Studies on the Text of Caesar's Bellum civile [1 ed.] 9780198724063, 0198724063

Citation preview

S T U D I E S O N T H E T E X T O F CA E S A R ’S BELLVM CIVILE

Studies on the Text of Caesar’s Bellum civile CYNTHIA DAMON

1

3

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries # Cynthia Damon 2015 The moral rights of the author have been asserted First Edition published in 2015 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2015934837 ISBN 978–0–19–872406–3 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY

Contents PART 1. P R O L E G O M E N A I. Why a new edition? II. The history of the text A. Corpus Caesarianum B. Direct tradition III. Constituting the text A. History of the stemma B. The evidence of the Bellum Hispaniense for â and í 1. The evidence for â 2. The evidence for í 3. Discussion

C. Looking for í 1. 2. 3. 4.

Bellum ciuile Bellum Alexandrinum Bellum Africum Discussion

D. The shape of the stemma 1. 2. 3. 4.

Pi vs. ì and S S vs. ì and ð Mu vs. S and ð Discussion

IV. The witnesses A. The archetype 1. 2. 3. 4. 5. 6.

Abbreviations Errors of word division Inversions Variants and corrections Glosses Book divisions

B. The hyparchetypes 1. Mu 2. Nu 3. Pi

3 10 10 14 16 16 23 23 29 32

33 33 34 36 43

43 44 46 48 51

55 55 55 56 57 58 59 60

61 61 64 66

vi

Contents a. b. c. d. e.

The relationship between T and V V’s contribution Contaminatio T’s omissions, filled T’s usable innovations, ignored

C. The extant manuscripts 1. M a. b. c. 2. U 3. S a. b. 4. T 5. V

Contents Corrections Innovations

Contents Innovations

66 67 70 72 76

77 77 78 79 82 84 85 85 88 91 92

PART 2. SCOPVLI! ON NOVEL AND UNUSUAL EXPRESSIONS IN CAESAR’ S BE L LVM C IVILE V. Chasmata salebraeque A. Omissions in transmission B. Innovations in transmission 1. 2. 3. 4. 5.

Inversions Roughly synonymous variants of the same length Roughly synonymous variants of different lengths Different constructions, both syntactically sound Omissions or substitutions that damage the syntax

99 99 100 100 101 102 103 104

VI. Nondum confecta

105

VII. Inaudita atque insolentia, or nouata atque translata?

109

VIII. Inaequata

121

PART 3. N O T E S O N T H E TE X T IX. Book 1

129

X. Book 2

175

XI. Book 3

207

Bibliography Index nominum Index locorum

311 319 322

Part 1 Prolegomena

I Why a new edition? Caesar’s Bellum ciuile needs a new edition.1 In 1963 Wolfgang Hering published his influential Die Recensio der Caesarhandschriften, covering the BC and non-Caesarian Bella as well as the BG, but the fine Teubner edition of the Bellum Gallicum that he produced in 1987 has no counterpart for the BC. The edition that will appear concurrently with the present volume, relying as it does on a fresh collation of the principal manuscripts by Virginia Brown, has been a long time in the making. Her 1972 book The Textual Transmission of Caesar’s Civil War, which is based on that collation, was not followed by the expected edition, but before her untimely death in 2009 she very generously passed her collation on to me, a debt acknowledged but hardly repaid by the dedication to her memory of the edition it made possible. Besides the new collation and the general clearing away of error that it permits, this edition presents a text based on a new stemma. The detailed argument for that stemma is presented below; here I will just say that it supports the reconstruction of the archetype in more places than was possible before.2 I looked at the evidence for Caesar’s text with more tolerance for stylistic liberties than has generally been applied by past editors, who tended to purge the commentarii of irregularities.3 I have perhaps been too willing to see development in Caesar’s style beyond the usages of the Bellum Gallicum, but I hope that my attempt to redress the balance will promote critical work on Caesar.

1

See Reeve (2000, 205). See Brown (1972, 9–10) on the accumulation of erroneous readings in the editions of Klotz and Fabre. Some of these errors surface in the discussion below. 3 As Winterbottom puts it (1983, 35 n. 1), texts of Caesar are ‘marked by remarkable indifference to what the manuscripts actually read’. 2

4

Part 1. Prolegomena

The focus of modern editorial work on the Bellum ciuile has been on rationalizing the list of witnesses to be used in constructing the text.4 The project of repairing the many problem spots of the tradition’s archetype is ongoing, while the problems that arose from the incomplete state in which Caesar left the work are probably beyond repair. Establishing a stemma has been an elusive goal. Indeed three fundamentally different accounts of the transmission have been proposed in the past century or so. The following paragraphs offer a survey of the major milestones. Karl Nipperdey’s 1847 edition made a content-based dichotomy in Caesar manuscripts the definition of the α and β families of the BG, the α family consisting of manuscripts that contain only the text of the BG, while manuscripts of the β family contain the five works of the corpus Caesarianum (BG, BC, BAlex, BAfr, BHisp).5 The oldest α manuscripts are roughly a century older than the oldest β manuscripts. The readings of the two families frequently diverge, and many of the divergences would be all but invisible were it not for the split tradition.6 This leaves the editor of the BC, which is preserved only in β manuscripts, uncomfortably conscious of the likelihood that the text is unreliable even where it is not obviously corrupt.7 In addition to this fundamental contribution to our understanding of the manuscript tradition, Nipperdey improved the quality of the text itself by using some of the manuscripts still used today for the constitution of the text of the BC, especially T (Paris, Bibliothèque Nationale, Lat. 5764), V (Vienna, Österreichische Nationalbibliothek, 95), and a descendant of U (see below).8 Nipperdey also made many successful or useful alterations to the transmitted text; I have accepted eighteen

4 For the humanist and early-modern phases see Hering (1963, 3–6) and Brown (1976, 101–32; 1972, 1–6). 5 See Nipperdey (1847, 37–49) on the principal manuscripts and their classification, with Brown (1972, 6–7) for further details. The manuscript S (Florence, Biblioteca Medicea Laurenziana, Ashburnhamensis 33), which will figure prominently in this chapter, is an exception to the α/β split (see p. 18 below). 6 Meusel (1885, 182) counts more than 1500 divergences in the BG. According to Dübner (1867, XVI), more than 500 of these stem from deliberate innovations by β. See further p. 100 (on inversions) and p. 101 (on divergences) below. 7 As Nipperdey remarked earlier, concluding that the situation warrants boldness but not temerity in emendation (1847, 49). 8 For the readings of these manuscripts Nipperdey used collations prepared by other scholars, conscious though he was of their discrepancies and errors.

I. Why a new edition?

5

of his emendations into the text of the BC and mention nearly fifty for diagnostic purposes in the apparatus.9 The edition produced in 1867 by Friedrich Dübner for the Imprimerie impériale of Napoléon III uses Nipperdey’s classification and emphasizes the defects of β, whose descendants are called ˜ for deteriores. Dübner’s imperial patronage brought him access to more than eighty manuscripts, among them U (Vatican, Biblioteca Apostolica Vaticana, Vaticanus lat. 3324) itself.10 His edition is primarily useful as a respository of variants. Heinrich Meusel was a central figure in late nineteenth-century work on the text of Caesar’s commentarii. In an 1885 article he revisits Nipperdey’s two-family classification, arguing that Nipperdey’s labels integri (for the α family) and interpolati (for the β family) should not be the basis of editorial choices between readings: the β family, in addition to supplying text with which to fill α’s numerous omissions, has some good readings where α goes astray. The assertion that β is not as bad as Nipperdey thought is of rather limited significance for the BC, of course, since β is the archetype of that tradition; there is no α. But it might help restrain emendation.11 Meusel also added R (Florence, Biblioteca Riccardiana, 541) to the witnesses for β as a (possible) sibling of U, identified a large number of codices descripti, and argued for the division of β’s descendants into two families (TV = π, RU = q).12 From his dismissive remarks about 9 Emendations by Nipperdey can be found in the text at 1.11.2, 1.14.4(x2), 1.38.1, 1.40.3, 1.87.1, 2.10.4, 2.15.1, 2.23.5, 2.28.2, 3.11.1, 3.63.6, 3.69.1, 3.70.1, 3.71.1, 3.71.3, 3.78.5, 3.82.4. 10 Nipperdey mentions U briefly as a member of the BG β family but is clearly unaware of its potential (1847, 46). The force of imperial patronage can be seen in the Avant-propos by the Director of the Imprimerie impériale, Anselme Petetin, who claims that ‘par l’intervention gracieuse de M. le Ministre des Affaires étrangères, l'Imprimerie impériale a obtenu communication de tous les manuscrits dont la collation pouvait offrir matière, soit à des restitutions, soit à l’étude des variantes’ (1867, VII). The exception was U, which had to be collated in the Vatican. Dübner himself reports that, in pursuit of a rumoured late-antique manuscript in Constantinople, ‘promptissime Imperator Augustissimus meis precibus annuit et a Sultano ut Parisios perferrentur illae membranae impetravit’ (1867, XVII n. 1); the manuscript was in fact fifteenth-century, not fifth. For a recent look at other Caesar-related projects of Napoléon III see Nicolet (2009). 11 Meusel puts it thus (1885, 184): ‘Offenbar wäre es ein bedeutender Gewinn für die Glaubwürdigkeit der Überlieferung von Cäsars bellum civile, wenn sich der Beweis erbringen ließe, daß  wenigstens nicht absichtlich gefälscht ist.’ 12 He and Felice Ramorino pursued the question of the relationship between R and U in an article published in 1891.

6

Part 1. Prolegomena

manuscripts with a mixed heritage, i.e., manuscripts that have α readings in the BG and switch to β for the BC and non-Caesarian Bella, it is clear that he did not yet appreciate the value of S, which is just such a codex mixtus (1885, 174; see p. 18 below).13 As the author of the monumental Lexicon Caesarianum Meusel had an unparalleled familiarity with Caesarian idiom. His many emendations, which can be reviewed in the Conspectus editionum provided in my OCT volume, generally offer a ‘normal’ expression, insofar as that can be determined.14 The work of Wilhelm Theodor Paul on the text of the BC is scattered across the editions of 1889 (the editio maior) and 1898 (a posthumous school edition) and Meusel’s Tabula coniecturarum; the Tabula records emendations proposed but not published by Paul, who died in 1894.15 For the 1889 edition Paul based his text on URTV, using Dübner’s collations, and generally preferred the reading of U, a good choice on the whole (see p. 85 below).16 His real contribution, however, lies in his clear-headed analysis of the sense of the text, which he often tried to improve by emendation and excision.17 His interventions are extra-ordinarily useful for defining the text’s problems, so while I have accepted only a handful of them, I report more than fifty in the apparatus.18 In Bernard Kübler’s 1894 edition the oldest witnesses were at long last taken into account: M (Florence, Biblioteca Medicea Laurenziana, Plut. lat. 68.8) and S (Florence, Biblioteca Medicea Laurenziana, Ashburnhamensis 33). ‘Oldest’, however, is no earlier than the middle 13 It was only in the decade after Meusel’s 1885 article that S began to attract scholarly attention (see Kübler 1898, VI). The manuscript was minutely described by Enrico Rostagno in his 1894 article. 14 Meusel’s emendations appear in the text at 2.25.6, 2.41.4, 3.66.6, 3.73.6, 3.83.4, 3.85.2, 3.86.2; his suggestions are noted in more than twenty other passages. 15 For details see Kübler (1899), a review of Paul’s 1898 edition, which differs from its predecessor in more than 300 places. The Tabula coniecturarum is a separately paginated section at the end of volume 2.2 of Meusel’s Lexicon Caesarianum (1887–93, 37–93 for the BC). It was also published separately in 1893 as Coniecturae Caesarianae. 16 Paul (1889, V). 17 Paul (1889, VI), esp. ‘Coniecturas . . . ubicumque aut sententia ipsa aut oratio flagitare uidebatur, consulto arripui.’ Cf. the concluding comments by Kübler (1899, 686): ‘Die ungemein große Zahl von Konjekturen zeugt von dem rastlosen Bemühen des Herausgebers; er suchte immer aufs neue den Text zu saübern und zu heilen. Unter seinen Vorschlägen sind gute, weniger gute und ganz verfehlte, aber kaum ganz wertlose.’ 18 Emendations by Paul in the text: 1.3.1, 1.33.4, 2.15.2, 2.32.4, 3.12.2, 3.15.6, 3.33.1, 3.60.5, 3.79.3, 3.79.6, 3.110.1.

