We use cookies to distinguish you from other users and to provide you with a better experience on our websites. Close this message to accept cookies or find out how to manage your cookie settings.
Published online by Cambridge University Press: 11 February 2009
By the time his Teubner De Officiis had reached its fourth edition, C. Atzert was running out of hope. So great was the number of the still-accumulating manuscripts ‘ut paene desperaverim in seligendis et ordinandis eis’. In fact, there were hundreds more of which he knew nothing. My own list approaches seven hundred in all, and there will be others lurking still. The present paper aims to impose some order on this vast army.2 It sketches in new detail the interrelationships of the old witnesses to the ξ tradition, and the transition to a twelfth-century Vulgate (Ψ); it throws light on the early story of the ξ family; and it makes a start on the task of showing how the f stream, from the twelfth century onwards, fitfully mingled with the Vulgate (for the stemma, see Fig. 1). It could never have been written without the help of Marina Luttrell and of countless friends and colleagues who provided information about manuscripts I could not see for myself
1 M. Tulli Ciceronis De Officiis quartum recognovit Atzert, C. (Leipzig, 1963), xi–xiiGoogle Scholar. Throughout the article, I give references to this edition, normally providing line as well as book and section numbers.
2 For manuscripts up to and including the twelfth century, I cannot add much to the list in Olsen's, Birger Munk invaluable L' étude des auteurs classiques latins aux XIe et XIIe siècles i (Paris, 1982), pp. 115–16Google Scholar, referring forward to the detailed descriptions on pp. 135–320. For an addendum see ii (Paris, 1985), p. ix. I normally follow Munk Olsen's datings, and add his reference numbers in brackets, thus freeing myself to abbreviate references to manuscripts. In an Appendix, I list the later manuscripts (not including extracts and abbreviations) at present known to me, with acknowledgement to all those who have helped me.
3 Here, and in my forthcoming Oxford Classical Text, I refer to the families long known as Z and X by the corresponding Greek letters, wishing to keep Roman capital and lower-case letters for individual manuscripts, Greek for groups.
4 See Fedeli, P., ANRW 1.4 (1973), 376–87Google Scholar for a statement of the present state of play in this tradition.
5 B has been very thoroughly corrected, by more than one old hand. Working from a microfilm, I have not ventured to distinguish the hands, and I refer to all corrections as B p.c. But it may be useful to signal a number of types of correction: (a) The removal of errors going back not to ξ but to β (the ancestor of the sub-group to which B belongs). For examples see n. 9 below, (b) The importation of errors of V: which indeed might have been the source of the corrections given under (a). Note, e.g. 1.72,12 adiumentum (also ξ); 2.87,22 del. ulla; 3.70,14 nam] num; 3.79,13 del. ilium; 3.113,1 mansisse. At 3.44,2 B p.c.V correctly give possit where B a.c.Pξ have posset. (c) The importation of errors familiar from the twelfth-century Vulgate (and so often found in (c)p also: see below, pp. 226–7), e.g. 1.41,27 dictum (est) (so cp); 1.49,28 morbo] modo (so L pc.p); 1.64,28 quo ] quod (so p); 1.97,23 reliquorum (so cp); 1.103,15 fortuitu (so cp). In the process B p.c. may be joined by P p.c, e.g. 1.130,8 dicere (so p). Once at least the truth is hit upon thus: 1.38,25 quorum (so cp). All these readings are found in a single representative Vulgate manuscript, Klosterneuburg 778 (except for that at 1.41, where the whole sentence is omitted), (d) Coincidence in truth and error with f against ξ. I agree with Atzert (ed. 4, xix) that we should not assume use of a ξ source. Conjecture and chance seem to be enough to explain the phenomena, especially in the case of banalization like 1.9,11 inquirunt (so P p.c.) and 3.38,29 inusitata (so P). (e) B p.c. occasionally attains truth unknown to ζξ, either by conjecture or from the Vulgate, e.g. 1.32,5 om. quid/quod; 1.150,7 cetarii (so Klosterneuburg 778); 2.43,11 qui (so also P p.c. and Klosterneuburg 778). At 1.92,6 it corrects the truth (the perhaps accidental parta) back to the reading of ξξ.
