Cynthia Damon (ed.), Studies on the Text of Caesar's Bellum civile

Find Location in text

Main Text

pg 16IIIConstituting the text


Over the past century or so three radically different stemmata have been proposed as representations of the descent of the extant manuscripts from ω‎. The bipartite stemma drawn by Holder (1898) is adopted, with some differences of detail, by du Pontet (1900), Klotz (1926, edn. 2 1950),42 Fabre (1936, rev. edn. Balland 2006), and Mariner Bigorra (1959–61) for the BC, and by Bouvet (1949, rev. edn. Richard 1997), Andrieu (1954), Pascucci (1965), and Diouron (1999) for the non-Caesarian Bella. Hering drew a new bipartite stemma in 1963, Brown a tripartite stemma in 1972. The new stemmata do not reflect new evidence; indeed one major development in the study of the tradition has been the elimination of codices descripti, so that Hering and Brown propose constructing the text on the basis of either four (Hering) or five (Brown) manuscripts where Klotz and others use as many as eight.43 The rival stemmata, with current sigla and hyparchetype designations, are represented in Fig. 1.44 (The pg 17

Fig. 1

Fig. 1

hyparchetypes μ‎ (MU) and π‎ (TV) are relatively uncontroversial, although the precise shape of these families varies somewhat from stemma to stemma and Hering questioned the utility of π‎.)

Brown's tripartite stemma, which constitutes a rebuttal of both of its bipartite predecessors, was based on evidence from the BC. Below, using evidence from all of the relevant Bella, I offer a more robust argument for Hering's bipartite division, together with a discussion of the possibility of horizontal transmission between the μ‎ and ν‎ branches and a more precise statement about the place and contribution of V. But before launching into what will prove to be a long and involved discussion it is worth considering what is at stake in evaluating these stemmata.

To discover a stemma's branches one looks for agreement in error between manuscripts, but to reconstruct the archetype one looks for agreement, in good readings and bad, between branches. The stemma is most useful—that is, it permits reconstruction of the archetype most securely—when there is agreement between branches at the first split. The stemmata offered by both Hering and Brown place M and U (μ‎) in a branch separate from T and V (π‎), making agreements between them (whole and partial: MUTV, MTV, UTV, MUT, MUV, MT, UT, MV, UV) evidence of the archetype.45 (Obviously not all of these agreements will always and necessarily be evidence of the archetype; the source and distribution of readings in each passage need to be taken into account.) This, given the waywardness of S pg 18(see p. 88 below), is a significant difference: μ‎ and π‎ or constituents thereof agree against S nearly eight hundred times in the text of the BC.46 When these four manuscripts are all regarded as descendants of β‎, however, as in Holder's stemma, their agreement simply offers a reading with equal weight to that of S. So if either of the new stemmata is proved correct we will have recovered much more of the tradition's archetype.47

The evidence for Holder's β‎ is scanty and editors have struggled to justify the family in the face of substantial contradictory evidence.48 Its core (UT) is identical with the β‎ family of the BG,49 but in constituting the text of the BC and non-Caesarian Bella one has to take into account two additional manuscripts, M and S, which for (most of) the BG were codices descripti belonging to the α‎ family. For the BC and other Bella, by contrast, M and S are the oldest manuscripts. Placing M in the stemma as a sibling of U in the μ‎ family is uncontroversial (see p. 61 below), but the position of S has proved elusive.50

The argument that μ‎ and π‎ had a common ancestor β‎ rests on a small number of significant but small shared omissions (roughly a pg 19dozen, none longer than two words), a paltry harvest for six substantial books of prose containing in total more than fifty thousand words.51 These are supplemented by an equally scanty list of fairly trivial shared errors.52 Furthermore, all of the editors who adopt Holder's S vs. β‎ stemma acknowledge the existence of a connection between manuscripts in the σ‎ family (now represented by S alone, but formerly including NL) and those in the π‎ branch of the β‎ family (TV). Various explanations for this stemmatic anomaly have been advanced, including an extra-stemmatic source (y) for the good readings in μ‎ that cannot be due to innovation,53 contamination of π‎ from the family of σ‎,54 and identical independent innovations by S and π‎.55

Hering dismissed the omissions and errors as insignificant, particularly given S's tendency, already evident in the BG, to supplement and alter its exemplar (1963, 59–73).56 Accordingly, he dismantled β‎. In his bipartite stemma the principal families are μ‎ (represented by MU) and ν‎ (represented by ST).57 That is, he separated T from μ‎ and made it a sibling of S. The connection Hering posited between S and T (his family ν‎) against μ‎ allowed him to discard both the extrastemmatic source for good readings in μ‎ and the hypothesis of pg 20contamination of π‎ with readings from σ‎. But the new evidence he presented for ν‎ in error against μ‎ and vice versa (1963, 76–7) was itself scanty: one item where ν‎ has an error and μ‎ has preserved the archetype's reading (BC 3.105.1), four (two of them admittedly weak) where μ‎ has an error and ν‎ has preserved the archetype's reading (BC 3.60.5, BC 3.83.2, BAfr 54.5, BAfr 62.3).58

Brown, after a fresh collation of MURSNLTV and a reassessment of the evidence for the BC,59 declared the arguments for both β‎ and ν‎ to be untenable: 'there is no solid evidence for placing TV on the side of S, as Hering has done, or of MU, the traditional position of editors' (1972, 31). She therefore drew a tripartite stemma, with S, π‎, and μ‎ independently derived from the archetype.60

How is an editor of the BC to proceed? There does not seem to be much point in making a new collation; Brown has done this. Given that Brown deemed the evidence for the BC inadequate to justify Hering's ν‎, that Hering in his 1973 review of Brown could do no more for ν‎ than restate his claims about BC 3.105.1, and that both Brown and Hering considered the evidence for β‎ inadequate (Brown on the basis of the BC, Hering on the basis of the BC, BAlex, and BAfr), the one remaining line of approach would seem to be to revisit the evidence of the BHisp, adduced by Diouron in 1999 for β‎ and against both Hering's ν‎ and Brown's tripartite stemma.61

pg 21Before starting, a clear statement of what we are looking for will be helpful. The argument for a tripartite stemma in an uncontaminated tradition needs to show that two branches never share significant innovations against a reading in the third that is unquestionably archetypal.62 (A 'significant' innovation is an error that could not be corrected by a medieval scribe or an innovation that is unlikely to arise simultaneously in unrelated manuscripts or families.) Our attention will be focused on passages that might offer evidence of S and π‎ joined in innovation against μ‎ or μ‎ and π‎ joined in innovation against S, since the remaining possibility, μ‎ and S joined in innovation against π‎, has not been suggested as the basis for a stemma (see further n. 71 below). 'Never' is a difficult thing to prove, of course, especially when, as will become clear below, virtually every analysis involves relative probabilities rather than absolute yes/no results.

So we will also look for positive evidence for a tripartite stemma, namely, the presence of apparently unstable relationships. If the first split in the stemma has the three branches S, μ‎, and π‎, we will sometimes find μ‎ agreeing in a correctable error with S, and sometimes with π‎, and sometimes S will agree in a correctable error with π‎ against μ‎; the exact distribution will depend on the character of each family. These errors have been inherited from the archetype by shifting pairs of its descendants (μ‎ and S, μ‎ and π‎, S and π‎); the reading of the other branch in each case will be a correction. It is important to note that the errors relevant for this kind of demonstration are trivial errors, correctable by scribal conjecture, quite unlike the significant innovations mentioned above, which are by definition incapable of correction by scribal conjecture. For a positive demonstration of a tripartite stemma it is important that all three 'occasional relationships' be represented by a respectable number of instances. On the other hand, evidence of a necessary or stable relationship pg 22between two branches of the three will constitute an argument for a bipartite stemma. We will have a necessary relationship if two branches agree in an innovation when the third has a reading that is both good and archetypal (i.e. a reading that couldn't have been reached via conjecture), and a stable one if two branches agree with each other in innovation far more often than either agrees in innovation with the third.

The argument is perforce somewhat lengthy; if the relationships among our principal manuscripts were easy to see, we wouldn't still have three fundamentally different stemmata in play more than 150 years after Nipperdey started the discussion.

To justify looking at the evidence of the BHisp for β‎ and ν‎, and to show how difficult it is to reach an unqualified verdict such as 'never' or 'cannot', I begin with BC 3.105.1, the foundation of Hering's ν‎ family and therefore of his bipartite stemmta (μ‎ vs. ν‎). In his view this passage offers a separative error 'der allen Ansprüchen an Unbestreitbarkeit genügen dürfte' (1973, 764).

BC 3.105.1 reperiebat (sc. Caesar) T. Ampium conatum esse pecunias tollere Epheso ex fano Dianae.

ampium U: appium MS: apium TV63

Titus Ampius is not mentioned elsewhere in the corpus Caesarianum, and his nomen is relatively rare. This means, says Hering (1963, 77), that his name cannot have been supplied by conjecture, but must come from the archetype.64 However, the very rarity of the nomen might have led scribes to alter it, even independently, to the much more familiar Appius, particularly if in the archetype the first syllable was written ā-.65 Hering's explanation, based in part on unreliable reports about the reading of M, is that μ‎ preserved the archetype's pg 23ampium and that ν‎ changed it, either inadvertently or deliberately, to ap(p)ium. This is neater than the alternative hypothesis (needed for either the S vs. β‎ or the tripartite stemma) of an archtypal ampium preserved by μ‎ and independently altered to appium and apium by S and π‎. But it is hard to feel that this one example is enough to define a family, particularly since a more accurate collation shows that one has to assume identical independent alterations of ampium to appium in M and either ν‎ or S anyway.66 Further evidence seems desirable.67


1. The evidence for β‎

In this section we are looking for the agreement of μ‎ and π‎ in a significant innovation against an archetypal reading in S. Diouron (1999, XCIII) lists three omissions common to μ‎ and π‎ against S. (I give Diouron's text throughout section B, and report the readings of the individual manuscripts rather than of hyparchetypes. Bold font in the text indicates the relevant problem spot(s) of each passage.)

  1. 1. BHisp 5.2 cum Pompeius cum suis copiis uenisset

cum2 S: om. MUTV

If S preserves the archetype's reading here, either haplography or the pursuit of elegance might explain the omission of the second cum. But it is at least equally plausible that S inserted cum to justify the case of copiis. Both usages are Caesarian (Meusel 1887–93, 2.2280), but the few examples in BHisp are all of the cum copiis variety, including one pg 24in the preceding chapter (4.4). This passage offers no decisive evidence for an association in error inherited by μ‎ and π‎ from β‎, or for the preservation of the archetype by S.

  1. 2. BHisp 22.3 Duo reliqui (sc. legati Bursauonenses) … fugerunt et Caesari rem gestam detulerunt <…> et speculatores ad oppidum Ateguam miserunt.

et1 S: om. MUTV | post detulerunt lacunam indicauit Nipperdey

The first et is omitted by μ‎ and π‎. The series of apparently parallel verbs in the archetype does suggest the desirability of et after fugerunt. But that is no guarantee that S has the right reading and has it by transmission. Furthermore, Nipperdey identified a lacuna before et speculatores on the grounds that Spanish legati don't dispatch scouts. If he's right—both Diouron and Klotz accept his argument—S's et may be a superficial repair to a faulty archetype. But even if Nipperdey is mistaken, the alternative explanations offered above suit this passage, too: either omission by β‎ or supplement by S. No firm argument can be built on this foundation.