I. Why a new edition?

7

of the tenth century, a full century later than the oldest witnesses for the BG. For M Kübler used collations by Meusel and Helm, for S a collation by Meusel.19 Most of the good readings of M and S were already available in the vulgate, which was based on hybrid manuscripts with readings from M and a descendant of S, so the new collations served mainly to clear away innovations.20 But in Kübler’s text we see the first manifestation of the reliance on S that would reach its fullest development in the edition of Alfred Klotz.21 Kübler also had access, in Meusel’s Tabula coniecturarum, to a vast compendium of scholarly attempts to fill the holes and correct the errors of a very imperfect archetype. In 1898 Alfred Holder published the stemma whose fundamental articulation of the witnesses for the text of the BC into two families— one principally represented by S, the other by MUTV—has been the basis of all subsequent editions of the BC and the non-Caesarian Bella to date, although it was challenged in 1963 and again in 1972 (see below).22 (At this point I should mention the irritating confusion that arises from the fact that Holder gave the name β to the common source of MUTV. Holder’s β is a descendant of the β of the BG tradition, which is roughly equivalent to the archetype of the BC tradition. Henceforth ‘β’ will generally refer to Holder’s β, unless the other is specified.) Holder also collated L (London, British Museum, Additional MS 10084) and placed it in the stemma as a sibling of S.23 The discussion of the relationships among the 19 Kübler collated T himself and took the readings of U, V, and (for BC 2–3) R from Dübner’s apparatus; for R’s text of BC 1 he used a collation by Rostagno published by Ramorino (1889, 253–83). 20 On the vulgate see Brown (1972, 48–49). 21 Unique readings of S newly accepted into the text by Kübler include: 1.31.1 uacuas mUTV : u- prouincias S, Kübler; 1.39.2 nobilissimo MUTV : n- et fortissimo S, Kübler; 1.41.5 omne prius est perfectum MUT : o- pr- est per- opus S, Kübler : omne opus pr- per- V; 1.59.2 fugiebant MUTV : ref- S, Kübler; 1.61.6 muniuntur MUTV : muniunt S, Kübler; 1.70.3 ante MUTV : et a- S, Kübler; 1.85.4 hominum MUTV : -nibus S, Kübler : -ni in Oudendorp; 2.28.2 contumelia Nipperdey : -a MUTV : -am S : -am Kübler; 3.58.2 rursum MUTV : -sus S, Kübler. 22 The fundamentals of Holder’s stemma were already present in Kübler’s (1894, XI), but Holder clarified the structure of the S-family and eliminated some codices descripti. 23 He also makes a vaguely worded claim to have collated all of the relevant manuscripts apart from U and V, which were collated for him by other scholars (1898, V), but his very full apparatus seems to be the source of at least some of the errors that spread subsequent confusion (for examples see p. 22 n. 63 and p. 61 n. 123 below).

8

Part 1. Prolegomena

manuscripts that he promises in his very brief praefatio does not seem to have materialized. He records a number of useful conjectures by Franz Bücheler, whose contribution is acknowledged in the preface; some of these reappear in my apparatus.24 In his 1926 editio maior Alfred Klotz accepts Holder’s stemma and focuses his attention on two of its implications: the authority of S and the difficulty of explaining good readings found in an unexpected combination of manuscripts or at the bottom of the stemma. He broaches here the possibility of an extra-stemmatic source for some of these readings (1926, IX), a hypothesis that he develops in more detail in his 1927 edition of the non-Caesarian Bella (1927, X–XI). He also explores the textual manifestations of the unfinished state of the BC. Ten years later Pierre Fabre added N (Naples, Biblioteca Nazionale, IV.C.11) as a new witness in the family of S for the text of the BC and showed Klotz’s extra-stemmatic source (y) contributing readings to one branch of β’s descendants.25 Thus in 1963, the date of Hering’s landmark study of the textual tradition of Caesar’s commentarii, the principal witnesses for the text of the BC and non-Caesarian Bella were, in one family, SLN, and in the other, MURTV. The story thenceforth is continued in chapter II below. As I mentioned above, the foundation of the present edition is Virginia Brown’s collation of SLNMURTV. I could not hope to better her palaeographical achievements, so I did not repeat that exercise in toto. I did, however, verify all of the readings in her collations of 24 Emendations by Bücheler in the text: 1.39.2, 2.32.13, 3.16.4, 3.42.5, 3.58.4, 3.84.1. I add a word here about René du Pontet’s OCT edition of Caesar, which appeared shortly after the major advances achieved by Meusel, Kübler, and Holder. The edition was not well received. Liebenam’s verdict, ‘Die Ausgabe von Pontet bringt keine Förderung’ (1901: I.90), is generally shared. Rice Holmes (1901, 174), for example, labels du Pontet a ‘reactionary’ for his Nipperdey-like proclamations about the authoritativeness of the α family over the claims of β advanced by Meusel and Kübler. In the BC preface we find the same desire to find an authority to follow in du Pontet’s assertion about the value of S (1900, [ii]): ‘At cum reputabimus quanta in commentariis de Bello Gallico huius codicis [=S] exstiterit cum codicibus familiae integrae [= α] similitudo, facile apparebit quanti faciendus sit’, which provoked Meusel to make explicit its false assumptions (1911, 37, see also 1901). The edition’s apparatus was deemed both inadequate and obscure (Rice Holmes 1901, 177; Meusel 1911, 39–41). 25 In adding N he was following the precedent set in editions of the BG by Bassi (1921) and Constans (1926). See further Cupaiuolo (1954, 59 n. 1).

I. Why a new edition?

9

SMUTV against photographs and, where photographs did not suffice, against the manuscripts themselves.26 I also accepted her elimination of 162 codices recentiores, which was based on an arduous investigation that Winterbottom rightly called ‘a notable advance’ (1983, 36). My own effort has been devoted primarily to the stemma, text, and critical apparatus.

26

Also, for BC 1.1–33, where M is missing, its descendants m and Vall. For further details about these manuscripts see pp. 77–95 below.

II The history of the text A. CORPVS CAESARIANVM The major milestones at the other end of the history of the text are much harder to discern. Almost every claim is disputed since the scanty evidence can be variously interpreted. The account given here represents my working assumptions rather than a full argument.27 When Caesar was assassinated in 44 bce, the BC was in an unfinished state, or so one may reasonably judge from intratextual references to an episode not present in the text that has come down to us, and from the fact that Caesar launched but did not complete the narrative of the war in Alexandria.28 In the conflict-filled months that followed the assassination members of Caesar’s inner circle and officer corps, specifically Hirtius and Balbus but presumably others as well, initiated the publication of narratives of Caesar’s campaigns 27 The bibliography on the corpus Caesarianum is extensive, and has political, historical, and literary as well as textual aspects; see Cluett (2009) for a brief overview and Gaertner–Hausburg (2013) for a comprehensive review of past work and some new ideas. On the physical history of the corpus and its textual implications see Pecere (2003). 28 The references to the defeat of Gaius Antonius, one of Caesar’s legates, come in Book 3 (3.4.2, 3.10.5, 3.67.5), but the episode itself belongs in Book 2. Further unanchored cross-references occur at 1.48.3 (ut supra demonstratum est) and 3.88.3 (docuimus). A finished work would not have such loose ends. The hypothesis of incompleteness underlies editorial acceptance of Nipperdey’s transposition of chapters 55–6 in Book 3: Caesar, it is assumed, wrote his chapters on separate pages and these two pages were reversed when the corpus Caesarianum was compiled. Also relevant—and difficult of precise interpretation—is the assertion by Asinius Pollio (paraphrased by Suetonius, Jul. 56.4) that Caesar would have revised his commentarii. See further Batstone and Damon (2006, 29–32), and, on Asinius Pollio, Morgan (2000, 55–60). And for the fullest argument in favour of the proposition that Caesar finished and published the BC see Barwick (1951, passim).

II. The history of the text

11

as a corpus. The ‘Letter to Balbus’ that introduces BG 8 represents one phase of that plan: its author Hirtius reports that at Balbus’ pressing request he has connected by a kind of metaphorical weaving (contexui) the two sets of commentarii written by Caesar, and completed (confeci) an unfinished commentarius about Caesar’s military campaigns ‘to the end of his life’ by finishing the narrative begun in BC 3 (BG 8 pr. 2): Caesaris nostri commentarios rerum gestarum Galliae, non comparantibus superioribus atque insequentibus eius scriptis, contexui nouissimumque imperfectum ab rebus gestis Alexandriae confeci usque ad exitum non quidem ciuilis dissensionis, cuius finem nullum uidemus, sed uitae Caesaris.29

It is unlikely, however, that Hirtius actually completed more than a portion of this ambitious plan, for events soon overtook the consul of 43, who died at Mutina in April of that year. Suetonius takes up the story with his report of a corpus of Caesarian texts that includes the Gallic war and the ‘Pompeian civil war’ by Caesar himself, and Alexandrian, African, and Spanish Wars by other variously identified authors (56.1): reliquit [sc. Caesar] et rerum suarum commentarios Gallici ciuilisque belli Pompeiani. nam Alexandrini Africique et Hispaniensis incertus auctor est: alii Oppium putant, alii Hirtium, qui etiam Gallici belli nouissimum imperfectumque librum suppleuerit.

This corpus bears only a general resemblance to the one that Hirtius describes, which does not demarcate by military theatre the campaigns of 48–45 and which carries the story to the end of Caesar’s life.30 But it matches the existing corpus Caesarianum in both its 29 Hering’s text is given. Vielhaber proposed deleting Galliae, Fuchs proposed adding et belli ciuilis after it, and comparantibus is variously emended. The meaning of confeci, too, is a disputed aspect of this precious but puzzling document; see Canfora (1993, 45–6) and Gaertner–Hausburg (2013, 22–30) for discussion. The letter’s authenticity has been challenged: see Canfora (1993) and (2000) for the argument that it is a late-antique composition foisted upon the extant corpus Caesarianum. Pecere by contrast views it as genuine ‘testimonianza “prenatale” del corpus Caesarianum’ (2003, 198). Gaertner–Hausburg conclude that its evidence is inconclusive on the subject of Hirtius’ contribution to the corpus (2013, esp. 30). 30 Two earlier readers of the BC, Asinius Pollio and Lucan, are occasionally cited directly (Lucan) or indirectly (Pollio, via Plutarch and Appian) in the apparatus for evidence relevant to the constitution of the text, but nothing can be said about the form of the text they read.

12

Part 1. Prolegomena

content and in the anonymity of the authors of BAlex, BAfr, and BHisp. Suetonius reproduces two sentences from Hirtius’ Letter later in his discussion of Caesar as author, so he seems to have had an exemplar in hand, but he makes a peculiar error in the bit quoted above in applying (a lightly modified version of) Hirtius’ description of BC 3, nouissimum imperfectumque (sc. commentarium), to the BG.31 It happens that our text of the BG is imperfectum, too, since the end of BG 8 has been lost together with the beginning of BC 1, but it is hard to know whether to invoke coincidence or to assume that Suetonius’ exemplar underlies the surviving manuscript tradition. In the BC the only visible sign of the corpus context is the sentence haec initia belli Alexandrini fuerunt, which was added at the end of BC 3 to ‘weave together’ that work and the BAlex, which begins bello Alexandrino conflato.32 When the corpus next comes into view, in Orosius’ fifth-century Historiae aduersus paganos, the authorship issue has taken a new turn and Suetonius himself is deemed the author of the BG (6.7.2): Hanc historiam Suetonius Tranquillus plenissime explicuit, cuius nos conpetentes portiunculas decerpsimus. (The portiunculae are taken from the BG.) Here again the presence of excerpts suggests that an exemplar of the work was at hand. Indeed it is even possible to discern that Orosius’ exemplar had readings that reach us through the BG’s β tradition (Hering 1987, V). The assertion of (some sort of ) Suetonian authorship survives in paratextual material found in the BG’s α family, as well, specifically in the opening titulus of the oldest Caesar manuscript, A (Amsterdam: Universiteitsbibliotheek, MS XV G 1), which concludes with the words Incipit Liber Suetonii.33 31

Does Suetonius mean that Caesar started BG 8 and Hirtius claimed to have finished it (for which there is no warrant in Hirtius’ letter), or, as Pecere rather implausibly suggests (2003, 217), that his (codex) copy of the BG, which Hirtius claimed to have finished, was defective at the end? I am inclined to view this as a quotation supplied by memory in an inappropriate context. Similarly Gaertner– Hausburg (2013, 30 n. 63). 32 Gaertner–Hausburg argue that contexui means ‘continued’ rather than ‘wove together’ (2013, 23 n. 31), but the textual evidence of ‘joins’ between corpus constituents (such as the concluding sentence of the BC, mentioned above, and the opening of the BHisp, 1.1 Pharnace superato, Africa recepta) supports a stronger reading of the metaphor. The aforementioned lacuna prevents us from seeing the ‘join’ most directly relevant to Hirtius’ claim, namely, that between the BG and the BC. 33 Suetonian authorship of the BC, BAfr, BAlex, and BHisp is also asserted in N (Naples, Biblioteca Nazionale, IV.C.11), a copy of S, but not in S itself. Paratextual material can obviously be transmitted independently of the text.

