¹ Cf. L.D., Reynolds (ed.), Texts and Transmission: a Survey of the Latin Classics (Oxford, 1983), pp. 102–9; for the mutili, esp. pp. 102–4. Some additions and corrections to this account, and the stemma underlying it, were later given by M. D. Reeve, ‘The Circulation of Classical Works on Rhetoric from the 12th to the 14th Century’ in Cl. Leonardi/E. Menestò (eds.), Retorica e poetica tra i secoli XII e XIV (Atti del secondo Convegno internazionale di studi dell'Associazione per il Medioevo e l'Umanesimo Latini (AMUL); Trento e Rovereto, 1985), [Florence, 1988], pp. 109–24 (for De oratore, see pp. 118–24). I will cite the following work by the author's name only: E. Ströbel, De Ciceronis de oratore librorum codicibus mutilis antiquioribus (diss. inaug., Erlangen 1883;Google Scholaralso in Acta seminarii philologici Erlangensis, 3 [1884], 1–74, which I have used). The edition of De oratore I have used is K. F. Kumaniecki's (Leipzig, 1969). Occasionally the previous Teubner edition by W. Friedrich (1891 etc.) is referred to.Google Scholar On derivation in general, and the means of establishing it, see Reeve, M.D., ‘Eliminatio codicum descriptorum: a methodological problem’, in J.N., Grant (ed.), Editing Greek and Latin Texts (New York, 1989), pp. 1–35.Google Scholar

² P. 48. The reader will find it reproduced in A. S. Wilkins' OCT edition (1902 etc.), p. iv; also in the Introduction to his Commentary (Oxford, 1892; repr. Hildesheim, 1965, 1990), p. 66.

³ See Kumaniecki, p. vii; the only change he has made is the addition of K (see below), which was still unknown to Ströbel. In Texts and Transmission Ströbel's/Kumaniecki's stemma is accepted (see e.g. p. 105: ‘E and K descend from a sister of A’): the only point on which they voice some doubt (see p. 103, n. 2) concerns the intermediary unnecessarily introduced between M and H by Ströbel. Cf. Reeve, ‘The Circulation…’ (above, n. 1), pp. 119–21.

⁴ Later, Reeve (op. cit., p. 122) saw that it is in fact independent of E.

⁵ As Book 1 is lacking in A, all passages adduced in this article are from Books 2 and 3. In citing them I add line numbers (Kumaniecki's edition) to the book and section numbers. All manuscript readings given have been checked with the aid of photographs (the many discrepancies with Kumaniecki's app. crit.—which is highly inaccurate—are only seldom made explicit); F I have collated in situ; in a later stage I have also seen A in situ and re-checked many of its readings. Manuscript abbreviations are mostly resolved without notice; sometimes, when it seemed to serve some purpose, line end has been indicated (/). I use the abbreviation p.c. (or ^pc) for corrections evidently made by the original scribe in the course of copying; m.p. (or ^mp) denotes a later corrector on whose identity I cannot be more specific.

⁶ It is noteworthy that the words concerned were evidently not missing in K's exemplar, as K's §428 starts with Haecfere maxime etc. (2.206,5).

⁷ A rather thoughtless emendation of a type that occurs more often in E: out of an impossible form it makes an existing Latin word, without bothering too much about the sense. As a result of all this E reads: si aut contra quicumque illi erunt etc.

⁸ On top of that, in F the words quae (codd.: idque Kuman.) simul atque (2.214,11) are also omitted. This is probably a case of deliberate ‘emendation’. F reads: Item misericordia aut inuidiam simulatque emissa est adherescit (cf. Kuman. ad 9.10; 12). Note that in isolation F's text could be the result of one single jump (from the first simul atque to the second).

⁹ There is yet another case of words omitted by E corresponding exactly to one line in A: 2.102,6–7 habeat—plus om. E¹ (A, f. 8^r, the last line (30) of the page. But here I do not consider the mistake to have arisen in copying from A, because the words concerned are not missing from F, while in E they are supplied, not by E³, but by an older corrector (E²), who emended many errors of E throughout the three books of De oratore, and who must have used a descendant of A closely related to EF for his exemplar (perhaps a, or else the copy of a from which F descends). Besides, the bottom line of a page will not, I think, easily be skipped by a copyist. So the mistake is better explained as one of E's own copying-errors, caused by the two occurrences of plus. For the source of E³, see below, n. 10.

¹⁰ The omission in E¹F concerns 2.38,10–11 suum munus—nonpotest. Note that the words are not missing in K (they form part of It's §422, which runs from 2.35,2 to 2.38,14); we have seen a similar case at 2.205–6,3–6: above, with n. 6. E³ has his supplements (and corrections) either from a manuscript very close to H (cf. Ströbel, p. 46), or—more probably—from H itself. Like E² (above, n. 9), he made numerous corrections in E. Moreover, he must be responsible for E's cryphias (which correspond continually to either a cryphia or a vacant space in H). He seems to have given up after he had finished 2.233. Note that Ströbel and Kumaniecki fail to distinguish E² and E³: to them all the corrections made by these hands are by one and the same corrector (‘E²’).