6 See esp. Schwenke, P., Philologus Supplementband 5 (1889), 397–588Google Scholar; also Beeson, C. H., CPh 40 (1945), 201–22Google Scholar. Beeson is quite wrong (pp. 217–18, 221) to countenance Mollweide's old view that K is close to T (for which see below, p. 229); Sabbadini, R., Storia e critica di testi latini (Catania, 1914), p. 160Google Scholar, correctly saw K as ‘indubbiamente’ of class 's readings led him to think it at all influenced by ξ.
7 Testard's β. Another German manuscript, Gotha, Membr. 11.198 (below p. 220), is related to B2 after starting as an ordinary Vulgate text.
8 If this provenance is correct, it would be tempting to think of the ancestor of our group as being a German manuscript, lent to Corbie (see the remarks of Reynolds, L. D. in Reynolds, L. D. (ed.), Texts and Transmission (Oxford, 1983), p. xxix n. 107)Google Scholar.
9 e.g. 1.8,3 atqui (B a.c); 1.42,7–8 om. et qui…convertant (B a.c); 1.47,13 diligimur (B a.c, ut vid.); 1.48,24 <si> modo; 1.96,12 generis; 1.160,16 consideratio (om. actio) (B a.c); 2.32,1 praeter eas; 3.60,17 sane ] satis (B a.c). On corrections in B, see above, n. 5.
10 All the errors given in the previous note except that at 3.60, where I is lacking. Schwarz, H., Philologus 54 (1895), 177CrossRefGoogle Scholar, thought I so near to B that complete collation was pointless. ‘It does not appear to have been copied from B’: Clark, A. C., Anecdota Oxoniensia (Oxford, 1892), p. lxivGoogle Scholar.
11 Those (see n. 9) at 1.8,1.47 and 1.48. It agrees with B a.c.KP2I at 1.91,23 versentur. I cannot see that it is near to I, despite the stemma in M. Testard's Budé i (Paris, 1974), p. 89.
12 e.g. 1.103,18 om. illo; 1.103,22 honestis (so Lp); 1.111,21 quidem; 2.13,22 necessarium.
13 To be precise: B does not capitalize, though the verses are placed in two lines with three blank lines following. P capitalizes only to viam. V preserves the line ends, as no doubt the hyparchetype did. It also capitalizes 1.121 optima…liberis.
14 It may be a corrected descendant of B, with whom it shares against the others, e.g. 1.104,28 vocamus (vocant is transmitted); 1.154,20 contemplatique (but also P a.c.); 1.155,26 quia (B a.c.B2 a.c). At 1.13,6 it gives contentio (B a.c.EI, but also V) as a marginal variant. It seems likely that V had a hand in its make-up: it agrees with Va t 2.74,12 om. sit; 2.87,22 om. ulla (so Bp.c.: see above, n. 5); 3.10,23 imitando.
15 P, like B, has been much corrected, particularly in Book 1. I again distinguish not hands but some types of correction, (a) Apparent free conjecture: e.g. 1.4,15 ysocrate (m.rec); 1.22,2 mutuatione (marked as a variant); 1.75,11 a se adiutum; 1.146,16 dedeceat. (b) Correction to readings found in the later Vulgate and/or (c)p: e.g. 1.22,1 aliis ] alius; 1.35,6 conservandi <sunt> (sop); 1.35,10 aliquos (so p); 1.36,25 patiatur (so cp); 1.137,21 ipso (so cp); 3.44,28 iurationis (so p?). The first three of these errors are found in the Vulgate manuscript Klosterneuburg 778; 1.137,21 ipso is omitted in the Klosterneuburg book, but is common in other Vulgate texts. The matter is complicated by the interrelated facts that P in its first hand not seldom coincides with p and that the Vulgate owes something to the influence of P (below, p. 222). (c) Correction to truth or error found in ξ. Here again (see above, n. 5) it does not seem that there was direct use of a ξ source. Many of the changes to truth are obvious enough, and many of the errors are mere banalization (e.g. 2.56,30 apparatione).
16 Popp, E., De Ciceronis de officiis librorum codicibus Vossiano Q.71 et Parisino 6601 (Progr. Hof, 1893), pp. 17–21Google Scholar; Fedeli, P., Studi sulla tradizione manoscritta del ‘De Officiis’ di Cicerone, AFLB 10 (1965), pp. 5–13Google Scholar. Atzert in his fourth Teubner edition, p. xxi, thought the two gemelli, a conclusion towards which Popp in the end inclined.