II. The history of the text

13

Like the identity of the author, the names of the constituent works are various. In addition to the ‘titles’ suggested by Cicero (commentarii . . . rerum suarum: Brut. 262) and Suetonius (commentarii belli Gallici, commentari belli ciuilis Pompeiani, but also Bellum Gallicum, Bellum Africanum, Bellum Hispaniense), we find Orosius’ generic historia, the Greek derivative ephemeris in singular and plural forms (ephemeris C. Caesaris: Symmachus, Ep. 4.18.3; Balbi ephemeris: Sidonius, Ep. 9.14.7; Caesaris historiae . . . quas ut ephemeridas condidit ipse sibi: Arator, ad Parthenium 39–40), and, in the titulus of A, the bizarre compound liber Gai Caesaris belli Gallici Iuliani de narratione temporum.34 This messy evidence provides us with glimpses of readerly engagement with the corpus Caesarianum, especially its first constituent, the BG, through late antiquity. This engagement comes most clearly into focus with the subscriptiones in which named individuals declare themselves to have ‘read’ each of the eight books of the BG.35 The basic formula for BG 1–7 is ivlivs celsvs constantinvs v(ir) c(larissimvs) legi. This is varied by a reference to an additional reader, one Flavius Licerius Firminus Lupicinus, in BG 2, and by relegi in a surprising β family subscriptio in U (see ed., p. xliii).36 In the subscriptio to BG 8 the addition of tantvm suggests that the ‘reading’ ceased at the end of the BG, but also that Constantinus was aware that there was more to the corpus. Beyond that it is difficult to go. Hering expresses little enthusiasm for the oft-expressed idea, which goes back to Nipperdey (1847, 37), that the ‘reading’ amounted to an editorial recensio of BG 1–8 and therefore explains the superior quality of the integri over the interpolati (1987, VI). We have already seen that these qualitative distinctions were called into question, and in any case the ‘reading’ has only the slightest of connections with the manuscript tradition of the BC. To trace the history of the text and more particularly the split between the BG’s α and β after the end of antiquity Hering and Brown use palaeographic evidence (see below). Hering dates the creation of

34 Many of these titles also turn up in manuscripts of the Caesarian texts; for details see Seel (1961, CXIV–CXXII). 35 For the BG subscriptiones see Hering (1987, XVI–XVII). 36 For the reference to Constantinus in U see ed., p. xlii. Lupicinus has been plausibly dated to the early sixth century but Constantinus is unknown; see Pecere (2003, 184–7).

14

Part 1. Prolegomena

the manuscript from which α and β were copied to the sixth century and the split itself to some time after the seventh century (1963, 95–6). That is, one copy survived from antiquity into the Middle Ages and was copied at least twice some time early in the Carolingian period. For the story thenceforth see the section entitled ‘Fortuna’ in Virginia Brown’s 1976 contribution on Caesar in the Catalogus translationum et commentariorum.

B. DIRECT TRADITION The text of Caesar’s BC survives in almost two hundred known manuscript books (Brown 1972, 42–65). Of these the oldest and most independent are used for the constitution of the text: S and M from the tenth century, T and U from the eleventh, and V from the twelfth (for details see ch. IV below). All of the extant manuscripts are ultimately derived from a single archetype, as is shown by errors grave and trivial common to them all. These accumulated as the text was copied in the course of the centuries after the publication of the corpus Caesarianum. For the Bellum ciuile alone, a work of some 33,000 words, they number in the hundreds.37 The most striking are the gaps (1.1.1 (the beginning of the work),38 1.39.2, 1.64.6, 3.8.4, 3.10.11, 3.22.2, 3.25.3, 3.38.4, 3.50.2), the insoluble problems (1.3.3, 1.5.3, 1.18.2, 1.35.4, 1.80.4, 3.11.1, 3.32.6, 3.48.1, 3.49.3, 3.49.5, 3.53.5, 3.69.4, 3.109.5), and the desperately garbled text of 2.29.3.39 But almost every page of the apparatus criticus shows one or more spots where the reading of the archetype (ω) is not the reading in the text. After more than a millenium of repairs the number of spots where the text is lost or uncertain is down to a few dozen, including

37 The word count used by Gaertner–Hausburg is 32,577, which breaks down as follows: BC 1, 10,992; BC 2, 6437; BC 3, 15,148 (see e.g. 2013, 286). 38 The end of the corpus is missing as well: the BHisp stops in mid-sentence. 39 The disturbances in the text of 2.29.3 are unlike those anywhere else in the tradition. What survives are discontinuous phrases about troops, municipalities, and the shifting relationships in a civil war context. These are all plausible topics for the fear-inducing sermones mentioned at 2.29.1, but the train of thought is not recoverable with any certainty; see Klotz (1950, IX–X) for some attempts. Physical damage to the archetype is probably responsible.

II. The history of the text

15

the passages listed above and some others where it is hard to feel confident about any of the emendations on offer.40 The date at which the archetype was produced can be established at least approximately by considering errors that arose from features of its script. The evidence of letter confusions is discussed in some detail by Brown (1972, 36–9). In addition to listing numerous confusions characteristic of copies of minuscule texts, she considers the absence of errors characteristic of copies of other early scripts (Beneventan, Visigothic, Insular) and the probable location of production. Her conclusion is ‘that ø could have been written in any of the preCaroline scripts used in France or even in Caroline miniscules’ (1972, 39). This gives us a rather late date for the archetype, probably no earlier than the eighth century ce.41

40 E.g. 1.4.3 [adulatio]; 1.5.2 menses uariarum; 1.25.9 ; 1.36.3 commeatusque . . . si accidat; 1.48.3 lacunam ante neutrum statuerim; 1.53.2 in forum; 2.25.1 [a] theatro; 2.32.13 Corfiniensem . . . an . . . an . . . en; 2.44.1 patresque familiae; 3.8.4 Caesaris complexum; 3.10.6 ipsi . . . essent; 3.10.11 urbiumque copias; 3.16.3 atque; 3.19.4 una uisurum quem; 3.21.5 uisa deque proditione oppidi appareret; 3.31.4 [prouincias] . . . uenisset; 3.32.2 cuius modo rei; 3.73.5 se notum; 3.75.2 quam suetissima; 3.84.3 electos milites ad pernicitatem armis; 3.86.5 cogitauissent; 3.108.2 conscios. 41 Brown’s date is consistent with Hering’s conclusion (1963, 95–6) that our ω (the β of the BG tradition) was copied from a surviving ancient manuscript some time after the seventh century ce.

III Constituting the text A. HISTORY OF THE STEMMA 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 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,

III. Constituting the text Holder (1898, VI)

Hering (1963, 87)

Brown (1972, 33)

ω

ω

ω

β

σ S

17

μ

π

M U’

T

L

U

R

ν S V

μ T

M

θ U

S

π T

μ V

M

U

x V

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 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.

18

Part 1. Prolegomena

(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 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 Æ on the basis of shared characteristics and then to call  everything that in reality is merely non-Æ.’ 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 Æ pour la Guerre des Gaules’ (1997 (1949), XLV), and du Pontet goes even further (1900, [ii], quoted at p. 8 n. 24 above).

III. Constituting the text

19

dozen, 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 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 prearchetypal 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.

20

Part 1. Prolegomena

contamination 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 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

III. Constituting the text

21

Before 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 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.

22

Part 1. Prolegomena

between 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 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

III. Constituting the text

23

ampium 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

B. THE EVIDENCE OF THE BELLVM HISPANIENSE FOR  AND 

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. 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 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’.

24

Part 1. Prolegomena

in 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. 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. 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). 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

III. Constituting the text

25

This 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. 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. 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. 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

26

Part 1. Prolegomena

preceding -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. 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

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

III. Constituting the text

27

9. 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 ϛ : -ret v | nuntiare v : -ri ϛ | dicere v : -ci ϛ | 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. 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

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

28

Part 1. Prolegomena

satisfying, so we have no evidence here of an innovation in β against an archetypal reading in S. 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. 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

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

III. Constituting the text

29

the 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 eo . . . medio 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 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

30

Part 1. Prolegomena

might 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. BHisp 4.4 Itaque Cn. Pompeius Vlia propre (sic) capta litteris fratris excitus cum 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 ?’) 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. 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 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.

III. Constituting the text

31

being 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. 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. 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 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.

32

Part 1. Prolegomena

S 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. 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] 1 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).

III. Constituting the text

33

C. LOOKING FOR 

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: 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. 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 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.

34

Part 1. Prolegomena

are 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.) 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 S 78 | 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 78

On S’s sublatuiri, if it doesn’t just represent the loss of a macron, cf. NeueWagener 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.)

III. Constituting the text

35

aside 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)is . . . uidebatur and STV’s his . . . uidebantur 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)is . . . uidebantur, 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, quod . . . sublatum 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 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 . . . ex fortissimis 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.

36

Part 1. Prolegomena

(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. 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 twiceperpetrated 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): 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), 81

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

III. Constituting the text

37

which 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 μ 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 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. 6. BAfr 18.1 Interim M. Petreius et Cn. Piso cum equitibus Numidis 1DC electis peditatuque eiusdem generis satis grandi . . . subsidio suis occurrunt. 1DC 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 1 MU : co STV, BHisp 31.10 mille] 1 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 1C than vice versa, although it could also be a deliberate alteration of the form 1C, which has no parallels in the Caesarian corpus. Could the error have arisen independently in S and

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.

38

Part 1. Prolegomena

π? 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. 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. 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

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

III. Constituting the text

39

admit 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.) 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. 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 ?’ 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 abripuisse . . . sese. 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 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.

40

Part 1. Prolegomena

misreadings 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 farreaching 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 9.2.145.77–84). 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. 11. BAfr 32.3 Interim Numidae Gaetuli 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. 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. 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

III. Constituting the text

41

the 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. 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 Rō 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. 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

42

Part 1. Prolegomena

applies 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. 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 v | 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. 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. 86

On the corrections in M see p. 79 below.

III. Constituting the text

43

4. 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.

D. THE SHAPE OF THE STEMMA 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

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.

44

Part 1. Prolegomena

a 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

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.

III. Constituting the text

45

further 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: 1.32.7 1.54.2 1.56.4 1.58.3 1.70.4 1.71.3 1.76.3 1.78.1 1.81.6 1.82.4 2.2.4 2.14.5 2.15.2 2.16.3 2.18.3 2.34.5 2.43.1 3.13.3 3.16.4 3.81.3 3.83.3 3.85.2 3.88.2 3.95.4 3.100.3 3.104.1

hortatur TV : hortat U : orat mS91 ex leui TV : et leuia [-ua U ac] 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 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.

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. 91

46

Part 1. Prolegomena

2. 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:

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 et Uin marg. : uel statim saltim V : om. S, Klotz; 1.6.7 priuatim mUTV : priuati S, Klotz, Fabre; 1.25.3 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 v : -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 contumelia Nipperdey, Fabre : c- MUTV : -am S : -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 re m- Aldus; 3.19.1 ut inter MUTV : i- S, Klotz, Fabre : at i- Herzog; 3.39.1 caninianus MUTV : caninius S : Caninianus Carter : Caninus Fabre; 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 V ex 3.59.1, 3.79.6 : uncum MUT : unum S, Klotz : Aecum Fabre; 3.88.5 . . . 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,

III. Constituting the text

47

1.13.4 L. S : SL mUTV compendiis indicatis 1.15.1 auximo mmrS ex 1.12.3, etc. : maximo mUTV95 1.25.9 ed. pr., Klotz, Fabre hoc loco, ante ad mmr : om. mUTV : et re [pro et ne?] S96 1.32.4 ultro S : ultra mUTV 1.36.3 commeatusque . . . si accidat S : commeatus [-tis M] qu(a)e . . . -ant MUTV97 1.40.1 diebus hoc loco S, ante superioribus Mmr : om. MUTV 1.44.2 fit S : sit MUTV 1.51.2 usi S : usu MUTV 1.59.2 longo S : -ge MUTV 1.68.2 ut S : om. MUTV 1.71.1 id MmrS : idem MUTV 1.75.3 statione MmrS : -es MUTV 2.12.4 posse S : posset MUTV 2.21.5 quibusdam S : quibus qu(a)edam MUTV 2.39.5 uideretur S : uteretur MUTV 3.7.2 uisus MmrS : usus MUTV 3.15.8 aliquid S : aliquod MUTVper compendia 3.18.3 libone MmrS ex 3.17.5 etc. : librone MUTV 3.20.2 treboni S ex 3.20.1 : trebani MUTV

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 v; 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 maximo . . . progressus 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/ν 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.

48

Part 1. Prolegomena 3.20.2 moderat(a)e ius Mmr : moderat | eius S : moderata eius UTV et M ut uidetur98 3.71.4 at labienus S : ad labienum MUTV 3.75.1 impedimenta omnia S ex 3.77.1 : o- TV : omni MU 3.83.1 spintherque S ex 1.15.3 etc. : spint(h)er .q. MUT : spinter et V 3.106.1 fufio S ex 3.55.4 : fusio MUTV 3.108.2 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 fiftyfour places.99 Occasionally the pattern is complicated by further tinkering in one manuscript or another.100 The relevant passages are these: 1.3.3 1.4.4 1.9.6 1.38.4

com(m)itium mU : -tum STV caesari mU : caesar STV interpunctione anteposita sanciantur mU : -iatur STV afranium MU : (a)erarium STV

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 v; 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.