  1. 3. BHisp 22.6 cum bene magnam manum fecisset et nocturno tempore per fallaciam in oppidum esset receptus, iugulationem magnam facit (sc. Pompeianus quidam).

et S: om. UTV (M deest)

Here one can explain the omission of et as an error due to haplography after fecisset, or its addition as a remedy for the asyndeton between the two halves of the cum-clause.

These correctable omissions are all innovations that might have been inherited from a common exemplar, β‎, where S followed the archetype. But they might also have been inherited from the archetype, directly or through β‎ or ν‎, with S making an innovation to improve the text. In other words, in none of these passages do we have the decisive evidence we are looking for.

I turn next to Diouron's eight 'erreurs communes de MUTV' (1999, XCIV n. 27).

  1. 4. BHisp 1.4 si qua oppida ui ceperat (sc. Cn. Pompeius), cum aliquis ex ea ciuitate optime de Cn. Pompeio meritus ciuis esset, propter pecuniae magnitudinem aliqua ei inferebatur causa.

ui ceperat MUTV: uice parat S aliquis S: aliis MUTV ex ea MUST: om. V aliqua STV: alia qua U: alia quae M

pg 25This passage comes from a summary of Pompeian methods for increasing their power in Spain. If S preserves the archetype at the aliquis/aliis split, β‎ has made a careless error, perhaps taking cum as a preposition. But the alternative hypothesis that S, prompted by aliqua later in the sentence, emended the archetype's nonsensical aliis to aliquis, does not seem beyond belief. A possibly correctable error is not the kind of proof we need.

  1. 5. BHisp 3.5 Qui (sc. L. Vibius) cum ad Cn. Pompei praesidia uenisset, incidit id temporis ut tempestate aduersa uehementique uento adflictaretur.

id S: idem MUTV

If S preserves the archetype, β‎ has made a trivial slip, repeating the formula used at the beginning of the chapter (3.1 idem temporis) in an unsuitable context. But the correction from idem to id is not difficult, since there is no possible antecedent for idem in the vicinity, and the content of id is immediately supplied by the ut-clause; furthermore, the phrase id temporis occurs two sentences after our passage (3.7). This is another possibly correctable error.

  1. 6. BHisp 5.5 propter pontem coagulabant, fluminis ripas appropinquantes coangustati praecipitabantur.

appropinquantes edd.] ac propinquantes MUTV: ut propinquantes S

The context is a battle for control of a bridge. It is difficult to see how the distribution of readings here can represent anything other than a conjecture by S to repair the puzzling ac in the archetype preserved by μ‎ and π‎, unless it is simply a misreading. In any case the passage does not show β‎ in error against a correct and archetypal reading in S.

  1. 7. BHisp 5.6 Hic alternis non solum morti mortem aggerabant, sed tumulos tumulis exaequabant.

alternis ϛ‎: -rius ω‎| aggerabant S: exa(g)gerabant MUTV

This passage offers a nice illustration of the anonymous author's penchant for rhetorical effect, but it is difficult as evidence of transmission. If S preserves the archetype's verb (and incidentally gives us the earliest attestation of aggero), β‎ has innovated, either to improve the parallelism between the verbs or to replace an uncommon verb with a more common one. If μ‎ and π‎ preserve the archetype, it is hard to see why S tinkered. But there remains the possibility that the pg 26preceding -em caused the omission of the ex-. On balance, this seems more likely to be an innovation in β‎ against an archetypal reading in S than an innovation in S against an archetypal reading in μ‎ and π‎, but it is far short of decisive.

  1. 8. BHisp 6.3 Caesar munitionibus Ateguam oppugnare et brachia circumducere coepit.

(The text is corrupt in a number of interrelated places here, so I give the readings of S and MUTV in full to facilitate comparison. Orthographical variants are not recorded.)

caesar munitiones antiquas oppugnare et brachia circumducere coepit S caesar munitionibus antequam oppugnaret brachia circumducere coepit MUTV

The context is immediately after Caesar's dash to Ategua, mentioned in 6.1 (Ateguam proficiscitur, where MUTV have the town's name correctly and S reads ad teguiam). If MUTV represent the archetype reading in our passage, would S have had any inducement to tinker? Editors beginning with Aldus do tinker, altering antequam into a place name. But S, which thought the name of the place Caesar was besieging was Teguia, would probably not have seen a name lurking under the unexceptionable antequam. Still, the syntax and position of munitionibus are peculiar. Does it go with oppugnaret? If so, is it dative or ablative? And why does it precede the conjunction? All of the innovations in S, if they are innovations, serve the end of restoring to oppugnare its proper transitive construction (as, most recently, at BHisp 3.1). If, however, the formally unproblematic reading of S is that of the archetype, we have to assume either that β‎ was innovating without warrant, or that two alterations were made (changing the case of munitiones, and turning antiquas into a conjunction), not necessarily at the same stage, but cumulatively forcing a further change in the construction of the sentence (oppugnare et to oppugnaret). This concatenation of events seems extremely unlikely. Here the evidence, on balance, suggests that S deviates from the archetype while μ‎ and π‎ preserve it. This illustration of S's capacity for making deliberate and substantial innovations should be borne in mind as we continue.68

  1. pg 279. BHisp 18.3–4 Eodemque tempore signifer (sc. Pompeianus) de legione prima transfugit et innotuit, quo die equestre proelium factum esset, suo signo perisse homines XXXV neque licere castris Cn. Pompei nuntiare nec dicere perisse quemquam. (4) Seruus … dominum iugulauit, etc.

eodemque MUST: eodem V | innotuit S: non timuit MUTV | licere ς‎: -ret ω‎ | nuntiare ω‎: -ri ς‎ | dicere ω‎: -ci ς‎ | quemquam edd.: quamquam MUST: quaquam V | seruus MU: -uos STV

The transmitted text of this passage about the siege of Ategua is a mess, but for our purposes we only need to consider the innotuit/non timuit split. Non timuit does not fit the syntax. If it is the archetype's reading, S has substituted innotuit, a verb (barely) capable of introducing the indirect statement that follows. This is not a perfect repair, since the usage innotescere=notum facere required here is only attested in late and almost exclusively Christian texts (see TLL 7.1.1713.27 ff., esp. 1714.22–6). In texts of the classical period—all post-Caesar—innotescere means notum fieri and does not govern indirect statement.69 Other repairs can be imagined, beginning with Aldus' suggestion nuntiauit (cf. 18.6 insequenti tempore duo Lusitani fratres transfugae nuntiarunt Pompeium contionem habuisse). If S's rather peculiar reading is that of the archetype, β‎ has made a baffling innovation, perhaps a misreading. But given the improbability of innotuit being Caesarian, or even 'Caesarian', it is more likely that S tried to fix the nonsensical non timuit inherited by μ‎ and π‎ from the archetype. There is no evidence for β‎ here.

  1. 10. BHisp 28.3 Ita hac opinione fretus tuto se facere posse existimabat (sc. Pompeius).

tuto se Lipsius: totos UTV: totum S (M deest)

The text of U and π‎ makes no sense, that of S is not much better. It is hard to imagine totos arising out of anything but faithful copying of a corrupt archetype. S will then have replaced the nonsensical totos with something that makes apparent sense, even though, if pressed, the scribe would have found it difficult to say just what totum referred to. If S's totum is the reading of the archetype, β‎ will have made a baffling innovation. The former explanation is distinctly more pg 28satisfying, so we have no evidence here of an innovation in β‎ against an archetypal reading in S.

  1. 11. BHisp 30.1 … cum leui armatura milibus sex.

milibus S: militibus UTV (M deest)

The context is a battle-line description involving thirteen eagles, cavalry, and light-armed troops. S's milibus is a trivially easy correction if the archetype read, with U and π‎, militibus: hard to imagine what impact those six light-armed soldiers would have had! If S preserves the archetype, we have to assume that β‎ misread milibus. Since the error in U and π‎, if it is an innovation, is correctable, this passage gives us no clear evidence about the source of the innovation.

Diouron finds in these passages satisfactory proof of the β‎ family familiar from pre-Hering editions, and her stemma, as was noted above, is that of Fabre et al. In my view the evidence can't sustain the burden of proof. Only one passage (BHisp 5.6) of those here discussed had a better chance of being a disjunctive error of β‎ against S than the other way around. And that one was not at all decisive.

Before looking to see whether BHisp provides positive evidence of Hering's ν‎ family—Diouron, of course, denies this—let us look at one other BHisp passage possibly relevant to the question of whether β‎ is a family; Diouron does not include it in her lists of omissions and significant errors.

  1. 12. BHisp 1.4 … ut eo de medio sublato ex eius pecunia latronum largitio fieret.

de MUTV: om. S sublato S: -ta MUTV eius S: eo MUTV

This passage, which follows on example (4) above, illustrates Pompeian brutality even towards their supporters, including their habit, so very reminiscent of Sulla, of raising false charges against a rich man in hopes of profit. Diouron follows Klotz in printing MUTV's de with S's sublato and eius.70 If S preserves the archetype, β‎ has innovated at three spots, adding de before medio, and changing the gender of sublato (and hence its referent) and the case of eius. But once de has been supplied it is hard to see why the latter two changes would have been deemed necessary; modern editors don't make them. If, on pg 29the other hand, μ‎ and π‎ preserve the archetype, S has innovated in two spots (sublata and eo2) and made one error (omission of de). In this scenario the first change (making eomedio depend on the participle sublato) requires the second; the error is just S being S (on S's errors see p. 88 below). The latter explanation seems the better, so this passage does not give us evidence of an innovation in β‎ against an archetypal reading in S.

That concludes the review of the evidence made available since the publication of Hering (1963) and Brown (1972) for the existence of β‎. If solid evidence had emerged for the agreement of μ‎ and π‎ in an innovation where S has an unquestionably archetypal reading, both of the new stemmata would have been disproved. But in my view no such evidence has emerged.71

2. The evidence for ν‎

Diouron (1999, XCIX–C, nn. 43, 48) lists seventeen places in BHisp where S and π‎ (or occasionally S and T, with V going its own way, as is its wont; see chapter IV.C.5 below) are associated in error against μ‎ (or against U alone after M falls silent at 22.5). Ten of these are dismissed as being corrections by μ‎ or identical independent innovations by S and π‎.72 Reasonably so: they are the same sorts of errors as those dismissed by Brown (1972, 30). I would eliminate two more as being orthographical variants.73 The passages where the reading of μ‎ pg 30might be archetypal are discussed below. Where the case for μ‎'s reading as archetypal holds up, and the innovation cannot have arisen independently in S and π‎, we will have evidence for ν‎ and against both β‎ and the tripartite stemma in which μ‎ and π‎ are independent.