17 See Sabbadini, , Storia e critica, p. 159Google Scholar.
18 Pa agree at e.g. 1.66,13 om. aut expetere oportere; 1.77,27 om. nonne; 1.118,10 otn. ea (added by a modern hand in P); 1.161,25 om. sit; 2.36,26 om. iis; 3.39,18 om. celare1; 3.63,18 om. enim; 3.77,17 om. Romano; 3.90,4 medicando. P is wrong where a is right e.g. 1.9,9 om. aut; 1.32,11 ora. quae; 1.43,10 visum iri] fieri; 1.128,15 om.sunt; 1.138,30m. viri; 1.158,26 om.quae; 2.6,21 om. dicere; 2.36,6 om. neutiquam; 2.57,17 om. cum; 2.65,15 om. qui…par esset; 3.60,15 om. ipsis. In all these places b follows P.
19 ‘Der älteste von den planmässig corrigirten codices’: Unger, G. F., Philologus Supplementband 3.1 (1867), 74Google Scholar. It appears in my apparatus as the earliest carrier of a conjecture at half a dozen places.
20 Its relationship to b, and therefore (though he did not know it) to P, was observed by its discoverer, Chatelain, E., RPh 5 (1881), 136Google Scholar.
21 2.76,7 om. tantum (the word is displaced in a); 2.76, 10–11 cessura (P a.c. +a); 2.79,2 sese (+a; B a.c. is uncertain); 2.79,4 adeptum (+a); 2.81,16 argys; 3.5,12 tune; 3.10,21 veneris. On the other hand, P and Q have their separative errors (P 2.76,5 illo ] illa; 3.5,12 ac ] et; Q 2.76,10 nihil; 2.81,20 quos <et>; 2.82,3 perfectique (so at: i.e. the correct reading is given as a variant). On the corrections in Q, see van den Bruwaene's, M. remarks in his review of Fedeli's edition (Latomus 27 (1968), 447–8)Google Scholar. At 2.74,14 enim <extranea> (from which Fedeli, (Gnomon 37 (1965), 263)Google Scholar made extraneae) is, besides being wrong, only the reading of a corrector (perhaps the same who added videri, equally wrongly, after populates at 2.78,21). His <a> coetu at 3.2,14 is not preferable to ξ's e. But he is right to join B in num quid at 2.76,11.
22 Popp (1893) was uncertain whether to think H gemellus or copy of V; Fedeli, , Studi, pp. 13–19Google Scholar, is clear that it is a copy.
23 The earliest known Italian manuscript of the De Officiis (the work is known to the eleventh-century catalogues of Bobbio and Montecassino). For V readings in other Italian books see below, p. 222.
24 For H, see Fedeli, , Studi, p. 15Google Scholar. For B3 note e.g. the avoidance of the following errors of V: 1.134,25 causam; 2.76,9 patriam; 2.86,10–11 praeteritis; 3.7,29 controversia <deos qui sine controversial>.
25 VHB3D share e.g. the following omissions: 2.74,12 sit (B3 before correction); 2.77,18 alia; 2.82,2 et eorum…tenebant; 2.87,22 ulla. D is not copied from H, for it avoids H's errors at e.g. 1.38,16 geratur; 1.136,12 om. alia; 2.83,11 civis boni; 2.84,16 habeam ] habeo; 2.88,26–7 om. et externa… externis1. Nor is it copied from B3, whose omissions it avoids at e.g. 2.85,1 et; 2.85,5 nostros; 2.88,1 ut.
26 P: 1.133,16 ipse; 1.135,2 destinandi; 1.138,3 om. viri; 2.23,27 tenentes; 2.23,3 maius; 2.25,24 ipse. B: 1.134,24 om. in; 1.135,30 nec <in>. V: 1.134,25 causam; 2.25,19 Thebenam modum diligere. For a collation, see Klein, J., RhM 22 (1867), 429–32Google Scholar.
27 I ignore the errors of B that a corrector of B has removed, some of them shared by B's relations (e.g. 1.136,12 recti B a.c.B2P2 I a.c, but not K or our fragment). This illustrates the difficulties posed by the careful correction of B. If it had not been corrected at all, we should be confident that the fragment was independent of B and its group.