III. Constituting the text 1.40.6 1.41.4 1.45.2 1.50.2 1.52.4 1.63.1 1.64.1 1.66.1 1.69.2 1.69.2 1.75.2 1.76.2 1.82.2 1.83.3 1.84.3 1.85.7 2.14.1 2.17.2 2.29.3 2.35.5 2.41.7 2.44.2 3.2.1 3.7.2 3.17.3 3.18.3 3.29.1 3.36.5 3.40.1

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] 1 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)

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.

50

Part 1. Prolegomena 3.59.1 3.63.8 3.66.4 3.67.3 3.67.3 3.67.4 3.69.1

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 3.84.3 iuberet MU : iubet STV 3.87.2 continenti MU : -tia STV 3.88.4 aciem MU : -es STV 3.89.1 adiunxit MU : ad STV 3.95.2 aestu fatigati MU : (a)estu TV : est S102 3.101.4 sunt MU : om. STV 3.102.1 quascumque . . . partes MU : quamc- [quac- V] . . . parte STV 3.102.7 soluerunt MU : -rent STV 3.105.5 reconditis MU : reconitis S : recognitis TV 3.107.1 tenebatur MU : -antur STV 3.108.2 sui MU : suis STV103 3.108.2 ex MU : om. STV 3.112.7 praemuniit] praemunit MU : praemuniti ST : muniti V 3.112.12 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

102

This passage is discussed in section c. 1 above. 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 . . . 103

III. Constituting the text

51

4. 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 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

52

Part 1. Prolegomena

one 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: 1.15.1 1.24.3 1.43.2 1.76.5 1.85.2 2.28.2 3.9.5 3.13.2 3.62.2 3.62.2 3.93.1 3.103.4 3.103.4

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 contumelia Nipperdey : contumelia MUTV : -am S : -am Klotz quare missis Brutus : qui remissis MUTV : cui remis sis S : cui rei missis Nipperdey : qua 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: 1.25.1 1.73.3 2.15.2 2.21.1

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

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.

III. Constituting the text

53

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 3.14.1 portus MU : -u S : -um TV 3.68.1 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

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).

54

Part 1. Prolegomena

possible 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): ω ν

μ M

π

U T

Fig. 2

S V

IV The witnesses A. THE ARCHETYPE With the stemma in hand we can reconstruct the text of the archetype of the BC (ω) in most of the more than 1,000 places where the manuscripts disagree. The archetype’s approximate date (eighth century) and script (pre-Caroline or Caroline minuscule) have already been mentioned. Other features relevant to the analysis of the transmission are the following:

1. Abbreviations It is clear that the archetype contained both abbreviations and errors arising from abbreviations. Those that we can detect are all rather simple ones, but they are the residue of several stages of transmission. At each stage outmoded or ambiguous abbreviations were a source of error for scribes trained in different abbreviation conventions. Among the abbreviations most prolific of error in the text of the BC are numerals (cardinal and ordinal), praenomina, prepositions and prefixes, measurement units, and the terminology of ancient political and military life (e.g. populus Romanus, res publica, consul, tribunus, eques Romanus, praefectus equitum, legatus, legio, senatus consultum).110 Others include abbreviations for common words both 110 Selected examples where the error is in ω (although the error is occasionally corrected by one or another manuscript): 1.5.4 ante diem Scaliger per compendium : ad v; 1.6.7 et Manutius : ex v; 1.9.2 populi Romani beneficium Aldus : pro beneficio v; 1.14.4 circa Nipperdey : contra USTVc : intra m : e Vac; 1.14.4 sese ed. pr. : esse v; 1.20.1 tribunos Beroaldus : -num v; 1.23.4 ab IIIIuiris Mommsen : ab (h/i)is uiris mUTV : habiis u- S; 1.24.4 N. Perizonius : cn. v; 1.32.4 ultro S : ultra mUTV; 1.32.6 legionibus

56

Part 1. Prolegomena

uninflected (e.g. sed, quoniam, uero, et, -que, non) and inflected (e.g. esse, se, noster, omnis), as well as internal syllables (e.g. -par-, -ber-, der-, -ter-) and various inflectional endings.111 The errors that arise from the archetype’s abbreviations are sometimes present in all of ω’s descendants, and sometimes just in μ or ν. But since the form in which abbreviations are transmitted is, like orthography, to some extent a matter of scribal choice, abbreviations, like the various forms of archetypal doublets, can obscure the lines of descent.112

2. Errors of word division In ω’s descendants there are a number of errors of word division.113 These may have arisen in the copying of ω’s scriptura continua, as

mmr : legibus v; 1.35.4 his] sibi Markland; 1.40.3 egressae Jurin : congressae v; 1.41.2 VI cohortes] sex -tes ed. pr. : ex -tis UTV : et -tis M : ex -tibus S; 1.61.3 [L.] ϛ : l. v; 1.61.5 miliaque Stephanus : milia V MUST : mille V V; 1.63.2 pons M : post USTV; 1.78.1 parandum ed. pr. : prandium v; 2.42.3 praefectus equitum ed. pr. : pr(a)eque MUScT : praef Sac : populus romanus aeque V; 3.77.3 quarta v : -to Morus. Regularly abbreviated words or forms with a split transmission: 1.11.1 reuerteretur mmrUcSTV : reuertetur mUac; 1.40.6 legionum MU : -nem ST : -nes Vac : -nis Vc; 1.61.1 pedum STV : pedes MU; 1.67.6 postridie MUV : post tridie ST; 2.2.4 conuoluta TV : euoluta MUS; 2.32.2 grauissime US : gauisissime MTV; 2.32.12 incolumem SV : -umen M et UT per compendium; 2.35.5 mille] 1 Uc : eo M : om. UacSTV; 3.7.2 IIII Ursinus : illi MU : ille STV; 3.31.3 post se MUV : posse ST; 3.67.3 nona MU : non STV; 3.67.3 centurionibus MU : centum STV. 111 Selected examples where the error is in ω: 1.7.4 omnia Markland : dona v; 1.21.5 uero Faernus : eo v; 1.23.3 a parte Vascosanus : aperte v; 1.30.4 imparatissimis mmr : -per- v; 1.44.4 non tenuit Aldus : continuit v; 1.48.7 utribus Aldus : utrisque v; 1.84.3 uoluerint Aldus : -runt v uel per compendia; 1.85.9 sed Aldus : sit MUS : fit TV. Regularly abbreviated words or forms with a split transmission: 1.7.1 deprauatum UV : -priu- mST; 1.32.3 esset SV per compendia et Tc ut uidetur : esset et U : et Tac (deest m); 1.73.3 et2 MU : ut T : om. SV; 1.75.2 praetoria cohorte MU : -iam -em STV; 2.41.7 nostros MU : -ris STV; 3.17.3 classe MU : -em STV; 3.63.8 nostros MU : -ri STV; 3.84.3 iuberet MU : iubet STV; 3.102.7 soluerunt MU : -rent STV; 3.112.12 deficeret MU : d- et STV. 112 Regularly abbreviated words or forms with a peculiar distribution of readings: 1.24.3 praetor1 UT : praeter mSV; 2.10.1 sunt confisi UST : confisi M : statim confisi V ; 3.4.3 equitum V et UT per compendium : equites MS. 113 Word-division errors in both branches: 1.40.3 ui uentorum Pantagathus : iumentorum v; 1.43.3 antesignanos M : ante signa nos USTV; 1.44.4 antesignanis MTV : ante signa nis U : ante signanis S; 1.53.1 multa rumores Oehler : multarum M : multa rumor MmrU : multorum ora S : multarum rumore orat T : multa rumore V; 1.61.5 conquiri et Otogesam Vascosanus : -rere toto gesma v; 2.16.2 qua sit V : quasit UST : quasi M; 2.16.3 data se uirtute U : datase u- MST : data seruitute V; 2.20.1 ac tam

IV. The witnesses

57

Brown suggests (1972, 38), but they might also be errors in ω that were the legacy of such a script from an earlier phase in the transmission. The fact that most of them are present in both branches suggests the latter, in my view; none is exclusive to μ or ν.

3. Inversions Given the frequency of word order inversions in ω’s text of the BG, where it can be checked against the α family, one has to suspect the presence of this sort of error in its text of the BC, too. In BG 7, for example, a text rather shorter than BC 3 (about 11,500 words vs. 15,000), Hering reports more than ninety inversions in the β family to which our ø belongs (in the α family, by contrast, there are eight); in the text of BC 3 we can identify with some confidence only one.114 Without an α at our disposal we are unlikely to be able to detect any but the most egregious inversion errors; others are merely suspected, and there are probably more that are now invisible.115 I only note proposed inversions in the apparatus if they serve to clarify an otherwise difficult passage. Many others that serve to smooth or regularize an otherwise tolerable text can be found in Meusel’s Tabula coniecturarum.

ed. pr. : actam MUTV : atiam Sac : autem Sc; 2.20.7 cui iusserit MmrUS : cuius|serit M : cuius erit T : cui erit Vac : cui praeerat Vc; 2.25.1 [a] theatro ed. pr. : atheatro MUST : a t- V; 2.25.6 Cornelia naues traduxisset Meusel : corneliana uestra duxisset S : corneliana traduxisset TV: corneliana uela duxisset MU; 3.19.4 una uisurum quem Elberling : una uis utrumque v : eundem uisurum quem Nipperdey; 3.20.2 moderat(a)e ius Mmr: moderat|eius S : moderata eius UTV et M ut uidetur; 3.28.4 salo naus(e/i)aque MV : salona usiaque UT : salonausiaque S; 3.68.3 prorutis Ciacconius feliciter : proutis S : prout (h)is MUTV; 3.74.5 antesignanos ed. pr. : antesigna nos M : ante signa nos USTV; 3.112.8 In eo tractu Hoffmann : neotractu [-trat- S] T et S ut uidetur : nec tractu MUV. Word-division errors in individual manuscripts (particularly M and S) can be found in the Appendix critica. 114 Inversions so obtrusive as to warrant emendation: 1.48.5, 1.65.4, 1.86.1, 2.14.3, 2.15.2, 2.28.1, 3.21.4. 115 Suspected inversions: 1.1.2, 1.4.3, 1.19.4, 1.39.2, 1.44.2, 1.61.3, 2.10.4, 2.24.4, 2.28.1, 2.30.2, 2.44.1, 3.6.3, 3.8.3, 3.10.11, 3.13.2, 3.19.4, 3.30.4, 3.32.6, 3.39.1, 3.40.2, 3.57.2, 3.61.2, 3.65.3 (2x), 3.72.3, 3.76.4, 3.84.3 (2x), 3.85.2, 3.88.4, 3.96.3, 3.101.4, 3.103.1. On inversions involving substantial displacements see the note on 1.4.3 [adulatio], p. 131 n. 5.