  1. 1. BHisp 4.4 Itaque Cn. Pompeius Vlia propre (sic) capta litteris fratris excitus cum <omnibus> copiis ad Cordubam iter facere coepit.

cn pompeius MU: facit ut STV | omnibus addidit Böhm

Just before this passage, with Cn. Pompeius beseiging Ulia and Caesar threatening Sextus Pompeius at Corduba, Sextus has written to his brother for help. Klotz (1927, XIII) considers Cn. Pompeius an intrusive gloss. (He nevertheless prints it, with the following note: 'an facit ut <frater uoluit et>?') Pascucci (1965, 75), on the other hand, considers facit ut a gloss for itaque. Diouron dubs the passage a 'cas épineux' (1999, CII), printing Cn. Pompeius although she deems facit ut the lectio difficilior. If μ‎ preserves the archetype, the text makes perfect sense, so it is difficult to see how S and π‎ ended up with facit ut, which doesn't make sense as is, even if with a further innovation such as that proposed by Klotz sense and syntax can emerge. If STV's facit ut is the reading of the archetype, the verb might certainly provoke a marginal gloss identifying its subject, since in the preceding sentence the 3rd person singular subjects were Sex. Pompeius in the main clause and Caesar in the subordinate cause. However, one then has to explain how the gloss eliminated its raison-d'être. With synonym glosses this is of course a regular occurrence, but here the substitution seems implausible. I am inclined to think that neither μ‎ nor STV preserve(s) the whole text of the archetype here. So no clear argument for ν‎ emerges.

  1. 2. BHisp 5.3 Caesar, ut eum ab oppido commeatuque excluderet, brachium ad pontem ducere coepit; pari item condicione Pompeius.

item Fleischer] idem MU, Klotz: denique S: dem T: om. V

Whether one prints idem with Klotz or item with Pascucci and Diouron, the best way to explain S's denique is to suppose a faulty archetype here, reading something like T's nonsensical dem, which S, V, and μ‎ tried to correct in different ways, with the innovation of μ‎ pg 31being the most successful. (If it is right, the word in the archetype lost its initial letter by haplography after pari.) So, as Pascucci argues (1965, 75), there are no good grounds here for seeing innovation by ν‎ against an archetypal reading preserved by μ‎.

  1. 3. BHisp 17.2 tua uirtute superati salutem a tua clementia deposcimus petimusque ut <…>

petimusque ut] petimusque et MU74: petimus ut STV75

The context is a petition addressed to Caesar. There must be a lacuna after ut (or et); the first person singular utterance that immediately follows in the manuscripts comes from Caesar's response: qualem gentibus me praestiti, similem in ciuium deditione praestabo. If anything can be made of this passage, it can only be that S and π‎ preserve the nonsensical archetype (ut does not work with praestabo), while μ‎ tries to patch it up by ending the petition after petimusque and beginning Caesar's speech with et. In any case there is no reason to think that the nonsense in S and π‎ is an innovation against an archetypal reading preserved by μ‎.

  1. 4. BHisp 24.376 … prohibiti a nostris sunt deiecti planitie. Quae res secundum nostris efficiebat proelium.

quae res U: quorum STV (M deest)

The enemy have just been repelled in their attempt to seize a hill. Klotz (1927, XII) explains U's good reading by recourse to y, his extra-stemmatic source, implying that the reading of S and π‎ is archetypal. (The extra-stemmatic source explanation is a last-ditch effort to defend β‎. It should not be accepted lightly.) Pascucci (1965, 75) suggests that both readings represent resolutions of an abbreviation; S and π‎ got it wrong, U got it right. But quae res is not the sort of thing that gets abbreviated in the archetype (see p. 55 below), and there are no comparable errors in the BC, at least. So we need another explanation. If the nonsensical reading in pg 32S and π‎ is that of the archetype—perhaps a word such as fuga had fallen out after quorum (cf. BAlex 20.4)—quae res is a surprising innovation by the relatively uninventive U (see p. 84 below); the absence of M here prevents us from ascribing quae res to μ‎, which does make a number of successful innovations (see p. 61 below). If quae res is an innovation, its basic form might have been suggested by haec res in the preceding sentence: 24.2 haec res necessario deuocabat ut ad dimicandum descenderet. The alternative is that U preserves the archetype and ν‎ makes a rather baffling innovation. This seems to me the best case for ν‎ innovating against an archetypal reading in U in BHisp, but it is not one that can be pressed very hard.

  1. 5. BHisp 31.10–11 Nostri desiderati ad hominum mille partim equitum, partim peditum; saucii ad D. (11) Aduersariorum aquilae sunt ablatae XIII et signa <…>, etc.

mille] ∞ UTV: co S ad D U: ad e- S (i.e. ad eaduersariorum): om. T: praeter V aquilae UST: maculas V ablatae UST: ablata V et signa UST: signa V (M deest)

The context is the 'cost of battle' assessment after Munda. A number seems clearly called for with saucii, so U's ad D may be a (rather feeble) conjecture. If it was the archetype's reading, one can say that T omitted it by a kind of haplography before aduersariorum. The reading of S is more mysterious, and V has made major changes here, but for our purposes these hardly matter: there is no sign here of an innovation in S and π‎ where U, the lone representative of μ‎ at the end of the BHisp, has an archetypal reading.

3. Discussion

The Bellum Hispaniense offers no decisive evidence for β‎ or for ν‎. At least one passage, however, BHisp 24.3, is more easily explained with Hering's stemma than with either of its rivals. The absence of decisive evidence for either bipartite stemma is of course a necessary argument in favour of a tripartite stemma, but it is perhaps not a sufficient one. That issue will be addressed in section D below. Meanwhile, we continue our search for ν‎ with an examination of the evidence for ν‎ alluded to but not discussed by Hering (see n. 58 above; the passage that he considered decisive, BC 3.105.1, was discussed earlier).

pg 33C. LOOKING FOR ν‎

1. Bellum ciuile

Brown's conclusion after her search for omissions in the BC as evidence of ν‎ is the following (1972, 31): 'the only STV omissions seemingly worthy of the name are 3.89.1 -iunxit and 3.95.2 fatigati. The paucity of their number gives rise to doubt, and to the suspicion that STV [i.e. S and π‎] omitted the words independently. Or it may be that these words were lacking in the archetype, and the reading of MU is a successful conjecture of the scribe of their common exemplar'. The first of the two passages mentioned is this:

  1. 3.89.1 huic [sc. legioni] sic adiunxit octauam ut paene unam ex duabus efficeret

adiunxit MU: ad STV

Huic shows the syntax to be lacunose, so this is an instance of a correctable error common to S and π‎; accordingly, I list it in section D. 3 below.77 The second needs a brief discussion.

  1. 1. BC 3.95.2 Qui (sc. Caesariani) etsi magno aestu fatigati—nam ad meridiem res erat perducta—tamen ad omnem laborem animo parati imperio paruerunt.

aestu fatigati MU: (a)estu TV: est S

If π‎ preserves the archetype, the absence of parallelism after etsi and tamen might well provoke innovation. The reading in S seems to be a casual error rather than an attempt to emend, but the reading of μ‎ is a different matter. First, fatigati is not a Caesarian word. On fifteen occasions he and his continuators use defatigo, never fatigo. And second, μ‎, like S although to a lesser extent (see above), seems to have a habit of supplementing its exemplar: of the fifty-odd places in the BC where μ‎ may have corrected an error in the archetype (see section D. 3 below), about six involve small supplements (cf. also BC 3.101.2 ad STV: apta ad MU: aptae ad ed. pr.; Hering [1963, 75] lists roughly fourteen more for BAlex and BAfr, and we just saw another at BHisp 17.2). If, on the other hand,μ‎ preserves the archetype, S and π‎ are associated in error and we have evidence of ν‎. Both explanations pg 34are viable, neither is decisive. I therefore list this passage with the correctable errors of S and π‎ in section D. 3 below.

2. Bellum Alexandrinum

Andrieu reviews the places where S and π‎ seem to be joined in innovation against an archetypal reading in μ‎ in the Bellum Alexandrinum (1954, LXXVI–LXXVIII). He is trying to assess the likelihood of contamination between S (his σ‎) and π‎, not looking for ν‎, but his examples will nevertheless serve our purposes. For some of μ‎'s good readings he posits identical independent errors arising in S and π‎ from misguided innovations, as at 33.3 legiones … ueterana sexta MU: legiones … ueteranas sexta STV, where he sees two potential sources of error (LXXVI): 'faux accord avec legiones' or 'suggestion de -s de sexta' Other such errors he ascribes to easily explicable misreadings, as at 26.1 missus MU: missis STV, where he sees a faulty interpretation of an abbreviated form, 'faite deux fois' (LXXVII). Similarly 10.4 occurrerunt MU: occurrerant STV, where the pluperfect is a 'faute banale sur la minuscule qui a pu se produire deux fois' (LXXVII). I myself would be more inclined to analyse these as archetypal errors correctable by μ‎, but it hardly matters, since there is no decisive evidence of ν‎ here. The most telling of the Balex passages mentioned by Andrieu and Klotz are the following two. (I give Andrieu's text.)

  1. 2. BAlex 19.2 fortiorem illum (sc. pontem) propioremque oppido Alexandrini tuebantur. Sed eum postero die simili ratione adgreditur (sc. Caesar), quod, his obtentis duobus (sc. pontibus) omnem nauigiorum excursum et repentina latrocinia sublatum iri uidebat.

fortiorem MU: certiorem STV | sed MUV et T per compendium: si S | quod (h/i)is MU: his STV | o(p/b)tentis MUTV: optenas S ut uidetur | sublatum [-ta U] iri MUTV: sublatuiri S78 | uidebat T. Bentley: -atur MU: -antur STV

Caesar has taken Pharos Island and is trying to secure it and its two bridges, one of which is still controlled by Alexandrian forces. I leave pg 35aside the string of unique innovations in S in order to focus on associations in innovation. At the first split in transmission, if STV's certiorem is the reading of the archetype, it is hard to see why μ‎ would alter it to fortiorem; the fact that the TLL offers no good parallel for certus modifying something like pons seems beyond the ken of a medieval scribe. On the other hand, the implausibility of deliberate alteration is even more pronounced if μ‎ preserves the archetype, since there is nothing obviously wrong with fortiorem.79 This seems likely to be one of roughly a dozen places where we have evidence of variants in the archetype (see p. 58 below).80 Archetypal variants render the distribution of readings an unreliable guide to relationships, since our hypothetical β‎ or ν‎ might have preserved both variants, whence their descendants might have made individual selections that align randomly.

The split between μ‎'s quod (h/i)isuidebatur and STV's hisuidebantur is somewhat more informative, since the plural verb doesn't construe with sublatum iri and quod doesn't work well with either verb form. Both words are therefore more likely to have been in the archetype than to be innovations. (Videbantur might be an innovation, but if so it is a careless recall of the plural subject of tuebantur and therefore not the sort of innovation to occur independently.) Quod cannot be explained as a casual mistake. This suggests that the archetype read quod(h/i)isuidebantur, of which μ‎ altered the verb and ν‎ omitted the unwanted conjunction. The alternative, proposed by Andrieu, is that S and π‎ preserved the archetype while μ‎ made a two-part emendation, supplying quod to set up its singular uidebatur. Andrieu calls this a 'leçon vraie et inauthentique' (1954, LXXVII) but he doesn't print it. After all, quodsublatum iri uidebatur—'because … it seemed that there was a going to stop (raids)'—cannot be called neat; it is easier to see this text as the result of a superficial repair such as changing uidebantur to uidebatur. This is all more complicated than I would like, but if the analysis is correct, we have evidence for both ν‎ innovating against archetypal μ‎ pg 36 (the omission of quod), and μ‎ innovating against archetypal ν‎ (the change from uidebantur to uidebatur). However, this is hardly the sort of decisive evidence we need.