28 Fedeli's Mondadori text (1965), p. 9.
29 Cf. 1.1,2, oporteat B; 1.7,25 pertineant B; 1.56,4 alliciat faciatque p; 2.8,14 verseris p. Contra, Fedeli, , ‘Sul testo del “De Officiis” Ciceroniano’, Ciceroniana 3–6 (1961–1964), 33–5Google Scholar.
30 Fedeli's edition can be consulted for individual readings of B, P and V. But I am afraid that caution is called for. When he and Atzert differ, it is by no means always Atzert who is wrong.
31 Bound to this group by an error at 1.156,13 (suam <sapientiam> prudentiam intelligentiamque) is Colmar frg. 690, s.XII2, French (93), containing 1.116–28 and 1.151–2.1.
32 To this category should probably be added the fragments of two texts in British Library, Harley 2567. The first, s.XII ex., French or English (239), scores 2 out of 2 where it is available to my test. The second, s.XII, German? (240), falls outside my sample passages. I make no assertion about the underlying text of either.
33 But this is also strongly affected by a β text. For the Leeds book see Martin, R. H., CQ 1 (1951), 35–8CrossRefGoogle Scholar.
34 All those I list that I do not place in other categories, except for the Kraków , Lambeth Palace an d Trivulziana manuscripts, which would need further investigation.
35 Related in fact to B2 (above, n. 7).
36 This manuscript, not in Olsen, is of some interest. It is dated by Professor A. G. Watson and Dr R. M. Thomson (personal communications) to s.XII (German A.G.W., French R.M.T.). It is full of striking archaic spellings (e.g. 1.133 optume, 1.135 desinundi, 1.136 utundum, 2.80 optumates, 2.82 (a)estumandis) that go beyond what we find scattered among the primary manuscripts, and it bears the flamboyant colophon: ‘Lucius Fulvius Tullii scriba et liberatus scripsit.’ ‘Codicem plane abieci’ (Atzert4, xl).
37 Add (perhaps) the fragmentary Cambridge, Trinity College R.16.34, c. 1100, English (83) (ends 2.9), which merits examination.
38 Clark, A. C., CR 5 (1891), 365–72Google Scholar.
39 Words reminiscent of many eulogies by Graevius on his manuscript: see Popp, , De Ciceronis de officiis librorum codicibus Bernensi 104 eique cognatis (Erlangen, 1883), 6Google Scholar.
40 So rare that I have not had access to any of them. I owe my knowledge of the reading at 1.40 to the kind services of Catherine MacCormack at the University of Chicago Library, whose Anemoecius is too frail to be filmed. Other readings I have gleaned from Heusinger's edition (Brunswick, 1783) and from Popp (1883), pp. 27 seq.
41 On which see Ruf, P., Mittelalterliche Bibliothekskataloge Deutschlands und der Schweiz iii. 1 (Munich, 1932), pp. 43–58Google Scholar.
42 It was Atzert's view that the ‘DM’ seen in the margin of L at this point stands for dolus malus (ed. 4, xxiv). Rather, dignwn memoria (The Bodleian Library Record 3 (1950–1951), 121)Google Scholar.
43 Anemoecius could not have been using c, which (apart from anything else) contains 1.40 only in part.
44 I remark especially 2.29,26 P. ] p. c: 1. ξ; 37,11 maioris partis animos c: maiores partes animi ξ;45,30 atqueadc, Non. .et ad ξ; 49,8 et apud populum c: om. ξ 50,15–16 ut hii quosanteadixi out ulciscendigratia c: om. ξ 50,17 AT. Aquilio ] aquilio c: manilio ξ. I have gone further than most editors in accepting readings of c in this passage.
45 After sampling 2.40–6 I think that the lost folios could be reconstructed (if we needed to) from MSV! (for which see below). Though showing Vulgate errors (especially 43,18 boni), the three are united by many errors that I do not find in representative Vulgate MSS (40,9 is ] idem; 43,12 retur; 43,13 errat; 45,30 firmiore eo), while, thanks to the known separate contamination in S and v1 we find agreements of MS against V1ξc (44,24 que) and especially MV1 against Sξc (41,24 constituendam; 41,25 obtinebant; 42,6 non pecunie; 44,21 om. ipso; 44,22 9111 ] quales; 44,27 versatur; 44,27 esse eius; 46,12 ratione (h)ac ingenio).
