam] disponsata G dispontam F
• A. III, 396 (Th. 412, 15) Calabria] G calabro F
• A. III, 455 (Th. 422, 5) vocet] G uouet F
• A. III, 517 (Th. 431, 9) serenitas] G senenitas F
• A, III, 590 (Th. 441, 24) Cyclopas] G cydopas F
• A. III, 590 (Th. 441, 26) Cyclopum] G cydopum F
• A. III, 678 (Th. 451, 10) unde] G ut de F
• A. V, 1 (Th. 587, 1) fletur] G flectur F
• A. V, 122 (Th. 608, 17) dammæ] damme G clammæ F

---

139




• A. V, 137 (Th. 610, 11) haurit] G ahurit (in ahuris) F
• A. V, 229 (Th. 614, 16) teneant] G eaeneant (in æneant) F
• A. V, 295 (Th. 619, 2) Sallustius] salustius G saltius F
• A. V, 324 (Th. 621, 2) a] G ad F
• A. V, 489 (Th. 629, 11) pro nominibus] G pro pronominibus F
• A. V, 823 (Th. 652, 1) recepisse] G recipis se F
• A. V, 823 (Th. 652, 2 in app.) herbam] G hebam (in hesbam) F

Moreover, it contains a few passages of text that are not found in F at all:

• A. III, 394 (Th. 412, 9) vel . . . timeas] G om. F To be sure, the words come at the end of a line in G and might have been added later. There is, however, nothing in the photograph to suggest that they are the work of other than the original hand. Thilo accepts them as authentic DS.
• A. III, 475 (Th. 425, 3) Scytha] ferenti uel furenti id est flanti. id est digne habito uel exaltato add. G This is a hodge-podge and certainly not correct DS ad loc. The first part of it is a gloss on "ferenti" (A. III, 473), and the second on "dignate" (A. III, 475). Neither has anything to do with what goes before. But the important thing for our purposes is that G writes it directly after "Scytha," and takes two lines to it. He must have gotten it from his original, and it is not in F.
• A. III, 576 (Th. 439, 25) ructatur] uel eruit add. G
• A. III, 629 (Th. 446, 14) Ithacensis] uel ita fensis ulixes add. G Although both of the foregoing are sheer nonsense and occur at the end of a line of text where they might have been added subsequently, the photographs give no indication that they are the work of other than the original hand.
• A. V, 790 (Th. 649, 7) QVAM MOLEM ut . . . moles] (tantis pro tantas) G om. F This looks like a clear case of a scholion in G that is not in F—certainly not where G would have looked for it in F, if he had been copying from that source, for F's left hand margin opposite the "Quam molem" in Vergil is entirely blank. By way of qualification to the force of this example, it must be said that the page in F is badly faded and could not very satisfactorily be photographed. It is barely possible that the note may have appeared in some part of the page that cannot now be read.

Instances of these two types are clearly fatal to the hypothesis that G was taken directly from F.

To complete the refutation, it is possible to cite an imposing number of cases where G is corrupt and F affords no grounds whatsoever for misunderstanding. A conspicuous example occurs early in Aeneid III, where F carries in the left margin the note, HOSPITIVM ANTIQVVM aut carum . . . regi (A. III, 15; Th. 338, 4-6), followed by the note on DVM FORTVNA FVIT (A.

III, 16; Th. 338, 10). Between the two notes, as a sign of cross-reference from the lemma in Vergil, F has the symbol '·d·', so that the apparatus entry must read: Since the insertion of such signs of cross-reference between two scholia, continuously presented, is unusual, we should not be surprised to find confusion in any manuscript copied from F, but certainly not the corruption found in G: Other aberrations in G point to the presence in G's original of well known palæographical difficulties that are absent in F, e. g.,

• A. III, 172 (Th. 374, 24}
• A. III, 359 (Th. 406, 5} dicit] F (written out in full) dr G
• A. III, 414 (Th. 416, 7} (sc. "dicitur") To make this mistake, the scribe of G must have had before him an original in which "dicit" was abbreviated as 'dt', with the horizontal stroke for a compendium over the 't'. Minuscule 't' and minuscule 'r' are readily confused.
• A. III, 73 (Th. 351, 9) Asterien] F asarien G The letters 'te', run together, can look very much like a flat topped 'a', and G must come from something on that order to contain the mistake it does. In F, however, the 'st' is ligatured and a plain 'e' follows, leaving no room for misunderstanding.
• A. V, 59 (Th. 597, 24) Anchise] F atichise G The letter 'n' and the letters 'ti', in ninth century script and when laterally compressed, can sometimes look very much like one another, but they do not in F, where the 'n' is unusually distinct.
• A. III, 607 (Th. 444, 13) vastoque . . . iuvabit] & rł usque periisse iuuabit F & rł quisq; perisse iudicabit G The scribe of G must have been copying from an original that used "et rłq" rather than "et rł" for "et reliqua."
• A. V, 380 (Th. 624, 3) dicat alacer tamen] (with correct word division) F dica ala certam G
• A. V, 553 (Th. 633, 19) certe] F arce G
• A. V, 238 (Th. 615, 5) iaciam] iaceā F iacet G In all three of the foregoing, G's difficulty apparently derives from misunderstanding of the 'oc' Beneventan form of 'a' and the very similar Beneventan form of 't'. In the first example, this could very easily have resulted in the loss of the 't' (between two 'a's) in "dicat alacer." In the second, the writer of G mistook a real 'c' for the initial loop of an 'oc'-type 'a'. In the third, there is simple confusion between 'a' and 't'. In each case, the script of F is plain.

---

141




• A. III, 645 (Th. 448, 7) ratione] F ratiore G
• A. V, 80 (Th. 601, 19) bonos ter enuntiatum] honeste renunciatē F honeste nentiatē G Both of these cases reflect confusion in G resulting from the similarity between 'n' and 'r' in insular script. The mistake could not have been made by a scribe who was copying from F.[28]
• A. III, 631 (Th. 446, 17) iacere] a sobriis cubandum (culiandum G) ostendit temulentos iacere add. F G The interesting point is the divergence in G from F's "cubandum." F's 'b' is unmistakable, but 'b' in insular script can very easily be taken for 'li'. There is a very similar case on A. V, 15 (Th. 590, 7) where, at "obliquatque," F fell into the trap and reads "obbquat que." G does much better, with "obliquat q".

To conclude, G is not a copy of F. Its similarities with F are so striking, however, as to suggest that both were copied from the same source—more cursive in style and exhibiting, it may be, a mixture of Italian and insular traits. F was the more careful of the two copies. Hence it is that, when F and G diverge, F is much more likely to be correct, and that there are a substantial number of omissions in G of text that appears in F, as compared with rather few and dubious instances of the reverse situation. Clearly G is a new source for Servius Danielis on Aeneid III-V, although its relationship to F is so close that it cannot be expected to contribute very materially to the establishment of the text.

Notes