58

Part 1. Prolegomena

4. Variants and corrections The archetype contained variant readings.116 In some passages both variants are simultaneously present in one or another of its descendants (usually in the μ branch).117 Sometimes the variants are synonyms, more often not. Even when no one manuscript contains two variants, a hypothesis of archetypal variants is often the simplest explanation for an otherwise puzzling distribution of readings.118 The presence of variants shows a reader at work on the text and makes it plausible that the archetype also contained corrections, though these are harder to spot in any one manuscript.119 The most 116 For variants in the β family of the BG and in the other Bella see Hering (1963, 42–3) and Klotz (1927, XII) respectively. 117 Both variants present in one or more witnesses : 1.6.2 saltem UT : statim saltem m : statim mmret Uin marg. : uel statim saltim V : om. S; 1.57.4 eorum Uin marg. : eodem USTV : eorum eodem M : eorum dem Mmr; 2.30.2–3 pati. Erant] pati proiecerant M et Uin marg. ut uidetur : proiecerant US : perpeti erant TV; 2.34.6 timore U : timore tempore M : tempore STV; 2.40.1 quos US : quod T : quae V : quos quod M; 2.41.4 hi U et spatio 11 litt. relicto Vac : hi ducenti Vc : L hi T : L MS; 3.4.2 supplementi UV : comsupplementi M : plementi ST; 3.9.4 uno tempore USV : t- M : omni u- t- T; 3.10.5 africani STV et Usupra lineam : -raniani MU; 3.11.3 autem etiam [e- a- Mac] v : autem ϛ : [a-] e- Dinter; 3.17.6 orationem MUcST : rat- UacV; 3.27.2 remisit MUSTc : redit r- Tac : dimisit V; 3.28.4 firmitudine USTV et Msupra lineam : fortitudine M; 3.50.1 aggressi MUin marg. : aduersi UST : om. V; 3.73.3 felicitate USTV : facili- MUin marg. et s.l.; 3.74.1 contione MUcST : orat- UacV; 3.92.2 ut1 STV : neut M : ne U : [ut] Oudendorp; 3.101.2 constratae STV : rost c- MU; 3.102.6 arcem captam Oudendorp : ar- aram -tam M et Uin marg. : aram -tam UTV : aram -ta S : arma -ta Forchhammer teste Meusel. 118 Distributed variants: 2.21.3 populis MU : publicis STV : populis publicis Meusel; 3.9.6 dispositis MSV : dep- UT; 3.24.4 obsessionemque [-quem S] UST : obsidionemque MV; 3.40.1 eam STV : etiam MU : eam etiam Oudendorp. 119 Selected passages where the hypothesis of a correction in the archetype provides the simplest explanation of the distribution of readings: 1.7.6 ne mT : nec USV; 1.13.5 auximatibus mS : auximabus UTV; 1.21.3 atque mT : aeque USV; 1.32.9 se mS : si UTV; 1.35.1 massilia MSV : -liam UT; 1.78.4 carperet UST : caperet MV; 1.79.3 erat in USacTV : erat MSc; 2.1.2 qua MS : qu(a)e UTV; 2.9.4 tabulationem MmrUS : tribMTV; 2.11.4 infulis se USc : infulsisse MSacTV; 2.15.1 aequa MmrUS : aeque MTV; 2.17.3 se [om. Vac] in ullam MTVc : se in nullam US; 2.21.1 seseque in MU : seseque TV : sese in S; 2.23.1 quas a caesare acceperat MS : q- a- a c- U : a- q- a c- TV; 2.25.3 numidae MS : numidi(a)e UTV; 2.27.1 curionis MV : centur- UT : centum rim S per compendia; 2.28.2 primam . . . memoriam M : -a . . . -iam UT : -a . . . -ia SV : primi . . . -iam Scaliger; 2.30.3 tutius MmrUTV : tot- MS; 2.31.3 non et UT : et non V : non MS; 2.41.1 exercitu MT : -um USV; 3.9.6 dispositis MSV : dep- UT; 3.13.5 sub pellibus McUTV : suppellibus MacS; 3.19.2 tuto Vossius : duo UT : duos MSV; 3.19.6 uatinio MmrUcSTc : uati(c/t)inio MV : uatio UacTac; 3.21.1 efficeret MV : -re UST; 3.51.2 reliqui se uerterunt USV : -quis euer- T : -quos euer- M; 3.53.1 duo TacV et Sc ut uidetur : duorum MUSacTc; 3.54.2 obstructis V : obstruxit ST : extructis MU; 3.60.5 facinus difficilius Clarke : facinus MU : facilius T : difficilius SVc (Vac non legitur);

IV. The witnesses

59

suggestive passage is the following, which is difficult to explain if the archetype had a single reading: 1.40.3 propiore legiones Nipperdey : proprio religiones [-nis Mac] McTac : proprio legiones MmrUSTcV

Here we have two unacceptable readings, each present in manuscripts belonging to different branches, and a contemporary correction, Tc, that crosses from one unacceptable reading to the other one, which is superficially plausible. The distribution of readings is much easier to explain if Tac and Tc represent what was in the (corrupt) archetype, and if that double reading was transmitted to μ, ν, and π. Both types of archetypal doublet—variants and corrections—may give us a glimpse of a phase in the history of the text when copies other than the archetype were still in circulation; the other copies have no known descendants. Alternatively, these double readings may represent conjectures in the archetype.

5. Glosses The text of the archetype incorporated expressions that were originally glosses; sometimes a gloss was adapted to fit the insertion context. These foreign bodies frequently gave rise to further innovations in the text of ω or its descendants. Most provide supplementary information, often from nearby in the text, but a few summarize the content or comment on it.120 More glosses have been suspected than those I list here and excise from the text. If excision of a gloss constitutes a 3.71.4 solerentne USV : -ret ne MT; 3.72.4 recordabantur MSV : -nt UT; 3.73.5 aliqui UT : -quis MSV; 3.76.3 alii quod STc : aliq- UTacV : alii quid M; 3.81.3 erant UacTV : erat MUcS; 3.85.4 sicut MS : si ut UTV; 3.97.5 quo UTV : quod MS; 3.102.6 dimissos v : dimissis UacTac; 3.103.4 qui ab MmrUTV : quia ab MS. 120 Glosses excised from text: 1.1.1 [a Fabio], 1.3.3 [plebis], 1.4.3 [adulatio], 1.7.2 [quae superioribus annis armis esset restituta]; 1.11.1 [legionibus], 1.21.6 [tanta], 1.39.2 [ad] VI [milia], 1.53.2 [ad Afranium], 1.82.1 [quae munitionis fiebat]; 2.4.4 [latitatis], [uehementiusque exterreamur], 2.5.3 [publicis custodiisque], 2.10.4 super [musculos]struantur; 2.14.3 [muro], 2.30.2 [consiliis]; 3.2.2 [inopia nauium], 3.2.3 [copiae], [Galli], 3.11.1 [Corcyrae], 3.11.1 [antequam de mandatis agi inciperet], 3.13.5 [castellis uigiliisque], 3.31.4 [prouincia], 3.38.4 [hostium], 3.41.1 [Macedoniam], 3.48.1 [Id ad similitudinem panis efficiebant.], 3.63.3 [munitiones], 3.63.6 [exercitus aduentus extitit], 3.71.1 [L], 3.89.5 [totique exercitui], 3.91.4 [eiusdem centuriae], 3.101.4 [pari atque eadem ratione egerunt], 3.101.4 [circiter XL], 3.112.12 [nutricius pueri et procurator regni, in parte Caesaris], 3.112.12 [Haec initia belli

60

Part 1. Prolegomena

viable alternative to the text I have printed, I mention it in the apparatus.121 Many many more can be found in Meusel’s Tabula coniecturarum; most are simply bits of text that seem superfluous and are therefore felt to be alien to Caesar’s spare prose style.

6. Book divisions In the archetype the beginnings and endings of the books of the Caesarian corpus were marked in some fashion. Precisely how the transition from one book to another was marked is not clear, however, since surviving manuscripts show different models, and sometimes more than one model in a single manuscript. At a minimum there is blank space between books, but there is very little consistency beyond that. In the paratextual material that sometimes fills the space between books we are often given conflicting information about author, title, book number, etc.; it is unlikely that this material derives from the same process of transmission as the text itself. More will be said about the information provided at these spots when discussing the individual manuscripts below, but for our present purposes the important point is that the location of the book divisions is unvarying, with one easily explained exception. The exception is in S, whose text of the BC is badly out of order. The major dislocation—a large section of BC 1–2 is found in the middle of BC 3 (for details see below)—affects book division, in that in S the first book division after the beginning of BC 1 (on f. 74r) is the division between BC 2 and 3 (on ff. 95r–v); the division between BC 1 and 2 comes later (on f. 105r) and is in fact invisible since S leaves no space between these books; a later hand supplied a vertical stroke to divide them.122 The immediately preceding and subsequent book Alexandrini fuerunt.]. Cf. also the obvious gloss preserved in m alone at 1.28.2 after iubet: caesar castrum subit. pompeius ultra mare fugit. The corrector of m excises it. 121 Excision of gloss as a viable repair: 1.1.1, 1.1.2, 1.2.3, 1.3.3, 1.5.1, 1.5.3 (2 exx.), 1.6.7, 1.11.2, 1.15.6, 1.21.6, 1.24.3 (2x), 1.24.5, 1.26.3, 1.39.1, 1.39.2, 1.59.2, 1.64.1, 1.64.7, 1.80.4, 1.82.4; 2.1.2, 2.6.3, 2.8.3, 2.10.4, 2.14.3, 2.16.3, 2.21.5, 2.23.2, 2.23.4, 2.30.2, 2.34.3, 2.36.1, 2.41.2, 2.44.3; 3.1.4, 3.9.2, 3.9.6, 3.10.6, 3.10.11, 3.11.1 (3x), 3.12.2, 3.13.2, 3.14.1, 3.24.2–3, 3.25.2, 3.27.2 (2x), 3.30.1, 3.32.2, 3.32.3, 3.32.6, 3.34.2, 3.36.3, 3.36.6, 3.37.1, 3.38.4 (several), 3.44.4 (several), 3.46.5, 3.51.5, 3.61.2, 3.63.6, 3.63.8, 3.65.3, 3.68.3, 3.69.4 (several), 3.78.3, 3.79.3, 3.79.5, 3.82.3, 3.83.2, 3.84.3 (2x), 3.92.2; 3.93.7, 3.105.6, 3.106.5, 3.108.2, 3.108.3 (2x). 122 S also omits the first eight words of Book 2. It may have been intended that the rubricator supply them, as he does the first seven words of BG 7 (f. 45v) and the first twelve words of BAlex (f. 121r).

IV. The witnesses

61

divisions in S are marked by a space between books, by starting the new book on a new page, and by an enlarged initial letter (the transition from BG 8 to BC 1 on ff. 73v–74r, that from BC 2 to BC 3 on ff. 95r–v); there is no original paratextual material. But a later hand supplied some: at the beginning of BC 1, ‘De civili bello. Incipit liber nonus’, and at the beginning of BC 3, ‘Incipit liber decimus de bello civili’. This last item has caused confusion, since it suggests that what we traditionally call BC 1–2 constituted a single book, the ninth from the beginning of the BG, and that our so-called BC 3 was the second of Caesar’s civil war books, the tenth from the beginning of the BG.123 There is a certain logic to this, since the events of BC 1–2 fall in the same year, 49 bce, and Caesar elsewhere aligns year and book boundaries. But even Hirtius, who says that this is Caesar’s policy (8.48.10), abandons it when the narrative requires (ibid.). There is no justification for following S in this matter. For further details see p. 86 below.

B. THE HYPARCHETYPES The difficulty of discerning the hyparchetypes μ and ν was discussed at length above. The absence of significant omissions is a particular puzzle. However, the existence of these hyparchetypes best explains the text that reaches us in the extant manuscripts.

1. Mu Numerous innovations common to M and U are consistent with descent from a hyparchetype, μ.124 In the BC the most telling are these, where the reading in ν is likely to be archetypal: 1.64.1 nonnumquam STV : nunc quam MU 2.1.2 portu STV : portui U : oportuit M

123

Klotz (1926, VI–VIII; 1950, VI) is the most influential proponent of this idea. But his report of the manuscript evidence is inaccurate (he doesn’t differentiate between original and later hands in S, for example) and seems to be tralatitious. See further p. 86 below. 124 Earlier stemmata have a further bifurcation, with a sibling of M (U’, or q) as the parent of U and R. Hering (1963, 52–4) showed R to be a descendant of U, an eliminatio in which Brown concurred (1972, 25–6).

62

Part 1. Prolegomena 2.15.3 tecto STV : texto MU 2.25.6 Cornelia naues traduxisset Meusel : corneliana uestra duxisset S : corneliana traduxisset TV: corneliana uela duxisset MU 2.31.4 noster STV : nostris MU 3.10.5 africani STV et Usupra lineam : -raniani MU 3.28.2 summissis SV : summisso T : summis MU 3.40.1 eam STV : etiam MU: eam etiam Oudendorp 3.66.2 opere STV : opera MU 3.83.2 acutius STV : actius MU 3.91.4 uoluntarii STV : -riae MU 3.94.5 aliisque STV : a- quam maxime MU 3.101.2 ad STV : apta ad MU 3.101.5 serperet STV : -perent MU 3.106.1 a Q.] a q. T : a quae S per compendium : quae V : atque MU

The distribution of these innovations suggests that the scribe of μ got increasingly careless as he advanced through the text. But the purposeful nature of some of μ’s innovations can be illustrated from 3.75.1: Itaque nulla interposita mora sauciorum modo et aegrorum habita ratione omnia silentio prima nocte ex castris Apolloniam praemisit (sc. Caesar). Haec conquiescere ante iter confectum uetuit. impedimenta omnia S ex 3.77.1 : o- TV : omni MU | haec STV : ac MU | confectum UST : fe- M : effe- V

The archetype’s omission of impedimenta left omnia nonsensical. S restored the baggage from 3.77.1, where the subsequent departure is said to have been managed similiter. But μ changes omnia to omni to go with one of the adjacent words, ratione or silentio. And this ‘repair’ entails the further change of the antecedent-less haec to ac. Significant innovations unique to M or U show that neither is descended from the other. Innovations of the sort that could not be corrected by conjecture, for example, include the following: 1.32.3 1.65.5 2.2.2 2.5.5 2.9.2

ferri . . . probasset cur om. m (uerba 6) itineris om. M atque hi maximis om. M si superauissent . . . confiderent om. M (uerba 12) duo om. M

IV. The witnesses

63

-cessitudinem . . . ignorare om. M (uerba c.11) praesidiaque . . . . reddebat om. M (uerba 8) -gatum . . . explora- om. M (uerba c.11) his rebus om. M tirones enim multitudine nauium perterriti] nam cum perterriti M 3.40.1 semper om. M 3.41.4 posse om. M 3.46.5 et . . . coegerunt om. M (uerba 4) 3.54.1 castrorum om. M 3.63.1 et . . . X om. M (uerba 8) 3.76.3 alii . . . progrediebantur om. M (uerba 6) 3.87.7 et1 . . . uictoriam om. M (uerba 9) 3.92.3 Caesaris om. M 3.94.4 factae . . . isdem om. M (uerba 7) 3.101.3 per dispositos om. M 3.108.4 ex duabus om. M 3.112.12 [haec . . . fuerunt] Forchhammer : om. M 1.23.2 P. om. U 1.36.5 armatisque diebus om. U 1.47.2 et initio . . . tenuissent om. U (uerba 5) 2.33.1 qua . . . uiderentur om. U (uerba 16) 2.33.3 sollicitandi milites siue om. U 3.88.3 cum cohortibus om. U125

2.17.2 2.18.5 2.24.2 3.2.1 3.28.4

The relative lengths of these two lists suggests that U is a more faithful copy of μ than M is; this is borne out by other sorts of singular innovations, which are far more numerous in M than in U, as can be seen in the Appendix critica. Where μ and ν diverge, as they do well over a hundred times in the BC, μ more often offers a good reading, either by conjecture or by preserving the archetype.