  1. 3. BAlex 51.1 Interim litteras accepit (sc. dux Caesarianus quidam) a Caesare ut in81 Africam traiceret … quod magna Cn. Pompeio Iuba miserat auxilia maioraque missurus existimabatur.

maioraque MU: maturaque STV

The reading of S and π‎ is perhaps superficially satisfying, but it is also troubling enough to provoke innovation: S's descendant N altered it to matureque. If it is the reading of the archetype, maioraque is a surprisingly deft emendation by μ‎, satisfying syntax, sense, and rhetoric by changing just two letters. Klotz, for one, finds this implausible (1927, XI–XII). If μ‎ preserves the reading of the archetype, the innovation by S and π‎ is hard to explain. Andrieu labels it a twice-perpetrated misunderstanding of a minuscule exemplar's maioraque (1954, LXXVII), but it seems implausible to me that a word that seems almost inevitable in its context should be misread in exactly the same way twice. This passage is better explained as an unconscious error transmitted by ν‎ to its descendants S and π‎, while μ‎ preserves the archetype. This is the clearest evidence for ν‎ in the Bellum Alexandrinum.

3. Bellum Africum

As was mentioned above, Klotz and Bouvet list quite a number of significant associations in error by S and π‎. The passages offering the strongest evidence for ν‎ are the following. (I give Bouvet's text):

  1. 4. BAfr 10.2 Omnibus in exercitu insciis et requirentibus imperatoris consilium, magno metu ac tristimonia sollicitabantur.

insciis MUTV: inscitis S | tristimonia STV: tristi M: tristicia U

The context is a raid by Caesar and a small force of seven cohorts. The men are baffled by their leader's strategy. Tristimonia is a rather uncommon word, unlikely to have arisen independently in S and π‎. If it is the reading of the archetype, μ‎ has deviated, perhaps substituting the bland adjective tristi (to modify metu in tandem with magno), pg 37which M preserved and U altered to a substantive, tristi(c/t)ia, more banal than the original. This scenario is compatible with all three stemmata, since we have ν‎ making a solo innovation. If M or U preserves the archetype, there is nothing to provoke alteration (blandness and banality being what they are), so it is hard to see how tristimonia arose. The second scenario would be evidence for ν‎ innovating against μ‎, but the likelier first scenario is, by contrast, a supplement to Hering's short list of Trennfehler of μ‎ against archetypal ν‎ (1963, 77; see p. 20 above).82

  1. 5. BAfr 10.4 Huic adquiescebant homines et in eius scientia et consilio omnia sibi procliuia omnes fore sperabant.

huic MU: hic STV | homines MUST: omnes V

This comes later in the same passage, when the men's worries have been allayed. Both readings construe but huic has more point (Caesar's confidence was mentioned in the preceding sentence). On balance, it seems more likely that the archetype read huic and that STV's hic is an involuntary error, certainly not the sort of thing that would occur independently in S and π‎. But it doesn't seem possible to rule out a conjectural improvement by μ‎. This passage therefore counts as modest but not decisive evidence for ν‎ in error against archetypal μ‎.

  1. 6. BAfr 18.1 Interim M. Petreius et Cn. Piso cum equitibus Numidis ∞DC electis peditatuque eiusdem generis satis grandi … subsidio suis occurrunt.

∞DC ex 19.4] 1C MU: CCC STV | subsidio MSTV: -dia U | occurrunt MUT: concurrunt S: procurrunt V

This is one of many instances of misread numbers (see p. 56 below and cf. BAfr 38.3 ∞ MU: co STV, BHisp 31.10 mille] ∞ UTV: co S). Both readings are wrong, historically speaking, so it's likely that one of them comes from the archetype. STV's CCC is more likely to arise as a misreading of μ‎'s ∞C than vice versa, although it could also be a deliberate alteration of the form ∞C, which has no parallels in the Caesarian corpus. Could the error have arisen independently in S and pg 38π‎? I don't think one can rule it out. (At BHisp 31.10, for example, π‎ and S differ in just such a reading.) So this passage gives us only weak evidence for ν‎ innovating against archetypal μ‎.

  1. 7. BAfr 18.2 Atque hostes, suis ex terrore firmatis, rursusque renouatis animis, legionarios conuersis equitibus recipientes nouissimos adoriri … coeperunt, etc.

atque hostes MU: at cohortes STV

This comes immediately after the example above. There is nothing to provoke deliberate alteration in either reading. Editors have two reasons to prefer the reading of μ‎. First, given the immediately preceding description of the troops involved, it is hard to see who, precisely, these cohortes were; the last cohortes mentioned (at 17.1; they will appear again at 18.4) were those of Caesar, whereas these are Pompeian troops, hostes. And second, the author of the BAfr does not use the word at. However, neither of these is likely to have motivated a medieval scribe to alter at cohortes to atque hostes. So μ‎'s atque hostes is likely to be archetypal, not conjectural.83 Klotz explains the reading of STV as a misreading of the abbreviation atq. hostes (perhaps with help from the cohortes in the vicinity). Since this is not the kind of innovation that would arise independently in S and π‎, we have here a little more evidence for ν‎ innovating against archetypal μ‎.

  1. 8. BAfr 18.4 Caesarisque equites iumenta ex nausia recenti siti, languore, paucitate, uulneribus defatigata ad insequendum hostem perseuerandumque cursum tardiora haberent, etc.

nausia MUS: -ias TV | recenti MUT: -tis SV | siti MUTV: sitis S | uulneribus defatigata [fa- V] MUTV: d- u- S | tardiora MU: tardiorem STV

In the same passage, a little further on, Caesar's cavalry is hard pressed and failing fast. It is much more likely that ν‎ altered the archetype's long-delayed tardiora to tardiorem, giving it a nearer prop in cursum, than that μ‎ changed tardiorem to tardiora and thereby restored the adjective to iumenta. Here, however, one has to pg 39admit the possibility of identical independent innovations in S and π‎. So this passage, while consistent with a ν‎ vs. μ‎ stemma, doesn't offer good evidence against either of its rivals. (The apparatus here shows nicely how widely distributed and frequent innovations are in this manuscript tradition.)

  1. 9. BAfr 20.2 audiebat enim (sc. Caesar) Scipionem … adpropinquare, copias suas cum Labieno et Petreio coniungere; cuius copiae legionum VIII … esse nuntiabantur.

cuius MU: quorum STV | nuntiabantur UScTV: -atur MSac

At the cuius/quorum split both readings construe but that of μ‎ is both historically true and less likely to have been reached by deliberate alteration than is STV's quorum, which looks back to the closer of the possible antecedents (Labieno et Petreio, rather than Scipionem). Cuius may well be archetypal, but, since here too identical independent innovations by S and π‎ must be considered possible, this passage offers only weak evidence for ν‎.

  1. 10. BAfr 26.5… principesque ciuitatum aut interfici aut in catenis teneri, liberos eorum obsidum nomine in seruitutem abripi; his se miseris suamque fidem implorantibus auxilio propter copiarum paucitatem esse non posse.

principesque MTV: principes US | abripi his se Oudendorp] obripi (h/i)is se McU: obripuisse84 Mac: abripuisse STV | miseris Wölfflin] in miseris S: in miseriis McUTV: miseriis Mac: 'an in miseriis <constitutis>?' Klotz | esse U: se esse M: sese STV

This passage comes at the end of Caesar's reflections on a long list of provincial complaints about Pompeian abuses, all of which are expressed with present passive verbs. The textual problems clearly go back to the archetype, which has been variously tinkered with. The important error for our purposes is STV's abripuissesese. The perfect infinitive seems to have arisen from a misreading of its exemplar, the pronoun either as a misreading or as a deliberate innovation to supply a subject for abripuisse.85 That S and π‎ made two identical pg 40misreadings independently is implausible, even if Mac goes similarly astray with the infinitive. (In the BC, at least, Mc's corrections always reunite it with U, even where (as here) the reading of U is problematic; see p. 80 below. So the error was probably not transmitted to Mac from μ‎ and to μ‎ from the archetype.) And the repair, if it is a repair, is unlikely to have been made independently since it leaves auxilio an orphan. (M's se esse is more successful.) Furthermore, the reading of μ‎ is unlikely to be a correction; none of μ‎'s other innovations is as far-reaching as this (see p. 48 below). The combination of an archetypal reading in μ‎ and a shared innovation in S and π‎ is the evidence of ν‎ that we have been looking for. Decisive? probably not, given the layering of innovations here.

A brief addendum on the question of whether the archetype read obripi or abripi. The former is a very rare verb (TLL–84). Abripi might be a normalizing innovation, but on the whole it seems more likely that the prefix ob- is a carry-over from obsidum earlier in the sentence. If so, this is another instance of μ‎ innovating against ν‎.

  1. 11. BAfr 32.3 Interim Numidae Gaetuli<que> diffugere cotidie ex castris Scipionis et partim in regnum se conferre, partim quod ipsi maioresque eorum beneficio C. Mari usi fuissent Caesaremque eius adfinem esse audiebant, in eius castra perfugere cateruatim non intermittunt.

<-que> edd. | usi MU: om. STV

If S and π‎ preserve the archetype at the split in the tradition here, μ‎ has made an impressive emendation. This seems rather unlikely, however. If μ‎ preserves the archetype, ν‎ has omitted usi by error, perhaps under the influence of mari; the omission is unlikely to have been made by S and π‎ independently. This passage is in my view the best evidence in the Bellum Africum for ν‎ innovating against an archetypal reading in μ‎.

  1. 12. BAfr 35.2 Itaque ex eius (sc. Caesaris) patientia in magnum timorem coniecti (sc. Pompeiani), ex Gaetulis duos quos arbitrabantur suis rebus amicissimos, magnis praemiis pollicitationibusque propositis, pro perfugis speculandi gratia in castra Caesaris mittunt.

pollicitationibusque MUST: pollicitationibus V | pro perfugis MU: propere fugisse STV

If STV's nonsensical propere fugisse is the archetype's reading,μ‎ has made an impressive emendation by adding a few letters and redividing pg 41the words. This is not impossible, since the two Gaetulians are referred to as perfugae later in the passage (35.5) and the topic of Pompeian deserters is frequent in all of the civil war narratives (e.g. BAfr 19.1, 32.3 to mention only the most recent). If, on the other hand, μ‎'s pro perfugis is the reading of the archtype, ν‎ fell into error by linking pro and per and made a half-hearted effort to patch things up. The (slight) possibility of emendation by μ‎ makes this example less clear-cut than the previous one, but it still seems, on balance, to be better evidence for ν‎ innovating against archetypal μ‎ than for either of the other stemmata, both of which would require S and π‎ to have made identical independent innovations here.