46 Ruf, , Mitt. Bibl., p. 54Google Scholar; Clark (1891), 370.
47 Cf. Graevius' words on the subject: ‘sed multum quoque debeo membranis pervetustis, quanquam laceris et mancis, quas aere suo comparatas mihi misit frater optimus et carissimus, Godefridus Graevius’ (cited by Popp (1883), p. 3; see his succeeding discussion).
48 Allgemeine Deutsche Biographie 30 (Leipzig, 1890), 661–2Google Scholar.
49 Ruf, , Mitt. Bibl., p. 49Google Scholar. He wrote a number of other books in the Munich collection (including a De Officiis that is now Munich 364), some dated from Nuremberg from 1463 to 1504.
50 Sometimes the Vulgate reading is so relegated: thus 3.10,21 veneris, where L had inchoaret.
51 There is a certain worry about 2.77,14, where for taetrius MS agree on deterius, though S gives the correct reading in the margin. But (see MD loc. cit., 138 n. 13) L has an erasure here, with tetrius in the margin.
52 Unless he was capable of producing a contaminated copy in one step. For divergent treatment of corrections and variants in L, note 1.38,19 praeclara ξ LM: praeclara sententia est L p.c.S; 1.38,25 virtuti LS: vitae L var. MS var. It may be worth adding here that the corrections in L adduced in my edition fall into two categories, though all may be taken from the same (Vulgate) source: (a) corrections from ξ readings to ξ readings (1.92,6 turn; 1.93,13 omnisque; 1.93,14 modus; 1.142,3 hone; 2.52,15 indigentibus; 2.72,26 civibus; 2.72,26 reipublicae) and (b) importation of Vulgate readings (1.38,19 (see above); 1.49,28 modo; 1.52,8 ipsi luceat; 1.76,14 dilatatum (this is sheer conjecture; the archetype omitted any participle); 1.82,9 et2; del. (a conjecture shared by c (see below, n. 70) and conditioned by the following, itself conjectured, cogitationibus); 2.11,4 inanimata; 2.73,2 diminutio). On the different correctors of L, which I do not distinguish, see Atzert, , De Ciceronis librorum de officiis quibusdam codicibus. I. De codice Harleiano 2716 (Osnabriick, 1914), pp. 7–20Google Scholar.
53 In my edition I use Me alone to reconstruct ξ where L is absent (except of course for the central part of Book 2). But where I record divergences of M and c, I show to which the support of S goes.
54 That it is formed thus, rather than by the infection of a Vulgate text from L, is strongly suggested by its behaviour in 2.25–51 (above, n. 45).
55 So much so that it is not quite absurd to wonder, as did Schiche, T. (WKlPh 4 (1887), 17)Google Scholar, whether p is the result of a ξ text being overlaid by ξ readings.
56 See below, p. 228.1 do not count 1.115,8, where I have preferred the word order given by Lp; here c and ξ do not agree with each other, but provide two other orders. Compare the three- way split Mp/ξ/c at 3.15,3 sunt.
57 Popp (1883), p. 10; cf. Popp, , De Ciceronis de officiis librorum codice Palatino (Erlangen, 1886), p. 16Google Scholar.
58 Not the intercalated leaves discussed above, p. 223 with n. 45. It may be remarked that p's last ξ reading before L's lacuna starts is as close as 2.24,8–9, and the first after it finishes as close as 2.51,8 (cf. also line 3).
59 A lacuna in p explicable from some special feature of L would be decisive. At 2.64,19–20 p (before correction) omits not (as one might expect: homoeoteleuton!) et…liceat but et…abhorrentem. The est that follows is in L below and a little to the right of et; that could have increased the possibility of a leap from et to est.
60 p has a Ψ count (above, p. 220) of 2, thanks to this contamination. Ψ's connection with P (above, p. 222) explains the agreements of p with P type manuscripts noted by Popp (1886), p. 18.
61 A point I stress in view of the assumption in Fedeli, , Studi, pp. 33–5Google Scholar (in a discussion of T) that agreements with p are ipso facto signs of contamination from f.
62 In his Budé, i.72. Not German , according to Dr Bruce Barker-Benfield, who very kindly examined a microfilm for me.
63 S.XIII in. according to Dr Barker-Benfield. P. Petitmengin thought of S. XIII2– XIV1 (but perhaps only for the pro Milone: Texts and Transmission, p. 82 n. 167).