125 Textual variation between U or M and ν in some of the passages omitted by M or U make it unlikely that these omissions were filled by contamination from the ν branch (see the apparatus for 1.32.3, 2.17.2, 3.76.3; 2.33.1). It is therefore likely that in all of the passages listed above μ’s text was complete and transmitted to one of its two descendants.

64

Part 1. Prolegomena

2. Nu Numerous innovations common to S and π are consistent with descent from a hyparchetype, ν. In the BC the most telling are these, where the reading in μ is likely to be archetypal: 1.4.4 1.41.4 1.45.1 1.45.2 1.52.4 1.55.1 1.69.2 1.76.2 1.77.1 1.83.3 1.85.7 2.14.1 2.29.3 2.38.4 2.41.7 2.44.2 3.7.2 3.17.3 3.17.4 3.36.2 3.36.5 3.59.1 3.63.8 3.67.3 3.67.3 3.67.4 3.68.2 3.87.2 3.88.4 3.89.1 3.89.1 3.102.1 3.106.3 3.112.12

caesari mU : caesar STV interpunctione anteposita uetuit MU : metuit STV nonam MU : nouam STV ilerda MU : -am STV iis om. STV quam MU : iam STV posse MU : possent STV iurent MU : uenirent STV milites aduersariorum qui MU : q- m- a- STV non om. STV horum MU : quo- STV se1 om. STV uellet MU : uel STV atque MU : ad STV nostros MU : -ris STV quorum MU : quarum STV IIII Ursinus : illi MU : ille STV classe MU : -em STV et om. STV tetendit MU : tendit STV : contendit ed. pr. tum MU : cum STV ex MU : om. ST : in V nostros MU : -ri STV nona MU : non STV centurionibus MU : centum STV et tametsi MU : etiam etsi STV pertingere MU : -tinere STV continenti MU : -entia STV aciem MU : -es STV erat MU : -ant STV attenuata MU : -uatae STV quascumque . . . partes MU : quamc- [quac- V] . . . parte STV existimabat MU : -mans STV deficeret MU : d- et STV

IV. The witnesses

65

Significant innovations unique to S or π show that neither is descended from the other. Omissions of the sort that could not be corrected by conjecture, for example, include the following: 1.6.1 1.6.3 1.40.5 1.74.4 2.1.1 2.6.3 2.18.6 2.24.4 2.27.3 2.31.6 3.13.2 3.30.6 3.43.3 3.60.5 3.99.2–3 1.42.4 1.70.5 1.83.2 2.7.3 2.32.10 3.13.4 3.33.1 3.49.2 3.86.1 3.101.1 3.109.5

Pompeius . . . senatus om. S (uerba 9) tota . . . refertur om. S (uerba 18) cuius .. legionibus om. S (uerba 7) interim . . . adducunt om. S (uerba 8) dum . . . legatus om. S (uerba 8) neque . . . deficiebant om. S (uerba 7) -re cognouerat . . . 2.19.3 custodias om. S (uerba c.107) abest . . . mille om. S (uerba 9) suas . . . 2.28.1 Varus om. S (uerba 11) an . . . tegenda om. S (uerba 10) simul . . . incidit om. S (uerba 8) per . . . peruenit om. S (uerba 18) haec uerba abhinc ad 3.48.4 iaciebant om. S (folium unum?) cum . . . 61.1 erant om. S (uerba 11) -cimus . . . dixe- om. S (uerba c.16) quas MUS : om. TV facit MU : fecit Sc in marg. (f*it Sac una littera rasa) : om. TV et rursus aliae om. TV ad cognoscendum US : ad ag- M : om. TV religio MmrUS : -one M : om. TV hos MUS : om. TV tolli . . . Scipio om. TV (uerba 19) non bona . . . multitudine om. TV (uerba 10) pelleretur MUS : om. TV prius Cassius ad Messanam MUS : om. TV accepto . . . alter om. TV (uerba 9)126

The relative number of words omitted in these two lists suggests that π is a more faithful copy of ν than S is; this is borne out by other sorts of singular innovations, which are vastly more numerous in S than in π, as can be seen in the Appendix critica. Where ν and μ diverge, ν is more often in error than μ is, either by preserving a corrupt archetype or by innovating. 126 Textual variation between π and μ in some of the many passages omitted by S make it unlikely that these omissions were filled by contamination from the μ branch. There are also some small discrepancies between S and μ in text omitted by π (see the Appendix critica for 3.33.1, 3.49.2). It is therefore likely that in all of the passages listed above ν’s text was complete and transmitted to one of its two descendants.

66

Part 1. Prolegomena

3. Pi A large number of shared innovations both link T and V and separate them from the rest of the tradition, more than 150 for the BC alone; the traditional siglum for the source of these innovations is π. Among the roughly forty-five unique omissions that T and V share in the BC, for example, are five involving more than a word or two: 1.83.2 et rursus aliae, 3.33.1 tolli . . . Scipio, 3.49.2 non bona . . . multitudine, 3.101.1 prius Cassius ad Messanam, 3.109.5 accepto . . . alter.127 These unique omissions are too long to have been repaired in S by conjecture and are therefore unlikely to have been inherited by π from their common exemplar, ν; on the implausibility of contamination as an explanation for text in S but not in its sibling π see n. 126 above. The separative errors in the BC also include a handful of inversions, another variety of innovation that, once made, tends to persist in an uncontaminated tradition: 2.22.1, 2.23.1, 2.26.1, 2.26.4, 2.32.14, 2.34.6, 3.1.5, 3.69.2. Finally, among the innovations unique to these two manuscripts in the BC are some two dozen that repair a corrupt exemplar and, once made, are unlikely to have been removed except by accident (see pp. 76–7); these too must have originated with π, not its exemplar. For the hyparchetypes discussed above, μ and ν, it was a simple matter to show the independence of the two earliest descendants of each (M and U; S and π). This is not the case for π.

a. The relationship between T and V Defining the relationship between T and V precisely is difficult. More specifically, determining whether V derives from T or from T’s exemplar. Hering believes that V is a descendant of T, while Brown restates the conventional view that they are independent copies of π. At issue is whether V’s readings help constitute the text. The relationship between the two manuscripts, whatever it is, is the same for all portions of the Caesarian corpus. In the fourteen books of this corpus T and V diverge too many times to count. For the three BC books alone the divergences number more than seven hundred. But the divergences that help us see V’s 127 Hering (1963, 40) lists six substantial omissions common to T and V in the BG. There is another at BHisp 40.2.

IV. The witnesses

67

parentage are a tiny and rather murky subset of this vast collection. There are no obvious signs of V’s dependence on T such as anomalies in V that can be connected with something specific to T: the omission of one or more of T’s lines, or the inclusion of extraneous material in T such as marginalia, or a reflection of physical damage to T.128 If, however, V is a descendant but not a direct copy of T, as Hering suggests, any such signs might well have been obscured.129 The most illuminating divergences available to us are innovations in T not shared by V, particularly those that V could not have removed by conjecture. In passages where T has innovated, the source of V’s archetypal reading must be either π (i.e. transmission), or another branch of the tradition (i.e. contamination), or coincidence. Where contamination and coincidence are implausible, we have evidence that T and V are independent descendants of π. The evidence is scanty, but working through even this limited sample is time-consuming, so before doing so we should consider what is to be gained.

b. V’s contribution Hering’s view is that it is not worth the trouble of sorting out V’s precise relationship to T. If V is a descendant of T (as in the 1963 stemma), and if all of V’s non-T readings can be shown to be errors, conjectures, or readings attested elsewhere in the tradition, then V is manifestly eliminandus except as the source of good conjectures (In the BC more than 40 of V’s conjectures have been adopted; see p. 94 below.) But even if V is T’s sibling (as Hering suggested tentatively in 1987; see p. 16 above), the manuscript is not much use, he says, since V will only be the unique representative of the archetype when the readings of the other four significant manuscripts (ABTU for the BG; MUST for the other Bella) differ from V and each other, or when the two other manuscripts in V’s branch and the consensus of the two in the other branch (T, U, and AB for the BG; S, T, and MU for the other Bella) differ from V and each other (1963, 28). These scenarios, he maintains, are likely to be rare. (In the BC they are in fact nonexistent.) However, a reading need not be unique to be useful. 128

See Hering (1963, 25); Brown (1972, 27). His suggestion (1963, 26) is based in part on the substantial time gap between the twelfth-century V, and T, which he dates as ‘10./11. Jh.’ Brown (1972, 85–6) and Munk Olsen (1982, 1.44), however, date T to the second half of the eleventh century. 129

68

Part 1. Prolegomena

Brown points out (1972, 28) that in the more than two hundred instances when T is in error and V agrees in a correct reading with another branch of the tradition, V’s role is ‘to aid in the reconstruction of the text of the archetype, thereby illustrating the operation and certainty of the stemmatic method’. In Brown’s tripartite stemma V’s stemmatic contribution is obviously greater than it is in our bipartite stemma, in which V is essential for the reconstruction of the archetype only where both T and S are independently in error and V alone provides evidence in the ν family of agreement with the μ family. This general scenario (MUV or MV or UV vs. S vs. T) occurs more than a dozen times, but for fewer than half of them is it even plausible that V got its reading from its exemplar rather than via conjecture or contamination, and in no case is the argument compelling.130 Furthermore, the possibility that the archetype contained corrections sometimes obscures the source of V’s readings.131 (In what follows I do not consider coincidence, since the numbers, small though they are, seem too large to make coincidence a convincing global explanation.) The real utility of V—if independent of T—lies elsewere. First, in its contribution to the reconstruction of ν. Where ν and μ disagree and T has gone astray, we can get the reading of ν from the agreement of S and V. This scenario turns up a handful of times in the BC.132 In the somewhat more numerous places—nearly twenty— where T’s innovation matches a good reading of μ against the agreement of S and V, V determines how one reconstructs the archetype. Consider, for example, the effect of an independent V on the stemmatic reasoning about the following passage: BC 1.39.3 Simul a tribunis militum centurionibusque mutuas pecunias sumpsit; has exercitui distribuit. centurionibusque MUT : centurionibus SV (L) 130 Possibly archetypal readings in V, against S and against T: BC 1.70.1 utri MUV : utti S et Tac ut uidetur : ut Tc (uti L); *1.71.3 misisse MUV (L) : missis sed T : misis S; 3.110.6 hi omnes MUV : homnes S : homines T (omisit L). I mention L’s readings here and below in this section because, as we will see, L has been proposed as a source of contamination in V. However, in the one spot where V and L agree here (marked with *), conjecture in V remains plausible, since L (a descendant of S contemporary with V) obviously arrived at its agreement with the μ family by conjecture. 131 V’s numerous successful unique emendations to the text suggest that V’s agreements with μ are perhaps more likely to result from innovation than from the transmission of an archetypal correction. But each instance needs to be analysed. 132 E.g., BC 1.43.1 planities MmrSV (L) : -ni(t/c)ia MUT; 3.28.2 summissis SV (L) : summisso T : summis MU; 3.51.7 tormentumue MU : -tum T : -to SV.