  1. 13. BAfr 44.1 … nauis una in qua fuerat Q. Cominius et L. Ticida eques Romanus… a Vergilio scaphis nauiculisque actuariis excepta est…

cominius UTV: comminius MS | Ticida eques Romanus] ticida eques romanus [romanus per compendia] MU: ticideq. Rō TV: ticideque S: ticida equites Romani Wölfflin, Klotz | uergilio MT: uirgilio SUV

Klotz (1927, XII) flags this as one of a handful of passages where his extra-stemmatic source (y) is particularly necessary for explaining a good reading in μ‎ where S and π‎ agree in error. However, Ticida's cognomen is preserved by the archetype at BAfr 46.3 Cominium cum Ticida, so one has to leave open the possibility that μ‎'s reading entered the tradition as a correction based on the later passage. However, it is obviously easier to see the reading of π‎ as an error transmitted from ν‎, and that of S as a rather superficial repair for ν‎'s error (omitting the apparently nonsensical but not worrying about the polysyndeton et-que). This is modest but not decisive evidence of ν‎ innovating against an archetypal reading in μ‎. One certainly doesn't need an extra-stemmatic source to explain it.

  1. 14. BAfr 45.1 Hac habita oratione Scipio, cum existimasset pro suo beneficio sine dubio ab his gratias sibi actum iri, potestatem iis dicundi fecit.

sibi actum iri MU: sibi iacturas S: sub acturis T: sibi acturis Vac: sibi acturas Vc

The basic division here is between μ‎, which has the infinitive form required by the syntax, and STV, which have a participle instead. (The small variations are irrelevant for our purposes.) If the participle was in the archetype, we must credit μ‎ with an impressive emendation, one that fully satisfies the needs of the sentence. And the same pg 42applies if the participle was in β‎. If μ‎'s reading preserves the archetype, however, the nonsensical participle was an innovation of ν‎. Independent innovations are implausible here, as, I think, is conjecture by μ‎. So this is another passage where the distribution of readings is better explained with a ν‎ vs. μ‎ stemma than with either of its rivals.

  1. 15. BAfr 58.2 Caesar item producit copias … sine dubio existimans ultro aduersarios, cum tam magnis copiis auxiliisque regis essent praediti promptiusque prosiluissent, acie secum concursuros propiusque se accessuros.

aduersarios MU: -ias STV | praediti MU: praetio STV | prom(p)tiusque US: prompti usque MTV | acie Schneider: ante ω‎ | secum] se cum USTV: se M | concursuros STcV: concursores MUTac

If STV's praetio (i.e. pretio) is the reading of the highly corrupt archetype here, μ‎ has made an impressive emendation. Praeditus is used elsewhere in the BAfr (22.4, 22.5, also 69.4), but always with abstract nouns in the ablative, so it was not an obvious choice here, even if it was obvious that some repair was necessary. It is more plausible (especially given the potentially misleading word division prompti usque) that μ‎ preserves the archetype. Since the error in S and π‎ is not such as to occur in two branches independently, this passage too counts as evidence for ν‎ innovating against archetypal μ‎.

  1. 16. BAfr 88.6 Quo (sc. Catone) interfecto L. Caesar, ut aliquid sibi ex ea re auxilii pararet, conuocato populo contione habita, cohortatus omnes ut portae aperirentur.

ut MmrS: om. MUTV | ex MUTV: haec S | auxilii pararet Mmr]86 auxilii raret MU: auxiliaret ST: auxiliaretur V | cohortatus USTV: -atur M

The nonsensical expression auxilii raret in μ‎ is likely to be the reading of the archetype here. The exemplar of S and π‎ will have produced a plausible verb and passable construction by dropping two letters and combining the rest. V, motivated by the fact the verb in question is regularly deponent, then innovated further. The appropriateness of auxiliare/-ior is in fact doubtful, since Caesar only uses its gerund and the non-Caesarian Bella don't use it at all. S and π‎ are unlikely to have arrived independently at so imperfect an innovation, as they would have to have done if ν‎ was not their exemplar. So this passage, too, counts as evidence for ν‎ innovating against archetypal μ‎.

pg 434. Discussion

It has been possible to add fourteen passages to the evidence of BC 3.105.1 for the existence of ν‎: two passages (one rather complicated) from the BAlex and twelve passages from BAfr with evidence, variously weighty, of ν‎ in error against archetypal μ‎. It has also been possible to add three passages (BAlex 19.2, BAfr 10.2, 26.5) to Hering's short list of passages showing μ‎ in error against an archetypal reading in ν‎. None of the evidence considered so far, however, can be called decisive. The shared omissions are small, the shared errors are either correctable or independently repeatable. Or double readings in the archetype may have obscured the lines of descent. So it will be worth exploring the shape of the stemma with a different approach.


To test the positive argument for a tripartite stemma I use evidence from the BC. Here, as was mentioned above, we are looking for unstable relationships defined by agreement in errors that are correctable by conjecture. This kind of argument needs substantial numbers, and I am reluctant to trust evidence I haven't verified myself. However, the picture that emerges from the BC on its own is, I think, sufficiently clear.87

Three points needs to be made at the outset. First, 'error' can be a constructed category, as can 'omission'. That is, they can be defined with reference to a particular text, which for the purposes of the analysis counts as 'correct' and 'complete'. In Brown's analysis, for example, these terms are defined with reference to Fabre's text. But if pg 44a text has been established on the basis of a stemma, the argument becomes circular if one then uses that text in the process of vetting rival stemmata. This is particularly problematic where the transmission offers two viable readings—Housman's infamous 'two bundles of hay'—and editors can differ as to which to print.88 In the lists below I use error and omission in the commonly accepted sense of the terms, for innovations that do not make sense and about which one can therefore be reasonably confident that they did not arise independently. Second, orthographical variants are not included in the category of error.89 And finally, it is important to remember that the analyses in this section only represent hypothetical or plausible transmissions. They are meant to test the stemmata under consideration. The actual (or probable) pattern of transmission will only become visible when a stemma has been established. The issue of horizontal transmission is taken up in the concluding discussion.

1. Pi vs. μ‎ and S

Mu and S agree in correctable errors and omissions against π‎ in twenty-six places.90 Occasionally the pattern is complicated by pg 45further tinkering by M or S. The corrections offered by π‎ are sometimes paralleled by corrections by a later hand in M (Mmr) or improved upon by editors. The relevant passages are these:


hortatur TV : hortat U : orat mS91


ex leui TV : et leuia [-ua Uac] MUcS


hae TV : h(a)ec MUS


armamentorum TV : armentorum MUS


hunc TV : huc MUS


uix TV : uis US : suis M


ii] hi TV : L MUS


copiam TV : -a MUS


caesar et TV : caesari MUS


uacabat TV : uagabatur MUS


conuoluta TV : (a)euoluta MUS


maiore TV : -ra MUSac : -ri MmrSc


lutoque crates Paul2: c- luto TV : c- lutoqu(a)e MUS


usum MmrTV : usu MUS


se certis TV : secretis MUS


memoria MmrTV : -am MUS


cohortatur MmrTV : -us MUS


ei MmrTV : et MUS


summam2 TV : -a MUS


erant UacTV : erat MUcS


tabellam iis <qui> scripsi: t- in MUS : t- (sc. iis uel eorum) qui MmrTV, Klotz, Fabre


pompeium MUS : -us TV92


una prima MmrTV : p- MUS


confecti MmrTV : -ta MUS


thessalia MmrTV : -iam MUS


regio ne TV : regione MUS

The opposite scenario, where π‎ is in error against the agreement of μ‎ and S, occurs more than 130 times.

pg 462. S vs. μ‎ and π‎

Mu and π‎ agree in correctable errors and omissions against S in twenty-five places.93 Occasionally the pattern is complicated by further tinkering in one manuscript or another. The corrections offered by S are sometimes paralleled by corrections by a later hand in M or m (Mmr, mmr).94 The relevant passages are these:

pg 47


L. S: SL mUTV compendiis indicatis


auximo mmrS ex 1.12.3, etc.: maximo mUTV95


<ne> ed. pr., Klotz, Fabre hoc loco, ante ad mmr: om. mUTV: et re [pro et ne?] S96


ultro S: ultra mUTV


commeatusque … si accidat S: commeatus [-tis M] qu(a)e … -ant MUTV97


diebus hoc loco S, ante superioribus Mmr: om. MUTV


fit S: sit MUTV


usi S: usu MUTV


longo S: -ge MUTV


ut S: om. MUTV


id MmrS: idem MUTV


statione MmrS: -es MUTV


posse S: posset MUTV


quibusdam S: quibus qu(a)edam MUTV


uideretur S: uteretur MUTV


uisus MmrS: usus MUTV


aliquid S: aliquod MUTVper compendia


libone MmrS ex 3.17.5 etc.: librone MUTV


treboni S ex 3.20.1: trebani MUTV

pg 48


moderat(a)e ius Mmr: moderat | eius S: moderata eius UTV et M ut uidetur98


at labienus S: ad labienum MUTV


impedimenta omnia S ex 3.77.1: o- TV: omni MU


spintherque S ex 1.15.3 etc.: spint(h)er .q. MUT: spinter et V


fufio S ex 3.55.4: fusio MUTV


pelusio MmrS ex 3.103.1: pulesio MUTV

The opposite scenario, where S is in error against the agreement of μ‎ and π‎, occurs well over seven hundred times.

3. Mu vs. S and π‎

S and π‎ agree in correctable errors and omissions against μ‎ in fifty-four places.99 Occasionally the pattern is complicated by further tinkering in one manuscript or another.100 The relevant passages are these:


com(m)itium mU : -tum STV


caesari mU : caesar STV interpunctione anteposita


sanciantur mU : -iatur STV


afranium MU : (a)erarium STV

pg 49


legionum MU : -nem ST : -nes Vac : -nis Vc


uetuit MU : metuit STV


ilerda MU : -am STV


tum MU : om. STV


iis MU : om. STV


sicorim MmrU : -rum M : -riam ST : -rim iam V


quae MU : om. STV101


correptis MU : -ti STV


se McU : om. MacSTV


posse MU : possent STV


praetoria cohorte MU : -iam -em STV


iurent MU : uenirent STV


instruit MU : -uunt STV


non committere MU : c- STV : non -ret ed. pr.: -ret Nipperdey: ne -ret Paul


suscensendum [-cendend- Uac] MUcTc : succensend- [suicens- Sac] ScTacV


horum MU : quo- STV


quieti se MU : quieti STV


nihilo minorem U : n- -rum SV : nihil ominorum T (deest M)


uellet MU : uel STV


mille] Uc : eo M : om. UacSTV


nostros MU : -ris STV


quorum MU : quarum STV


perfici(u/e)ndis MU : -fac- STV


IIII Ursinus: illi MU : ille STV


classe MU : -em STV


adhibito MU : adhibitoqu(a)e ST : a- .Q. V


iuuit MUc : iuit Uac : lu(i)it ST : om. V, spatio litterarum 5 relicto


tum MU : cum STV


funibus MU : fin- ST (deest V)

pg 50


ex MU : om. ST : in V, Klotz, Fabre


nostros MU : -ri STV


quibusdam MU : quibus STV


nona MU : non STV (u. et infra)


centurionibus MU : centum STV (u. et supra)


et tametsi MU : etiam etsi STV


V legiones … deductas Nipperdey: V -nem -ctam MU : quinta -ne … -cta STV per compendia


iuberet MU : iubet STV


continenti MU : -tia STV


aciem MU : -es STV


adiunxit MU : ad STV


aestu fatigati MU : (a)estu TV : est S102


sunt MU : om. STV


quascumque … partes MU : quamc- [quac- V] … parte STV


soluerunt MU : -rent STV


reconditis MU : reconitis S : recognitis TV


tenebatur MU : -antur STV


sui MU : suis STV103


ex MU : om. STV


praemuniit] praemunit MU : praemuniti ST : muniti V


deficeret MU : d- et STV

Altogether, ν‎ is found to be in error against μ‎ roughly twice as often as either of the pairings considered above, and more often than both of them combined. This disproportion is the more striking in that μ‎'s errors where ν‎ has a good reading number roughly thirty-five, whereas the singular errors of S and π‎ far outnumber the shared errors of the pairs opposed to them.104

pg 514. Discussion

These rather lopsided numbers do not support the tripartite stemma hypothesis. They do, however, align with the evidence presented in section C above in favour of a bipartite stemma with μ‎ and ν‎ as the main branches, and reinforce long-standing doubts about the coherence of β‎. Furthermore, the fact that S and π‎ agree in error against μ‎ more often than μ‎ is in error against the agreement of S and π‎ in a good reading constitutes an argument against contamination being the cause of the agreement between S and π‎ in good readings and bad. For while it is possible for the occasional error to be propagated via contamination, particularly if the error is a plausible innovation, the wholesale propagation of error in ν‎ cannot be so explained.