64 Popp (1883), pp. 11–18. See also Atzert4, pp. xii–xv.
65 I print six readings on the authority of cp: 1.38,25 quorum (also B p.c); 1.43,11 ab; 1.76,15 disciplinae; 2.37,16 om. est; 3.17,14 nec; 3.19,6 esse. Most if not all of these can be also found in, and no doubt came from, the Vulgate. They are no less conjectural than c's successes at 1.50,9 natura and 3.57,15 rhodius. See also n. 68.
66 It will be clear from my discussion why I disregard them in my apparatus. Such contamination could of course result in cp joining ξ to leave L in sole possession of the truth. In fact this seems rarely to happen (I only print the following on L's (or M's) sole authority: 1.37,9 effecit; 1.38,14 cum cive aliter contendimus; 1.54,22 domus (hardly significant); 1.128,15 sint; 3.112,15 T. Note also 3.90,1 sibine uterque ] sibi neuterque M solus). Of course, we cannot know which of L's individual errors in fact go back to ξ, having been ironed out in cp. In any event, I do not trouble my apparatus with them.
67 Note also three cases of inanimatus for inanimus at 12,14; 14,1 and 19,27; common enough in Vulgate texts; and 77,14 digressa (a banalization of the ξ reading). It will be noticed how many of these readings are easy errors or banalizations, liable to coincidental efflorescence. The real extent of contamination in c may be minimal (note the dismissal of cp agreements in Popp (1886), pp. 12–15).
68 See above, n. 65. So too 41,25 infims; 48,32 alliciant (both given by cpΨ, but also by P after correction).
69 There is an overlap. The older hand resumes at iudici]o tueri. I am very grateful to Mark's Stahli for help concerning the new folios.
70 That is, there is no build-up of Lp agreements against ξc. But note that at 1,82,9 occurs the agreement of L p.c. with c in the deletion of et2 mentioned at n. 52. That could be a sign of contamination in c. So could other agreements of L p.c. with c outside these folios (see 1.38,19; 1.76,14; 2.11,4).
71 I find all but the third of these readings in my usual Vulgate manuscripts.
72 I use the asterisk elsewhere to show which way p goes where there is a split between L and c (and in 2.25–51 to mark places where p ‘supports’ c).
73 Thus at 1.145,7 ‘vel ξ: aut L*‘; contrast at 1.147,1–2 ‘qua de L* (+c): de qua ξ’.
74 For two others see below, p. 231 and n. 93.
75 c cannot be right, however, for semper autem… does not cohere with 1.39.
76 Not counting books already discussed, or the older Vienna 315 (above, p. 225). Apart from the Lincoln College book* they are, in abbreviated description (all from the fifteenth century unless otherwise indicated): Amherst 14; Berkeley 62*; Florence, Strozzi 36, s.XIII? (mg. in similar hand); Florence, Riccardiana 564*, 577; Ithaca 564.A4*; Kraków 518; Naples IV.G.8 (mg., different hand); New York, Plimpton 99, s.XIV ex. (mg.)*; Princeton 68; Ravenna 107; Seville 5–5–19 (mg., but same hand; there are other intriguing ξ indications, and the manuscript would repay further investigation)*; Trento 1781 (mg., later hand); Vatican, Ottob. lat. 1781 (mg.). I do not know which version is found in Skokloster 8° 13. For the use of the asterisk, see below, p. 231.
77 And those I mention are not found in Avranches 225 or in Oxford, New College 251 (see below p. 231).
78 It is true that there are also stray (chance?) agreements with c against L, notably 2.6,26 om. tantum; 36, 3 <esse> putant; 53,27 mala. The many agreements with p are a sign of influence not from f but from the Vulgate (above, n. 61). The Vulgate count is 4 (plus one p.c).
79 It is, however, worrying that at least in the Lincoln College book nostris is present in 1.40,15, though it is omitted by L and its descendants, M, S and V1 Perhaps it was added by conjecture, or introduced by contamination; or perhaps we should postulate a gemellus of L (see also below, n. 88). What do the other manuscripts listed in n. 76 do?
80 Studi, pp. 19–39. Fedeli's study at once underplays the extent of f influence on these manuscripts by recording only ξ errors, and overplays it by regarding coincidences with p as evidence of f rather than Vulgate affiliation (above, n. 61); thus his evidence in effect shows no ξ influence on Vatican, Borg. lat. 326 (where, it will be recalled, 1.40 is only added in the margin). As for c (as opposed to Lc) readings, Fedeli claims none for Paris, lat. 6342 or Avranches 225, and does not press his data for Borg. lat. 326. As for Troyes 552, he does not know that many of his examples, often taken from Book 3, are in fact also found in M and so derive from ξ.