IV. The witnesses

69

If we take V into account, we reconstruct ν as lacking -que, since independent omissions by S and V are implausible. This means that T’s agreement with MU comes about by innovation not by transmission.133 In this scenario it is uncertain whether the archetype did or didn’t contain -que. If it didn’t, the addition of the connective by μ (and T) is easy to accept, but so is the possibility that the archetype did contain -que and that ν overlooked it. If we don’t take V into account, we conclude that the archetype contained -que (agreement of μ and T) and that S made a careless error. That is, the presence of V in the stemma as a sibling of T reverses the conclusion reached when V was not considered.134 Second, V sheds light on the nature of the other members of its family. Specifically, it shows us that T may have arrived at the truth by conjecture (as in the example just discussed) more frequently than the few instances in which T alone successfully repaired a corrupt archetype would lead one to expect (see p. 92 below). And that S may have departed from its exemplar somewhat less frequently than the many divergences between T and S would lead one to expect. Overall, in providing a clearer picture of ν, and therefore of disagreements between μ and ν, an independent V prevents us from treating as archetypal (ω) readings that are only possibly archetypal (μ vs. ν, or ν vs. μ).

133 Unless the archetype had a correction here and both readings were transmitted to ν and π and hence available to T to select between. In the present passage this is not implausible, and the fact that T makes rather few uniquely successful innovations may tilt the balance against conjecture here and in the passages listed in the following note (but see. n. 146 for other connectives added by T). However, changing the explanation of the source of T’s reading doesn’t change the point at issue here, since without V the question of T’s source wouldn’t even arise. 134 The other instances where V makes a potentially decisive contribution to stemmatic analysis are these (but see the preceding note for a caveat): BC 1.70.2 exercitu MUT : -ui SV (L); 2.12.4 quin MUT (L) : qui in SV; 2.16.1 posset MUTVc : possit SVac (L); 2.16.2 coniceretur MUTVc : -rentur SVac (L); 2.19.1 praegreditur MUT : prog- NV (L) (S deest); 2.31.7 ut media MUT : media SV (L); 3.6.3 quietam MUT : que aetiam S : qui etiam V (quae etiam L); 3.14.1 accipit McUT : -cep- MacSV (L); 3.14.2 recipit MUT : recepit SV (L); 3.14.3 exiguo MUT : in e- SV (L); 3.41.1 postridie MUT : post trid(i/u)(a)e SV (post triduum L); 3.82.3 ac de MUcT (L) : ac UacSV. Similar but more complicated are these: BC 1.66.1 iis . . . correptis MU : his . . . -ti T : hi . . . -ti SV (hi . . . correpti sunt L); 1.73.3 et2 MU : ut T : om. SV (L); 2.28.2 primam . . . memoriam M : -a . . . -iam UT : -a . . . -ia SV (L). In other passages more complicated still V affects the analysis without being decisive: e.g. BC 1.7.5 legibus mUT : legionibus SV (L); 1.48.2 ambo MUT : ambos SV (L); 1.82.1 tertio MUT : -ia SV (L).

70

Part 1. Prolegomena

c. Contaminatio I begin by considering Hering’s ‘Beweisstück’ for contaminatio in V (1963, 27). It is one of a number of lists of Gallic tribes; as is usual in such lists, many of the names are unique occurrences, and textual problems, particularly of word division, abound. I give Hering’s text. The readings are presented in full to facilitate comparison. BG 3.27.1 Tarbelli, Bigerriones, Ptianii, Vocates, Tarusates, Elusates, Gates, Ausci, Garunni, Sibulates, Cocosates tarbelli bigerriones ptianii uocates tarusates elusates gates ausci garunni sibulates cocosates ABU with some small variations trabelli berones saniua catesta rusates aut scisi bulates cossates T135 tabelli bigerriones phticiani uocates tarusates gates ausci garumni liberones saniua catesta rusates auscisi bulates cosates V

V lists more tribes than does T, and shows partial agreement with the other β manuscript (U) and with the α family (AB) in so doing, but V also omits their Elusates. On the other hand, V has T’s garbled text saniua catesta rusates (cf. -anii uocates tarusates ABU). Furthermore, V’s attempt to combine discrepant (and corrupt) lists leads to doubled references to the Vocates, the Tarusates, the Ausci, and the Bigerriones (transformed into Liberones). Hering explains the difference between T and V here by saying that a predecessor of V supplied the information not in T as a marginal note, having found it elsewhere, and that this note was incorporated into the text that came down to V (1963, 27). Hering implies that the supplement was not available to T: the predecessor he speaks of occupies a generation between T and V. But what we read in T and V is also explicable if the aforementioned ‘predecessor of V’ was π: T will have copied, say, what was in the text of π, V both that and a marginal supplement in π.136 135

Hering doesn’t mention that T has a suprascript correction here, as follows (bold letters indicate innovations in Tc) : tarbelli bigerriones ptianii uocates tarusates frustates gates ausci garunni sibuzates cocosates. This is clearly not the source of the surplus text in V. 136 The supplement might even have been added to π after T was copied from it. One tiny indication that V preserved marginal material in π that T ignored is seen at BC 1.6.2 (saltem UT : statim saltem m : statim mmret Uin marg. : uel statim saltim V : om. S), where V, like m and U, has two archetypal variants, while T has only one. Another at BC 2.22.4 (tempestatis MUS : om. TV (post concursu add. V)), where V puts a word omitted by π back into the text, but at a spot where it makes no sense; tempestatis was presumably retrieved from π’s margin. We may get a glimpse of a correction in π in the bizarre collection of readings at 1.53.1 (multa rumores Oehler :

IV. The witnesses

71

In this reconstruction there is no need to assume contaminatio in V’s text, since π’s marginal note could have been a correction based on its source β and (indirectly) the archetype, whose more correct list (no duplicates—so far as we can tell) is reflected in both α (AB) and β (U); see also p. 51 above for the possibility of horizontal transmission between the family of π and the family to which U belongs. This passage therefore does not prove contaminatio in V, nor does it help determine the relationship between T and V. In the following sections the relative plausibility of contamination and transmission will need to be weighed. Hering does not identify the source of the contamination he posits in V, limiting himself to ‘aus einem anderen Zweig der Überlieferung’ (1963, 27). Since he doesn’t explore any further discrepancies between T and V, he never puts the contamination explanation to work, although it is implicit in his treatment of T’s omissions (on which see below).137 Only one specific source has been suggested anywhere, namely, a descendant of S commonly called L (London, British Museum Additional MS 10084). Because of agreements between V and L that were surprising in view of the then-accepted stemma (basically bipartite, with S and its relatives NL in one branch, MUTV in the other; see above), the possibility of contamination of V by L was raised by Cupaiuolo in 1954, but then dismissed on the grounds that their agreements reflect a shared tendency to deliberate innovation (1954, 67). The matter was investigated more thoroughly by Brown in connection with her tripartite stemma, in which L and V again find themselves in different branches. She deemed the evidence of the BC to represent ‘insignificant preferred readings’ and ‘indecisive common errors and omissions’ (1972, 27–8). The hypothesis of contamination of V by L is rendered still less appealing by Brown’s redating of L from the

multarum M : multa rumor MmrU : multorum ora S : multarum rumore orat T : multa rumore V), where it looks like V, with multa rumore, has made sense of a correction to an obviously faulty passage while T has transmitted an amalgamation of the original text (perhaps multarum orat?) and the replacement. 137 He does suggest, apropos of BG 6.21.5 in fluminibus perluuntur et pellibus aut paruis renonum tegimentis utuntur (renonum V : rhenorum A : rhenonum BU : renorum T), that V corrected T’s -orum with -onum from the family of B or U (1963, 30). But here too the indication is very general, and the case for conjecture is strengthened by the fact that Q, a relative of A, also has renonum.

72

Part 1. Prolegomena

eleventh century to a date closer to that of V: ‘the end of [the eleventh] century or the beginning of the twelfth’ (1972, 15; Munk Olsen agrees with the later date for L, ‘S. XII1’). Although the relevance of L to V’s text now seems tenuous at best, I cite its readings in this section. As we proceed it will be worth bearing in mind the desirability of identifying the source of text that V owes to contamination.

d. T’s omissions, filled Hering lists four substantial omissions in T’s text of the BG where V’s text is complete (1963, 40): BG 1.38.5 (5 words), 1.53.2 (6 words), 3.26.2 (13 words), 4.19.1 (6 words). These omissions were filled in T itself by a hard-to-date corrector drawing on a manuscript from the α branch of the tradition.138 Hering leaves open the possibility that V got its text from the corrected version of T, but he has contamination to fall back on if the corrections were added too late for V. However, it is unlikely that V’s source is a post-correction T. In one of the passages on Hering’s list there is a discrepancy between α and β readings, and here V follows β in omitting ex (BG 3.26.2, the omitted words in bold, illi, ut erat imperatum, eductis iis cohortibus, quae praesidio castris relictae integrae ab labore erant, et longiore itinere circumductis, ne ex hostium castris conspici possent . . . ); Vc supplies ab (for ex) above the line. This suggests that V’s source was not the corrected T but either π (transmission) or a manuscript from the β branch of the tradition (contamination, presumably from U or one of its descendants).139 The omissions in T’s text of the BG therefore do nothing to clarify the relationship between T and V. In the BC Brown finds two passages indicative of an independent V. Here, as she explains it, V’s scribe has ‘copied T’s exemplar . . . with better results’ (1972, 28). In the first of these T has omitted the second of three rhetorical questions beginning with the word cur (8 words). 138 An ante quem date is established by the presence of the corrections in a thirteenth-century manuscript now in Leiden (Bibl. publ. lat. 38D), but no post quem date is given by Hering (1963, 39–40). 139 Further evidence that V does not descend from a post-correction T comes from the four substantial omissions where V failed to benefit from the corrections in T (see Hering [1963, 40]). For U’s extant progeny see Brown (1972, 48).

IV. The witnesses

73

BC 1.72.2 (omitted words in bold) Cur etiam secundo proelio aliquos ex suis amitteret? cur uulnerari pateretur optime meritos de se milites? cur denique fortunam periclitaretur?

Here the particle denique, attached to what is apparently the second of two items in a list, suggests that an omission has occurred, since Caesar elsewhere uses denique in longer lists. This anomaly might have prompted a search for the missing text, so contamination remains a possible (if not particularly likely) explanation.140 And there is nothing distinctive about V’s text of these words that would help identify their source. So this example is inconclusive. But the second passage, where T skips from one occurrence of acie to another, omitting twenty-four words, is another matter. BC 3.89.3–4 (omitted words in bold) Sinistro cornu Antonium, dextro P. Sullam, media acie Cn. Domitium praeposuerat. Ipse contra Pompeium constitit. (4) Simul iis rebus animaduersis quas demonstrauimus, timens ne a multitudine equitum dextrum cornu circumueniretur, celeriter ex tertia acie singulas cohortes detraxit . . .

Here the likelihood of the scribe of V (or its exemplar) realizing that T’s text was lacunose is small: the passage as it stands in T construes perfectly well, although its content would perhaps surprise on close examination. In other words, there is no pressing cue for contamination. And here, as often, the hypothesis of uncued contamination leads to a peculiar inference about scribal behaviour: the scribe was so scrupulous a collator as to catch a nearly invisible omission like this one and the others discussed below unique to T, yet he failed to repair dozens of obvious omissions that T (according to my stemma) inherited from π, or (according to Hering’s stemma) originated.141 It is simpler to argue that V got these words from π. This passage also offers evidence against the idea that V drew its non-T readings from L. That theory is particularly hard to credit here, since there are two significant differences between V and L in this passage, differences that are unlikely to be copying errors. BC 3.89.3 praeposuerat MUV : proposuerat S (L) : om. T

140 The four BG omissions mentioned above, which seem to result from eye-skips, are likewise detectable. 141 On π’s omissions see p. 65 above.

74

Part 1. Prolegomena

Here V, in altering the verb’s prefix (supposing it to be copying from L), has simultaneously returned to what is probably the archetype’s reading, from which S (with its descendant L) has deviated. Conjecture by V cannot quite be ruled out here, however, because this usage of praeponere (with a troop formation in the dative) occurs elsewhere in Caesar (4x, including two passages in BC 3). But the second difference is more telling. BC 3.89.4 rebus animaduersis MUSV (animaduersis rebus L) : om. T

If V (or its exemplar) filled T’s gap by consulting L, it also normalized L’s word order.142 V does alter its exemplar’s word order quite often,143 but how likely is it that a scribe would take the trouble of seeking L’s help in improving the text and then ‘correct’ his helper en passant? I suppose V’s exemplar could have used L, and V itself been responsible for the word order change, but the explanation for V’s readings is getting increasingly unwieldy. There remains the possibility that the missing words were supplied from the μ branch, with which V agrees exactly. So it will be worth looking more closely at the behaviour of our scribe(s) with an eye to the plausibility of contamination. The other omissions in T are of much briefer compass, no more than a word or two, and V’s retrieval of the text of the archetype (or of ν) can often be explained as a conjecture.144 Here I consider three of these brief omissions from the BC and other Bella where conjectural repair is implausible since no omission is visible. BC 1.10.1 Acceptis mandatis Roscius cum Caesare Capuam peruenit ibique consules Pompeiumque inuenit; postulata Caesaris renuntiat. (2) Illi deliberata respondent scriptaque ad eum mandata per eos remittunt . . . cum mUacSV (L) : cum lucio Uc per compendium : a T | per eos remittunt mUSV (L) : permittunt T

142 The only ablative absolute in the corpus in which the participle of animaduertere precedes its substantive is BC 1.81.3, where the substantive has a dependent genitive: animaduerso uitio castrorum. 143 On V’s inversions see n. 183 below. 144 E.g. for the BC, 1.55.1 cohortibusque] -que om. T; 1.57.2 numero om. T; 1.61.6 castraque] -que om. T; 1.74.5 illi om. T; 2.5.3 manus om. T; 2.9.3 trabes om. T; 2.23.5 ad om. T; 2.33.1 dicentem om. T; 2.34.5 sibi om. T; 2.43.3 imperiumque] -que om. T; 2.44.1 rebus om. T; 3.13.6 et om. T; 3.15.1 erat om. T; 3.69.4 et2 om. T; 3.86.1 diebus om. T.