However, it remains the case that μ‎ and ν‎ are frustratingly elusive. The numbers presented above are suggestive but hardly conclusive, and different decisions about inclusions and exclusions (see nn. 93 and 99) would change the picture significantly. The absence from both branches of uncorrectable omissions remains a puzzle. (In the tradition of the BG, by contrast, the α‎ branch is defined by five substantial omissions, the β‎ branch by two. Similarly, the archetype of our tradition seems to have at least ten substantial lacunas, while π‎ has five.)105 So it is worth considering the possibility that horizontal transmission has eliminated the most distinctive errors of μ‎ and ν‎ and contributed to the impression of the waywardness of S.

The difficulty is that if the μ‎ vs. ν‎ stemma is correct, we won't be able to see horizontal transmission in action. An agreement between, say, μ‎ and π‎ that arose when ν‎ was corrected with reference to a μ‎ manuscript (or vice versa) will be indistinguishable from a good reading inherited from the archetype.106 The best evidence available (apart from the above-mentioned absence of the sort of errors that pg 52one would expect to find) comes from passages where it is plausible that π‎ has joined μ‎ in a spuriously attractive innovation, while S preserves a difficult or corrupt archetypal reading. These are a small subset of passages that earlier editors considered to be evidence of β‎. The most plausible are the following:


auximo … progressus S : maximo … progressu mUTV107


alba S : -am mUTV


contulerant SVac : -at MUTVc


terrore oblato MUTV : -or ablatus S : -or oblatus Aldus


noluerit] noloerit Sac : -int MUScTV


<cum> contumelia Nipperdey: contumelia MUTV : -am S : <per> -am Klotz


quare missis Brutus: qui remissis MUTV : cui remis sis S : cui rei missis Nipperdey: ua <de> re missis Aldus


diei MUTV : die S, Klotz, Fabre


pertinebant MUScTV : -at Sac, Klotz, Fabre


aberant MUTV : -at S, Klotz, Fabre


cursum MUTV, Klotz: om. S, Fabre


praestaret MUTV : -stare S : -starent ϛ‎


despiceret MUTV : -rent S108

The advantage of contamination over transmission from β‎ as an explanation for these passages is that contamination is generally unsystematic, yielding occasional but not consistent agreements. And that it is more likely to result in a good reading than in a bad one. So the relative paucity of conjunctive errors defining β‎ is not likewise an argument against occasional contamination of π‎ by μ‎. The fact that the process was unsystematic can perhaps best be seen in passages like the following, where π‎ agrees neither with μ‎ in a good (or acceptable) reading nor with S in a bad (or worse) one:


at mU : cum S : ut TV : ac ϛ‎ teste Dübner


et2 MU : ut T : om. SV


et MU : at S : ut TV


seseque in MU : seseque TV : sese in S

pg 53


uoluerunt … perueniunt MU : -erit … -it S : -erunt … -it TV


Cornelia naues traduxisset Meusel : corneliana uestra dux isset S : corneliana traduxisset TV : corneliana uela duxisset MU


non2 MU : nam S : nonne TV


ne1 US : nec TV : nam M


portus MU : -u S : -um TV


tum2 MU : dum TV : om. S

This brings us to a rather messy three-part conclusion: (1) Hering's bipartite μ‎ vs. ν‎ is basically correct, but (2) an unsystematic collation of ν‎ against μ‎ in a generation between ν‎ and π‎ probably resulted in the filling of ν‎'s major omissions and perhaps in the removal of some of its innovations, and (3) μ‎'s omissions may have been filled in the course of the same collation. The fact that all of our manuscripts—with the exception of M's occasionally cited Beneventan copy m and possibly Vall.—may have originated in France makes the circumstantial case for such a collation at least plausible.109 One result of this horizontal transmission is that S, with its many unfilled omissions and uncorrected errors, looks quite different from its exemplar ν‎. The other is that ν‎ and μ‎, which are only visible through their innovations, are hard to see since the most significant innovations have been removed by the horizontal transmission of archetypal readings.

We began this long discussion of the stemma with the hope that, if it could be proved that μ‎ and π‎ belonged to different branches of the archetype's descendants, we would be able to recover more of the archetype than was possible with earlier stemmata. The results stated above mean that this hope is only partly fulfilled. That is, μ‎ and π‎ do belong to separate branches, but there has also been some horizontal transmission between these branches. The extent of that transmission cannot be determined with any precision: it is somewhere on the spectrum between occasional and systematic, probably closer to the former endpoint. This makes the reconstruction of the archetype less mechanical and more a matter of editorial judgement, with horizontal transmission, along with innovation (both inadvertent and deliberate) and archetypal doublets (variants, glosses, corrections) as pg 54possible explanations for the many stemmatically peculiar distributions of readings that characterize this tradition.

The stemma to be used in constituting the text of ω‎ is therefore the following (for the constituents of π‎ see IV.B.3 below):

Fig. 2

Fig. 2


42 In his 1950 edition Klotz accepted the refinements to Holder's stemma proposed by Fabre in 1936.

43 Hering (1963) and Brown (1972) differ over the contribution of V, on which see pp. 66–77 below. In his 1987 edition of the BG Hering set V beside (instead of below) T in the stemma and added '(?)'. He only reports a reading from V if it is an innovation both good and unique (1987, IX).

44 See Hering (1963, VIII) for a table listing the various sigla used for the principal mss. Those that I use are the following. M: Florence, Biblioteca Medicea Laurenziana, Plut. lat. 68.8 (with m: Biblioteca Medicea Laurenziana, Plut. lat. 68.6 for BC 1.1–1.33); U: Vatican, Biblioteca Apostolica Vaticana, Vaticanus lat. 3324; S: Florence, Biblioteca Medicea Laurenziana, Ashburnhamensis 33; T: Paris, Bibliothèque Nationale, Lat. 5764; V: Vienna, Österreichische Nationalbibliothek, 95. Reference is occasionally made to N: Naples, Biblioteca Nazionale, IV.C.11; L: London, British Museum, Additional MS 10084; R: Florence, Biblioteca Riccardiana, 541; and Vall.: Rome, Biblioteca Vallicelliana B. 45.

45 In the following discussion lists of manuscripts such as 'MUTV' or 'STV' are a shorthand for 'M, U, T, and V' or 'S, T, and V.' To refer to hyparchetypes I use the sigla β‎, μ‎, ν‎, and π‎.

46 Of course with Hering's stemma it is impossible to reconstruct the archetype securely in some situations where reconstruction was possible with the S vs. β‎ stemma, namely, where S agrees with π‎ or T or V against μ‎. But these situations—again, given S's waywardness—are much less frequent than agreements among MUTV. For the numbers see p. 84 n. 166 below.

47 The difference to the printed text is reduced by the fact that editors have generally treated S as inferior and followed (or emended) β‎. However, my text of the BC differs from that of Fabre, for example, in more than three hundred spots for a variety of reasons, including the new stemma. For some of the passages where the stemma has made a difference to the printed text see nn. 93 and 99 below. For the complete list of differences see the Conspectus editionum.

48 Holder's introduction is extremely brief (1898, V–VIII). By way of explanation for the σ‎ vs. β‎ shape of his BC stemma he lists fourteen omissions by σ‎ where β‎'s text is whole. Cf. Timpanaro (2005, 175) on 'the tendency to identify one class of manuscripts α‎ on the basis of shared characteristics and then to call β‎ everything that in reality is merely non-α‎.'

49 The two long omissions in UT that help define the BG's β‎ family (3.9.10, ten words, and 7.77.13, fifteen words) do not similarly define the common ancestor of MUTV, since the BG's β‎ is the archetype (ω‎) of the entire tradition for the BC and other Bella (see p. 7 above).

50 To complicate matters further, S makes the transition from its α‎ exemplar to one belonging to the BG's β‎ family in two passages near the end of the work (7.58.4–62.6 and 8.23.5 to the end of book; see Hering 1963, 12–20). Despite its move to the BG β‎ family (≈ω‎ for the BC and non-Caesarian Bella) S retains some undue credibility from its α‎ associations. Bouvet, for example, presents it as one of the 'manuscrits appartenant à la classe α‎ pour la Guerre des Gaules' (1997 (1949), XLV), and du Pontet goes even further (1900, [ii], quoted at p. 8 n. 24 above).

51 Conveniently collected by Hering (1963, 61): BC 1.39.2 et fortissimo, 1.40.1 diebus, 1.41.5 opus, 1.64.6 arrepta; BAlex 57.3 legionem, 60.1 orant, 60.3 uideret, 64.2 uenit; BAfr 2.4 mandatis, 19.3 equoque, 61.5 frumentandi gratia, 83.2 plumbique itata, 86.3 cohortibus. (See p. 24 below for omissions by MUTV in the BHisp.) In no case, Hering concludes after going through this list, can one exclude the possibility that the surplus text in S is the result of a well–judged innovation or that the omission in MUTV is the result of simultaneous error by μ‎ and π‎. For two significant items missed or underestimated in his list see n. 97 below on 3.75.1 impedimenta and 1.36.3 si accidat.

52 Also collected by Hering (1963, 69): BC 1.61.6, 3.84.5, 3.93.1; BAlex 1.1, 3.3, 28.3; BAfr 15.3, 26.3, 83.2, 98.2.

53 Klotz (1950 (1926), VII; 1927, X–XI) lists more than fifty passages from the BC and non-Caesarian Bella in which the reading of μ‎ is in his view to be attributed to an extra-stemmatic source. In Brown's view, however, for the BC at least these readings are 'barely right, thanks to the saving grace of one or two letters; such a slim margin hardly demands the assistance of a lost manuscript' (1972, 34). Fabre represents y as completely independent of the ω‎ tradition, a source preserving evidence from a pre-archetypal phase of transmission (2006 (1936), XIV).

54 See e.g. Fabre (2006 (1936), LIII–LIV).

55 See e.g. Andrieu (1954, LXXVI, LXXIX).

56 Cf. the dictum of L. A. Post, cited by M. D. Reeve apropos of this issue (1989, 6): 'It is not really safe to discuss the relationship of a manuscript without a fairly comprehensive knowledge of the methods and weaknesses of the scribe who wrote it. Readings that would prove relationship in one case may mean nothing in another.'