81 It should be mentioned that in the first part of Book 1, T is hardly more than an ordinary Vulgate manuscript, despite the presence of 1.40. It gets closer to Avranches 225 later.
82 In particular, Prof. R. M. Thomson points out to me that the words added by William of 1 Malmesbury to the Rawlinson manuscript at 2.43 to fill a lacuna reappear in the text of the Poppi I book; for this intervention of William's see Prof. Thomson's remarks in de la Mare, A. C. and Barker-Benfield, B. C. (eds.), Manuscripts at Oxford: R. W. Hunt Memorial Exhibition (Bodleian Library, Oxford, 1980), p. 28Google Scholar.
83 Written for William Gray, perhaps in Cologne, by the Dutch scribe T. Werken (see Mynors, R. A. B., Transactions of the Cambridge Bibliographical Society 2 (1950), 97–104)Google Scholar.
84 At least it does not show up on my Vulgate count, and ξ's tempting lacuna at 2.89,4 remains unfilled. The agreements with cp are trivial, those with p sometimes more striking; 31,18 putant; 39,3 om. eam; 46,14 <tum>in suos; 54,5 constat; 84,21 ne <non>. See further, Fedeli, , Studi, p. 30Google Scholar.
85 The source then would seem not to be P; and it cannot be a, where there is no capitalization at all.
86 The Ψ texts make a fist at filling this gap, but read reddiderit or the like.
87 Of these readings the Lincoln College manuscript gives those at 2.4,3; 69,18–19; 84, 26–7 (a.c).
88 L's seems to be correct: the position of eum looks idiomatic, and for senatus et C. Fabricius consul cf. 3.86 ‘a C. Fabricio consule iterum et a senatu nostro’ (also 87 ‘vel Fabricio…vel senatui nostro’) and esp. Polybius 18.46.5 (cited by Thomas, K.B., Textkritische Untersuchungen zu Ciceros Schrift De Officiis (Munster, 1971), p. 73 n. 250Google Scholar) ἠ σ⋯γκλητος ἠ ‘Pωμα⋯ων κα⋯ T⋯τος Kοἵντιος στρατηγ⋯ς ὔπατος. The c version might seem to arise from a misreading of some abbreviation for consul as eū, with consequent omission of the first eum. The third version would be parasitic on that, perfugam glossing eum. But we could instead assume, e.g., that all three versions arose independently from a fourth in which no object was present. In that case it would seem that the ξ readings in the Avranches book (supposing always that they are from the same source from which it got its version of 1.40) do not come from L itself, but from a close relative. Cf. also above, n. 79.
89 c itself is younger than Av, and does not carry 1.40 complete.
90 Geography is also relevant. If L was always in the vicinity of Augsburg, one would think of a gemellus nearer to France and England. Cologne might be a different matter.
91 The Vulgate count is very high (14).
92 Some of the variants, like the quite common omission of tum, might suggest a conjectural restoration.
93 The ξ reading ianum at 2.87,21 is found very rarely, and may, at least sometimes, be the result of learned conjecture. I have noted it in (apart from M, S mg. and VJ Florence, plut. 76.19 (mg.), Conv. Soppr. 529 (iani); Florence, Magi. VI.208 (a.c); Padua, Antoniana 50 (iani); Seville 5–5–19 (see above, n. 76, for this manuscript); Vatican, lat. 11465; and Wolfenbuttel, Gud. lat. 2. Conversely, the ξ omission of quid tertium male pascere at 2.89,4 is occasionally found, not only in the old ξ witnesses and in the early members of the Avranches group, but in a few twelfth-century books (Gotha, Membr. 11.198 before correction – to q.t. bene vestire as in all cases in this list); Leeds 21 (a.c); Milan, Ambros. C.29 inf. (a.c); Paris, lat. 13340 and 18419), and in the following later texts: Berlin, Hamilton 159/1 (a.c); Bloomington 219; Budapest 142; Erlangen 618; Escorial S.1.18; Florence, Riccardiana 567; Leiden, Periz.F.25; London, Burney 152; Madrid 9502; Oxford, Laud. Lat. 62; Paris, lat. 6357; Paris, Ste Genevieve 2394; Prague 2369 (a.c); Vatican, Ottob. lat. 2988 and Pal. lat. 1525; and Wroclaw R.436. Such an omission could be accidental, but it could also be a sign of a late survivor from the pre-Vulgate strain of ξ;. The correct reading is not found here except in M and S; but in Oxford, Canon. Class. Lat. 252 we find ‘quid tertium male pascere bene vestire’ (it is sad that this manuscript seems to have no other virtues).