IV. The witnesses

75

Here T, confused by the two Caesars in this passage (L. and C. Caesar respectively), makes two interrelated innovations, producing a superficially satifying repair to a somewhat puzzling text by collapsing the two men into one. (Uc makes a different repair.) V follows T in neither change, and in fact agrees with the archetype. The scribe could not have retrieved this reading by conjecture. If he did so with the help of another manuscript (be it L or a book belonging to μ’s family), this passage did not prompt his consultation of it. So transmission via π is a more likely explanation than contamination. Equally unmotivated is V’s restoration of absentem in the following passage, where it is required by neither syntax nor sense: BC 1.76.1 Quibus rebus confectis flens Petreius manipulos circumit militesque appellat, neu se neu Pompeium absentem imperatorem suum aduersariis ad supplicium tradant obsecrat. absentem MUSV (L) : om. T

A similarly uncued addition by V is found in the BAfr. At BAfr 26.3 we are told that Caesar sent a letter containing orders to Sicily. Events move quickly, and in the next sentence he gives new orders without waiting for reinforcements: BAfr 26.4 atque ipse erat in tanta festinatione et expectatione ut postero die quam misisset litteras nuntiumque in Siciliam, classem exercitumque morari diceret. in Siciliam MUSV (L) : om. T

Here again the words omitted by T are not required by syntax or sense, and might even be felt to be a gloss. In sum, it seems more likely that for all of the omissions discussed above, with the possible exception of BC 1.72.2, V got its text in the same manner as the other witnesses that have it, namely, by descent from the archetype. For V, this means via π, not T. This omission-based evidence is complemented by that pertaining to hard-to-spot inversions145 and

145 Nearly invisible inversions in T’s text of the BC: 2.32.2 inquit omnia] o- i- T; 2.35.1 aliquid uelle] u- a- Tc : uellet a-Tac; 2.35.2 circumuentus interficitur] i- c- T; 2.37.2 omnibus rebus] r- o- T; 3.85.3 cotidianam consuetudinem] cot- cotidianam cons- S : conditionem cot- Tac : cons- cot- Tc. Note that the corrector of T, when working on two of these passages, did not correct the word order. V has the archetypal text in all of these passages.

76

Part 1. Prolegomena

additions,146 where V, if dependent on T, innovates in such a way as to arrive at precisely the text of the archetype. It remains to consider V’s behaviour where T’s unique innovations produce an acceptable text.

e. T’s usable innovations, ignored Below I give two lists of passages where T, deviating from the archetype, produces a usable reading that V ignores. Only occasionally do these readings make it into the text, but there is nothing obviously wrong with them. They are too numerous to discuss in detail. The first list is a selection of T’s usable innovations in the BC. In order to make the list as perspicuous as possible I only include passages in which the usability of T’s reading is obvious; there are many more where one could make a case that T’s reading makes sense. Most of the innovations listed here represent (near-) synonyms or adverbs; a few involve changes of subject that make sense in context. A bracket follows the reading of the archetype. 1.4.3 1.7.8 *1.8.3 1.14.4 1.18.6 1.54.4 1.58.2 1.63.1 1.70.2 *2.7.4 2.30.1 2.33.1 2.34.4 2.37.2 2.39.4 2.40.3 3.12.1 3.27.1 146

re publica per compendium] rebus T conuenerant] uen- T ut T : ne m : om. USV productos] dedu- T ad pompeium] a pompeio T institutum] instituit T artificiis Tc : -cis MUSTac : -cum V coniungunt] iun- T deducta] ducta T defensionem T : -es MUSV conuocato] uocato T neu Mmr : ne ubi MUS : necubi T non male : nec ibi V uides] uidens T ac] et T equites] milites T deducit] ducit T eius] cuius T hic] huic T

Hard-to-spot additions in T’s text of the BC: 1.80.4 celeriter] celeriterque T; 1.85.9 aetatis] etiam a- T; 1.87.3 atque] autem a- T; 3.1.4 item] itemque T; 3.80.3 longinquam] longinquamque T; 3.88.5 castris] in c- T. V has the archetypal text in all of these passages.

IV. The witnesses 3.49.3 3.56.2 3.79.3 3.101.7 3.106.1

77

a(d/g)gesserat] aggre- Vac : adiecerat T abiectis] proie- T etiam] autem T arbitrabantur] arbitrabatur T a Q.] a q. T: a quae S per compendium : quae V : atque MU

This list shows V deviating from T without cause, and complements the evidence adduced above in connection with T’s invisible omissions, transpositions, and additions. Drawing on evidence from the other Bella I give a supplementary list of T’s usable innovations where V agrees with other branches in a corrupt reading (marked above with *). Vertical transmission is a much more likely explanation of V’s readings in such spots than is contaminatio. BG 2.34 dicionem edd. : deditionem ABV : ditionem UT BG 3.7.4 essuuios edd. ex 2.34 : esubios AB : unellos T : unellos sesuuios UV BG 7.3.2 ubi que Hering : ubicumque Seel : ubique ABUV : ubi ST BAlex 34.5 in ciliciam UT (L) : in ciliam MS : in siciliam V BAlex 64.3 in UT (L) : om. MSV BAfr 36.4 potitus MUT : potius SV (L deest) BHisp 21.2 e quibus T : de equitibus S : ex equitibus M : equitibus UV (L deest) BHisp 29.8 desinunt T : deseruant S : deserunt UV (ML desunt) BHisp 41.2 uiuos T : uiros RSV (MUL desunt) V’s failure to adopt T’s usable innovations supplies either more evidence of the hyperscrupulous remover of T’s innovations who somehow failed to catch the errors of π, or evidence that V derived from π the errors and the good readings discussed above, while adding more than a few of its own in both categories (see above). The latter alternative seems to me distinctly more likely. I therefore report the readings of V throughout. C. THE EXTANT MANUSCRIPTS

1. M M is a manuscript in the Biblioteca Medicea Laurenziana in Florence, shelf mark Plut. lat. 68.8. M contains the fourteen books of the corpus

78

Part 1. Prolegomena

Caesarianum on ff. 2–163; the first folium contains some blotted-out material but is otherwise blank. The text of the corpus Caesarianum is written on parchment sheets of about 334 x 245 mm, with one column per page. The gatherings of this book come from different phases. Two are of humanistic origin and unimportant for the constitution of the text (ff. 17–24 and 157–63, likewise f. 2). The majority are dated to the tenth–eleventh centuries: ff. 3–16, 25–116, 125–32.147 The remainder are dated to the twelfth century: ff. 117–24 and 133– 56.148 The latter two sets are reported as M; the relationship between M and its sibling U appears to be the same for both sets. The earlier pages are written in a neat Carolingian minuscule. Each page contains about 29 lines, each line about 55–60 letters. In the last gathering of these early pages the text has faded unevenly, occasionally making the distinction between original text and correction hard to determine. The later pages are much more densely written, with about 35 lines per page and about 90–100 letters per line. In these pages decorated initials mark the beginning of books.149

a. Contents 2r–92v 93r–104v 105r–116r 116r–133v 133v–143r 143r–154v 154v–156v 157v–163v

BG 1–8 BC 1.33.3 -duum disputationibus – end BC 2 BC 3 BAlex BAfr BHisp 1.1–22.5 petiit ab end of the BHisp in a later hand

The beginning of the BC is missing and unreplaced in M. Two descendants of M survive that were copied before these pages were lost. Both are written in a Beneventan or Beneventan-influenced script, which suggests that M was in Italy when the copies were made in the twelfth and early thirteenth centuries. The earlier, a codex preserved in the Biblioteca Vallicelliana, B 45, was discovered by Virginia Brown (1972, 78–81). She dates Vall. to ‘c.1100’ and For the ‘s. X/XI’ date see Brown (1972, 83) and Munk Olsen (1982, 40). See Brown (1972, 83). 149 For details about collation, provenance, and ownership see Brown (1972, 83) and Munk Olsen (1982, 40). 147 148

IV. The witnesses

79

provides a collation of BC 1.1–33.3. Slightly later is the manuscript m (Florence, Biblioteca Medicea Laurenziana, Plut. lat. 68.6) that is traditionally used to supply the missing chapters of M; it is dated to the late twelfth century or early thirteenth century.150 The differences between the two in the missing chapters are trivial.151

b. Corrections Most of the BC is preserved on the early pages of M: ff. 93–116 and ff. 125–32, for a total of sixty-four pages. Corrections were added to 150

Lowe (1914, 339), Brown (1972, 82). Brown suggests that m is a copy of Vall. (1972, 78–80). In my view the evidence that they are independent copies of a copy of M is at least as strong. There are divergences from M common to Vall. and m that are unlikely to have arisen independently. For example, at BC 3.5.1 both Vall. and m have the nonsensical text frumenti iura where M follows the archetype with frumenti uim. Their common source also fails to implement correctly a correction in M to BC 3.9.3 infecit: the correction involves deleting ‘in’ via sublinear dots and adding con- via a tiny suprascript abbreviation, but in both Vall. and m the reading is fecit, not confecit. An even more striking misreading of a correction occurs at BC 3.15.4, where Mmr emends M’s rerum (a corruption of rorem) to zephirum, which appears in both Vall. and m as triumphum. Furthermore, although most of M’s rare marginalia from both its earlier and its later pages are reproduced exactly in Vall. and m, where they are written in the same hands as the base texts, the marginal comment at BC 3.88.1 is supplemented in both Vall. and m, which have ‘senatus consulto’ in addition to M’s ‘belli dispositio’ (M 128v/m 124r/ Vall. 124v). All of these phenomena are consistent with both hypotheses mentioned above. But Vall. also has singular errors in which it is not followed by m. For example, at BC 1.17.2 Vall. and m diverge differently from the archetypal reading: magnumque USTV : magnoque Vall. : magumque m. And at BC 2.8.2 the marginal comment ‘de machinis’ is in M and m but not Vall. (M 106v/m 96r/[Vall. 96r]). These phenomena run counter to Brown’s hypothesis. A third explanation would be that Vall. is a copy of m. Against this is the fact that m has (at least) two substantial lacunas (BAlex 71.1 quo celerius – BAfr 35.2 in magnum; BHisp 38.2 Pompeius umero – 42.7 et uirtute) where Vall.’s text is whole. Strictly speaking, this only argues against the possibility that Vall. is a copy of m in its current state, but the relative dates of the manuscripts and the fact that Vall. ignores m’s fine innovation at 1.5.5 (lenissimis for leuissimis) support the stronger claim that Vall. is not a copy of m, period. I also considered the possibility that Vall. and m are independent copies of M itself made before the (hypothesized) tenth-century pages were replaced by a twelfthcentury copy. Against this is the fact that they implement corrections noted on the extant twelfth-century pages of M. One might explain this by assuming that the corrections bring the extant M back into alignment with the hypothesized tenthcentury exemplar of all three texts. But this explanation seems very unlikely where the reading in M represents an archetypal corruption: one would have to assume that the innovation in Vall., m, and Mmr, which ex hypothesi was in the tenth-century version of M, was ignored by the copyist of the twelfth-century version of M (e.g. BC 3.42.5 in Petram Bücheler (in petra iam ϛ) : in porta v : del. Mmr : om. Vall., m). 151

80

Part 1. Prolegomena

these pages in several phases; the total number of corrections is approximately seven hundred. The first phase of corrections seems to be the work of the original scribe.152 These corrections number roughly fifty and generally recapture the reading of M’s exemplar.153 The corrections by one or more later hands are much more numerous.154 They are also more variously motivated. Some certainly or probably recover the reading of the exemplar.155 But a great many introduce innovations into the text. Some of these are successful repairs to problems in the archetype.156 But most merely address 152 Corrections by the original scribe (Mc) are indicated by additions and overwriting in the hand and ink of the base text (
thirty instances), erasures (especially where the surrounding letters are undamaged;
ten instances), some sublinear dots (especially where combined with supralinear dots and/or characteristic additions;