57 For Hering on V see p. 67 below.

58 He maintains, however, that it is 'nicht schwer' (1963, 76) to find such errors, pointing to lists of passages showing unexpected associations between σ‎ and π‎ (e.g. Klotz 1927, X–XII, Andrieu 1954, LXXVI–LXXVIII, and Bouvet 1997 [1949], XLVIII–XLIX). As we saw above, of course, Klotz et al. explained these associations differently, as the result of extra-stemmatic readings, contamination, or simultaneous error. The BC is strikingly absent: Fabre gave no such list.

59 Brown also conducted a search for underappreciated manuscripts, turning up one of significance for the construction of the text where the original chapters of its parent M are missing (Rome: Biblioteca Vallicelliana B 45, henceforth Vall.; see 1972, 10 and Appendix).

60 Brown (1972, 23) in fact posits the existence of an intermediary θ‎ in the transmission of S to explain how in N, which is otherwise dependent on S, a major transposition in S has been repaired (see p. 86 below). For determining the text of the archetype, however, θ‎ and S are indistinguishable, so for convenience in this discussion I simply refer to S.

61 Cf. Winterbottom (1983, 36 n. 8): 'Hering … was right to say [sc. in his review of Brown] that the evidence of the other Bella (including the B.G.) should have been taken into account.' In my view, the evidence of the BG at least is unlikely to help establish either β‎ or ν‎ because the relationships among the manuscripts are different for the BG. As was mentioned above, S and M move from the α‎ family to the β‎ family at or near the beginning of the BC. Furthermore, in the two BG passages where S is dependent on a β‎ source the only innovations reported by Hering show S going its own way (7.60.1, 8.28.4). So although UTV stay in fixed positions relative to one another throughout the corpus, the altered positions of S and M make it seem unlikely that solid evidence will emerge from the BG tradition for either ν‎ (STV) or β‎ (MUTV). Certainly one would expect Hering to have found any evidence there was in the BG for his novel ν‎ family. See also n. 50 above.

62 This assumes that the archetype had a single reading at any given spot. Since, as will become clear below, our archetype contained double readings such as variants, glosses, and probably corrections, the dichotomy between 'innovation' and 'unquestionably archetypal' is not always relevant.

63 Hering, relying on contradictory and inaccurate reports about M at this spot, presents the evidence thus: 3.105.1 ampium M1U: appium McS: apium TV (thus Klotz; Fabre says M has ampium). According to Brown's collation and my own autopsy there is no correction here in M, which reads Appium, although it is true that the unevenly faded ink of this part of the manuscript makes it more difficult to distinguish between original text and correction here than it is elsewhere (see p. 78 below). Page images of M (Plut. lat. 68.8) are available on the website of the Biblioteca Medicea Laurenziana <>; Appium is on f. 132v, line 7.

64 Unless we want to return to the hypothesis of the extra-stemmatic source. But this is a counsel of despair.

65 Of course Appius is familiar as a praenomen, not a nomen (although it exists as such as well), and as a praenomen it would not be paired (as it is here) with Titus, but Appius is a praenomen that looks and functions a lot like a nomen in, say, its ability to generate an adjective, as it does in, e.g., via Appia (cf. aqua Marcia, lex Iulia).

66 Or a correction in the archetype. If both ampius and ap(p)ius were transmitted to the first generation of ω‎'s descendants, the lines of descent leading thence to our extant manuscripts might well be obscured by scribal choices. It would be tedious to mention this possibility everywhere it might be relevant in the coming discussion. The question of corrections in the archetype is taken up in general terms on p. 58 below.

67 Cf. Timpanaro (2005, 161 n. 6) on the implication of families defined by a single conjunctive error: you have to assume 'a subarchetype whose copyist committed only one serious error'.

68 As Hering observes (1963, 61), S seems to have been 'systematisch durchgearbeitet'. See further p. 88 below.

69 Carter (1997, 254) translates innotuit impersonally—'it became known that'—but this conflicts with the personal referent of suo signo.

70 An error in Klotz's apparatus ('eius S: ea β‎') obscured the situation before Pascucci's edition.

71 An association in a significant innovation by μ‎ and S against archetypal π‎ would also disprove both Hering and Brown. The evidence is extremely weak. For the BC, according to Brown (1972, 30–1) 'there are so few instances common to SMU as to be negligible'; see further p. 44 below. For the BAlex, Andrieu says that all instances are the result of identical independent innovations in S and μ‎ (1954, LXXIII: 19.1 propior TV: prior MUS; 50.3 legionis TV: legiones MUS; 68.2 armatura] matura TV: natura MUS). (One might quibble about his analysis of these passages, but that in itself shows that they do not offer decisive evidence.) For the BAfr, no mention. Diouron finds nothing in the BHisp (1999, XCIX).

72 Diouron dismisses from consideration five corrections by μ‎ or U (1999, CI): 4.2 se dederunt MU: sederunt STV; 5.1 transire MU: ire STV; 9.1 salso MU: salsone ST: falso V; 30.3 temere MUTc: timore S: timere TacV; 33.3 caput ex proelio Cordubam] capud ex p- co- U: apud ex p- c- ST: apud co- ex p- V; on the same grounds Pascucci (1965, 75) dismisses 28.4 id quod U: .d. quod ST: quod V. Another such is 31.4 legio MU: legi STV. Diouron also dismisses three 'fautes à faire' (1999, XCIX n. 43): 8.4 item MU: idem ST: id est V; 9.3 Caesaris MUc: -ri UacSTV; 32.7 mitterent U: -ret STV.

73 Orthographical variants: 7.3 ateguam MU: adteguam S: atteguam TV; 23.4 eius] is U: eis STV (M deest). In the latter passage the issue is debatable. I am inclined to see is and eis as ablatives (generated in error by the preceding cum); on the interchangeability of is/iis/his/eis in this tradition see my edition's Appendix orthographica. A nominative fits the syntax but makes no sense.

74 Diouron indicates in her apparatus that U makes a correction here, deleting et. This is incorrect: et is rubricated, not deleted.

75 Diouron indicates in her apparatus that S makes a correction here, adding -que. This is incorrect. The 'q' in the right margin, placed next to a line of which petimus is the last word, is not a supplement to the text but rather a 'query' mark indicating that a reader noticed the textual problem (f. 156r, 12 lines from the bottom). On the query marks in S's margins see my edition, p. li.

76 In Klotz (1927) this is numbered 24.4.

77 The fact that huic is a problem for its text seems to have been felt by S, which altered it to hunc (see the Appendix critica), an innovation less successful than that of μ‎.

78 On S's sublatuiri, if it doesn't just represent the loss of a macron, cf. Neue-Wagener 3.177, which lists this passage among its examples of a late Latin development of the future passive infinitive form. (My thanks to Ted Courtney for this reference.)

79 Stark (1964, 239) discusses the problems of fortiorem at some length in order to justify emending it to inferiorem. Fortis, in his view, pertains to material strength (as at, e.g., BC 2.2.4 testudo … …facta …… exfortissimis lignis), which is not at issue here. But it seems unlikely that similar semantic objections led a medieval scribe to substitute certiorem. On semantic innovations in our tradition see further pp. 101–4 below.

80 Stark (1964, 242) suggests instead that the archetype was partially illegible here, and that μ‎ and ν‎ read it differently.

81 Andrieu 1954 omits in by accident. It is in all the manuscripts.

82 Another possibility is that tristimonia and tristitia were both in the archetype, either as variants or with tristitia as a gloss on tristimonia, and that M hesitated between them (on double readings in the archetype see p. 58 below). This scenario too is compatible with all three stemmata, since archetypal doublets obscure the lines of transmission.

83 Bob Kaster suggests per litt. that the innovation may be inadvertent: 'if at co-became atque through an aural error, atque hortes would very quickly become atque hostes.' That the three innovations required for this scenario—atque hostesat cohortesatque hortesatque hostes—should end up with the putative original text seems less likely than the proposition that μ‎ preserves an archetypal reading.

84 Klotz misreports the reading of Mac as obripuise. Both he and Bouvet say or imply that M reads in miseriis, but in is a supralinear addition by the corrector.

85 For perfect active infinitives associated with word-division problems see, e.g., example (12) below and the apparatus notes at BC 1.71.3, 2.11.4, 2.14.1, 3.20.5. Esse and sese are frequently confounded in the process of expanding a common abbreviation for esse; see the Appendix orthographica.

86 On the corrections in M see p. 79 below.

87 Brown (1972, 29–30) conducts a similar review. Her numbers, which include orthographical variants as errors, are as follows: for MUS in error against TV, 'approximately 60', for MUTV in error against S 'approximately 75', and for STV in error against MU, 'about 75 … but nearly half are orthographical variants'. The last figure is particularly misleading (see p. 48 below), but it is difficult to see this since she lists omissions and characteristic errors rather than complete sets, her aim being to show that there are no indisputable innovations that define β‎ or ν‎. My argument in this section takes a different approach.

88 Housman (1932, xxxi). See nn. 93 and 99 for passages where editors choose differently.

89 The distinction between orthographical variant and error is another judgement call, particularly where variant spellings could represent distinct words or forms. Quod, for example, is often spelled quot, and nauis is sometimes but not always the equivalent of naues. My policy has been to treat a variant as orthographical (and therefore insignificant) if the distinct word or form that it might represent makes no sense in the context. When in doubt (as, e.g., with 2.13.1 deducunt S: diducunt MUTV) I have treated the variant as significant. With names the issue is more complicated and each set of readings has to be evaluated on its own terms.

90 Not included in the following list—given that the cue for correction seems negligible—are passages where both readings offer regular syntax and some sense, even if one is distinctly preferable: 1.64.3 magnitudini TV, Klotz: -idinis MUS, Fabre; 1.84.5 habeat MUS: -ant TV, Klotz, Fabre; 1.86.1 uicti aliquid scripsi [a- iam Mmr]: -qui u- MUS: -qui iusti TV, Klotz, Fabre; 2.5.2 eas MUS: eos TV, Klotz, Fabre; 2.8.1 si pro MUS: si ibi pro TV, Klotz, Fabre. At 3.16.4 editors report and adopt summam esse as the reading of π‎, but it is in fact the reading of MUSTV. The π‎/MUS split applies instead to summam belli later in the same sentence, as is noted above. A possible addition to the list is 2.33.1 neu Mmr: ne ubi MUS: necubi T non male: nec ibi V, where T seems to transmit an innovation in π‎.

91 The manuscript m is a copy of M that preserves the first 33 chapters of BC 1, which M now lacks (see p. 78 below).

92 I include this passage in the present list because π‎'s nominative is clearly a shortsighted repair for the archetype's omission of the subject of this sentence.