94 1.40, 10 quod]qui; 14 om. est.
95 1.40,18–19 cum scelere tnteritum (transposition).
96 Hence my difficulties about priority between Langius and Lambinus' edition of 1565.
97 The Parcensis (from the Abbaye du Pare, near Louvain), then, was not c. Nor, to judge from other readings given by Langius, was it of any particular merit. In 1.40, Langius seems to have been contaminating the Hervagiana with his Parcensis; at 1.40,10 he reads qui, but notes: ‘liber script, quod peierassent.’ The Parcensis therefore was not T either (see n. 94). It, or (better) Langius himself, was apparently responsible for the conjecture dedidit in line 18.
98 The reference is of course to Anemoecius (not that his version of 1.40 was the same).
99 Not, anyway, at the Cistercian Abbey of Val-Saint-Lambert. For even if a book there could count as ‘Gallicus’, the readings Langius gives from it are not in c (though he says it did have toga at 2.66,20: perhaps he is confusing it with c for this reading alone?).
100 At 3,3–4 se appears instead of iam in M, in front of iam in S, after fere in c (and O13). We should deduce, perhaps, that the reflexive is being added by conjecture, as it is so variously in the Vulgate. It is remarkable that Vt gives illos with no se (as later did Lambinus), another way of solving the problem. What the truth is, I am not sure: perhaps Mas (Facciolati). For aequare used like that, cf. Livy 27.16.7.
101 See MD, loc. cit., 140 with n. 23.1 am less confident than I was that Gallum is correct. It is noteworthy that Vt reads Sulpitium G. for C. Sulpicium, without adding Gallum. Taken together with the fact that Florence, plut. 76.20 has Gallum as a gloss over G., this may suggest how Gallum arrived in the text (it is also found in another Vulgate text, Paris, lat. 2335, before correction). The coincidence, C. Sulpicius being a Galus, would be extravagant. The equally unattractive alternative is that one or more medieval scholars identified this not altogether prominent person.
102 I do not know what Cicero wrote here, plus is attractive in view of Parad. Stoic. 52 ‘avari… plus semper appetunt’. e quo is immediately suspect both for its form (though it is true that MS have ex), and because of ex quo just before, aequo (Boot) governed by plus is by no means absurd. Or one might think of deleting.
103 Probably add 4,24, where et colendo (M, S and doubtless V1) stands against cψ's colendo.
104 See above, nn. 5 and 15 for f and the correctors of B and P (and also n. 52 for intrusion of Vulgate readings into L by correction).
105 Note also, for what they are worth, agreements of M with O13 (17,8–9 agendis (perhaps also V,); 27,11 om. aliquo; 28,20 expetant) to counteract the agreements of c with O13 mentioned below. If these really were L readings they have been ironed out by contamination in S and V1 (just as the latter pair has lost ξ readings at 23,4 iustitiae est; 28,20–1 gladiari). Or they may intrude from the Vulgate.
106 She in turn was much helped by the Institut de Recherche et d'Histoire des Textes in Paris, whose microfilm holdings enabled her to report on manuscripts in the places starred in this list. I was myself saved a good deal of Italian travel by the microfilm holdings of the Biblioteca Nazionale in Rome.
107 For the exceptions see the lists above, n. 76, p. 230, p. 231 and n. 93.
108 See Sabbadini, , Storia e critica, pp. 145–53 (Milan)Google Scholar; Gnesotto, A., Atti e Mem. Ace. di Padova 18 (1901–1902), 303–10 and 20 (1903–1904), 157–70, 287–303 (Venice)Google Scholar. I am most grateful to Alessandro Schiesaro for procuring me copies of these and other articles by Gnesotto.
109 I give references to the scribes listed in Colophons de manuscrits Occidentaux des origines au XVIe siècle (Fribourg, 1965–1979)Google Scholar under the form B followed by the number(s) in that collection.