93 The authority accorded to S in the S vs. β‎ stemma induced Klotz and Fabre to accept a substantial number of S's unique readings into the text as archetypal, sometimes with further emendation. I exclude the following passages from my count of places where μ‎ and π‎ are associated in correctable error against S because in each case it can be argued that the reading in S is an unnecessary or unsuccessful innovation and that μ‎ and π‎ or components thereof are associated in a good (or almost good) and authentic reading, though editorial repair is sometimes still necessary: 1.6.2 saltem UT, Fabre: statim saltem m: statim mmr etUin marg.: uel statim saltim V: om. S, Klotz; 1.6.7 priuatim mUTV: priuati S, Klotz, Fabre; 1.25.3 <cum> extremis Morus] e- mUTV: ex ultimis S, Klotz, Fabre; 1.31.1 uacuas mUTV, Fabre: uprouincias S, Klotz; 1.34.5 omnibus ϛ‎: in o- MUTV: ex o- S, Klotz, Fabre; 1.39.2 nobilissimo MUTV: n- et fortissimo S, Fabre, Klotz; 1.41.4 castrorum contra Beroaldus: castra contra MUTV: contra S, Klotz, Fabre; 1.41.5 omne prius est perfectum MUT: o- pr- est per- opus S, Klotz, Fabre: omne opus pr- per- V; 1.54.4 perficit MUTV, Klotz: -fecit S, Fabre; 1.59.2 fugiebant MUTV, Fabre: ref- S, Klotz; 1.61.6 muniuntur MUTV: muniunt S, Klotz, Fabre; 1.63.1 traduxerunt MUTV: -erant S, Klotz, Fabre; 1.67.6 euincit MUTV, Fabre: uincit S, Klotz; 1.70.3 ante MUTV, Fabre: et a- S, Klotz; 1.76.2 postulant ω‎: -at Sc, Klotz, Fabre; 1.76.5 terrore oblato MUTV: -or ablatus S: -or oblatus Aldus, Klotz, Fabre; 1.79.2 suos MUTV, Fabre: desuper S, Klotz; 1.82.1 rei [quae munitionis fiebat] causa Faernus: rei qu(a)e -onis [-ones Tac] f- c- TV: rei qu(a)e -onis [-ones M] c- f- MU: rei -onis c- S: reliquae -onis c- [f-] Forchhammer, Klotz, Fabre; 1.85.4 hominum MUTV: -nibus S, Klotz, Fabre: -ni in Oudendorp; 2.17.4 pr(a)escribebat MUTV: pers- S, Klotz, Fabre; 2.22.3 missu MUTV: iussu S, Klotz, Fabre; 2.24.4 uoluerunt MUTV: uoluerit S, Klotz, Fabre; 2.28.2 <cum> contumelia Nipperdey, Fabre: c- MUTV: -am S: <per> -am Klotz; 2.38.1 regno MUTV, Fabre: r- et S, Klotz; 3.9.5 quare missis Brutus: qui remissis MUTV: cui remis sis S: cui rei [re Klotz, Fabre] m- Nipperdey: qua <de> re m- Aldus; 3.19.1 ut inter MUTV: i- S, Klotz, Fabre: at i- Herzog; 3.39.1 caninianus MUTV: caninius S: <M. Acilius> Caninianus Carter: Caninus Fabre; <M. Acilius> Caninus Klotz; 3.61.3 detulerant MUTV, Fabre: -runt S, Klotz; 3.68.3 coniuncta MUTV: -am S, Klotz, Fabre; 3.71.1 flegmatem2 MUTV hic et supra: felgmatem S, qui felginatem supra habet: felginatem Klotz, Fabre; 3.71.1 militum Vc: mill MUTVac compendiis indicatis: .mil.L. S: militum quinque Klotz, Fabre; 3.80.2 auxerant MUTV: -at S, Klotz, Fabre; 3.84.2 ex castris exercitum MUTV, Fabre: exer- e c- S, Klotz; 3.84.5 egum Vex 3.59.1, 3.79.6: uncum MUT: unum S, Klotz: Aecum Fabre; 3.88.5 <se> … conuerterant scripsi: c- MUTV: conuenerant S, Klotz, Fabre (with erroneous reports of the manuscript evidence); 3.92.3 quod MUTV: om. S, Klotz, Fabre; 3.93.1 cursum MUTV, Klotz: om. S, Fabre; 3.109.5 adiret MUTV: aud- S, Klotz, Fabre.

94 Not included in the following list—given that the cue for correction seems negligible—are passages where both readings offer regular syntax and some sense, even if one is distinctly preferable: *1.24.3 alba S: -am mUTV; *1.43.2 contulerant SVac: -at MUTVc; *1.85.2 noluerit] noloerit Sac: -int ω‎; 2.13.1 deducunt S: diducunt MUTV; 2.17.3 omnibus ferebat sermonibus MUTV: s- o- f- S; 2.27.3 postero S: -ra MUTV; *3.62.2 pertinebant] -at Sac; *3.62.2 aberant MUTV: -at S. See also p. 52 below, where the passages marked here with an asterisk appear in the discussion of contamination between the µ and ν‎ branches.

95 This set of variants is connected with another later in the sentence: 1.15.1 progressus S: -u mUTV. Hering (1963, 72) suggests that the archetype read maximoprogressus and that µ and π‎ independently altered progressus to progressu, while S altered maximo to Auximo and retained progressus. But see also p. 52 below, where this passage figures in the discussion of possible µ/v contamination.

96 On this passage see n. 174 below.

97 This correction and the addition of impedimenta at 3.75.1 are the most impressive of S's innovations in the BC. The reading of S at 1.36.3 is easier to accept as a correction if S's exemplar read commeatusque. For what it is worth, that is (or could be) the reading of T. (T usually abbreviates -que as q; and writes quae with -e caudata. Here, after a word ending in a suspension, -que is written in full; compare brut(us)que on f. 84va.) The correction at 3.75.1 can be derived from similiter at 3.77.1. Neither passage is discussed by Hering in his evaluation of the S vs. β‎ stemma (see nn. 51 and 52 above). He omits 3.75.1 because Klotz and Fabre state incorrectly that M joins S in reading impedimenta.

98 The line break in S suggests that the scribe did not understand the passage in the same way that Mmr did. On the distinction in S between the corrector and the scribe see p. 89 below. The reading in S could also be the result of a careless omission of the final -a.

99 The authority accorded to S in the S vs. β‎ stemma induced Klotz and Fabre to accept a number of readings shared by S and π‎ into the text as archetypal, sometimes with further emendation. I exclude the following passages from my count of places where S and π‎ are associated in correctable error against μ‎ because in each case it can be argued that the reading shared by S and π‎ is an innovation, while μ‎ has a good and authentic reading: 1.77.1 milites aduersariorum qui MU: q- m- a- STV, Klotz, Fabre; 2.29.3 liceret MU, Klotz: -re STV, Fabre; 3.17.4 et si MU: si STcV, Klotz, Fabre: sib Tac; 3.67.5 t. MU: tito STV, Klotz, Fabre; 3.106.3 existimabat MU: -mans STV, Klotz, Fabre.

100 Not included in the following list—given that the cue for correction seems negligible—are passages where both readings offer regular syntax and some sense, even if one is distinctly preferable: 1.45.1 nonam MU: nouam STV; 1.55.1 quam magnum MU, Fabre: iam m- STV: [q-] m- Nipperdey, Klotz: permagnum Paul; 1.58.1 detergere McUin marg.: deterrere ω‎; 2.38.4 imprudentes M: p- U: prudent(e/i)sque STV; 2.38.4 atque MU: ad STV; 3.36.2 tetendit MU: tendit STV: contendit ed. pr.; 3.62.3 uelit MU: uellet STacV: uellit Tc; 3.68.2 pertingere MU: -tinere STV; 3.89.1 erat MU: -ant STV; 3.89.1 attenuata MU: -uatae STV; 3.101.2 ad STV: apta ad MU: aptae ad ed. pr.; 3.101.5 serperet STV: -perent MU; 3.108.2 praefecit MU: -fic-STV.

101 According to Klotz (1950, VII), the addition of quae here by µ‎ is beyond the capacity of even a learned scribe. But it is clear from a related innovation later in this complicated sentence (1.64.1 cernebatur MTV: -antur US) that scribes were trying to make sense of it. The juxtaposition of two verbs (erant coniuncta and cerneba(n)tur) suggests a relative clause, and the addition of quae to supply its subject would have been a relatively straightforward emendation, especially for a scribe whose exemplar read cernebatur, as I believe µ‎ did. The change of number is motivated by the words immediately after cernebatur, equitatus nostri, which look like nominatives until premi arrives four words later. It is also possible that in the archetype Caesaris castris erant coniuncta was a gloss on locis; if so, it was worked into the text differently by different descendants. This does not change the stemmatic analysis.

102 This passage is discussed in section c. 1 above.

103 This passage was corrupt in the archetype, but the alteration of suis to sui is believable as an attempt at emendation.

104 Not considered above are nearly twenty passages whose transmission involves two or more innovations, where the distribution of readings can be explained in more than one way. These cannot be assigned confidently to any one of the three 'associations in error' discussed above: 1.25.1 at mU: cum S: ut TV: ac ϛ‎ teste Dübner; 1.48.5 ac ciuitates S: at c- MU: a -te TV; 1.69.1 nos M: nos nec U: nec TV: om. S: nostros Morus; 1.73.3 et2 MU: ut T: om. SV; 1.82.1 rei [quae munitionis fiebat] causa Faernus: rei qu(a)e munitionis [-ones Tac] f- c- TcV: rei m- c- S: rei qu(a)e munitionis [-ones M] c- f- MU: reliquae munitionis causa [fiebat] Forchhammer, alii alia; 1.84.1 lignorum TV: -no S: -ni MU; 2.8.2 quoquo S: quoque MU: om. TV (but I'm inclined to think that quoquo and quoque are orthographical variants here); 2.15.2 et MU: at S: ut TV; 2.21.1 seseque in MU: seseque TV: sese in S; 2.24.4 uoluerunt … perueniunt MU: -erit … -it S: -erunt … -it TV; 2.25.6 Cornelia naues traduxisset Meusel: corneliana uestra duxisset S: corneliana traduxisset TV: corneliana uela duxisset MU; 2.32.8 non2 MU: nam S: nonne TV; 2.33.3 ne1 US: nec TV: nam M; 2.39.5 omne S: omni TV: homini MU; 3.8.3 terreri TV: deterrere MU: om. S; 3.14.1 portus MU: -u S: -um TV; 3.68.1 tum2 MU: dum TV: om. S; 3.102.6 id si MmrTV: ipsi si MU: si S.

105 For the BG omissions see Hering (1987, XII). For the archetype and π‎ see p. 14 and p. 66 respectively.

106 I'm not inclined to locate contamination further down the stemma (say, T or V contaminated by μ‎, or π‎ contaminated by M or U). If the contamination occurs after π‎, it doesn't solve the problem of the absence of omissions. If M or U is the source rather than μ‎, it is hard to explains why there are instances of both MTV vs. US and UTV vs. MS (e.g. 1.39.1 caetratae UTV: -t(a)er- MS; 1.64.2 dolere UTV: dolore MS; 1.65.5 pugna MTV: -am US; 3.97.5 quo UTV: quod MS; 3.103.4 qui MmrUTV: quia MS).

107 See n. 94 above for a different explanation of the distribution of readings here.

108 The reading of M is misreported by Fabre and Klotz as despicerent. But the singular form is clear.

109 Brown (1972, 83) locates the production of M in Italy, but Munk Olsen suggests France instead (1982, 40). On the Beneventan features of Vall.'s Carolingian script see Brown (1972, 78).

logo-footer Copyright © 2018. All rights reserved.
Access is brought to you by Log out