¹¹ In A, per- is not very legible, but autopsy (above, n. 5) has convinced me that this is in fact the correct reading (not con-, as not only Ströbel but also Friedrich and Kumaniecki have it).

¹² A² correctly reads p(er)tineret, anyway.

¹³ Here, of AEH, only H has the ‘obviously correct reading’. The readings of EF and m look like attempts to emend the senseless solut. Thus, on the strength of Ströbel's stemma, it is uncertain—to say the least—whether solut is indeed one of A's peculiar errors. Cf. 2.128,18 peta A: petam Em: peto FHL, edd.; 2.197,12 causu A: causa E'F: casu HE³L, edd.

¹⁴ In line 20 F's alisne is p.c. (aliisne F^ac).

¹⁵ Unlike H, A nearly always writes quom in full, probably to forestall confusion with quō quoniam: see Lindsay, W.M, Notae Latinae (Cambridge, 1915; repr. Hildesheim, 1963), s.v. quoniam, esp. §§328, 330.Google Scholar

¹⁶ Unlike EH, A does not generally abbreviate -us, probably to forestall confusion with -ur: in some pre-Caroline scripts, and also in early Caroline minuscule, the apostrophe was not seldom used for both -us and -ur. See Notae Latinae (above, n. 15), s.w., esp. §§468, 470–1, 474–5

¹⁷ E.g. at 2.72,20 in E sci- is p.c. (si- E^ac), but hau/haud is the relevant point; at 2.111,14 EFH^mp read hi instead of ii (H¹), but in the Middle Ages this is a question of spelling; 2.196,23 lacrumis E; 2.327,1 uirs- H¹ (uers- Hmp); at 3.190,2 H has st instead of est (with aphaeresis, as in A), but here the final -m is the point concerned; likewise, 3.219,1 est EF: st H^pc (om. H^ac).

¹⁸ To be more precise: the parts of Books 2 and 3 present in A, minus those written by A²; all in all about 400 sections.

¹⁹ One cannot argue that he systematically corrected himself: there are relatively few places in A (much fewer, for example, than in H) where the original scribe corrects errors of his own.

²⁰ By way of comparison: I estimate that in the same parts of the text H—by no means a careless copy—has about 150 to 170 peculiar errors, about 30 to 35 of which concern wordomission (e.g. 2.84,18 sentio; 2.119,11 dumtaxat; 2.128,20 modo; 2.219,15 persaepe; 2.305,15 dicas; 2.312,7 animos; 2.327,10–11 narrantur; 3.139,23 doctissimum; 3.202,11 quam dixeris; 3.227,10 quiddam).

²¹ In many places the text of E shows signs of this striving for emendation. Success in at least some of these places need not surprise us. Ströbel (p. 44) gives a few (of the many) examples of failure; cf. also my notes 7, 13 and 53.

²² E.g. 2.24,9–10. Note that here (and elsewhere) E probably owes its emendation to the activity of a corrector in A; I will return to this later.

²³ A mark used to indicate a corruption (cf. Isid. Etym. 1.21.10). For the cryphias in H,see Beeson, Ch. H., Lupus of Ferrières as Scribe and Text Critic: a Study of his Autograph Copy of Cicero's De oratore (Cambridge, MA, 1930), p. 27.Google Scholar

²⁴ Erased by a later hand. Cf. Beeson (above, n. 23), pp. 34–6. By some slip, he lists the present case under H's marginal corrections (see p. 37, s.v. ‘82^V.2.15’), rather than under the marginal variants, [uoluisti E¹: -s add. E², s.l. (prave Kuman.)]

²⁵ At the same time he begins to abbreviate superlative forms in -mus in this way: e.g. prim; (3.7,7and 3.137,17), optim;(3.135,20),grauissim; (3.177,12—gravessumusedd —and 3.209,14). The same goes for eius: from 2.366,13 onwards this is mostly written ei;. Hui; occurs at 3.121,7 and 3.227,13, cui; at 3.132,13 (Cous edd.). All these forms are correctly transmitted in E, except for 3.7,8 ei; A: ei EF. To be sure, I found an isolated case oiei; as early as 2.46,16 (et EF) and one of cui; at 2.233,12 (cuf EF), both well before the turning-point near the end of Book 2. A similar change of habit occurs in H, though a book earlier. Lupus starts abbreviating -mus to -m', but changes to -m; at 1.98,15 (quaerim;). From then on forms like grauissim;, ei; etc. are also frequent in H (hui; already at 1.96,4). The ‘semicolon’ normally looks in H more like a ‘colon’, but that makes no real difference.

²⁶ For the various means of abbreviating -us, see Notae Latinae (above, n. 15), s.v., esp. §§475–6; for the one with ‘semicolon’, esp. §§476 I, 477.

²⁷ Apart from an isolated case as early as 2.80,8 faciam; A: faciam' E (faciamus F). The preceding word is nobis, which may have prevented a copying-error, or else may have inspired emendation.

²⁸ The two instances where they commit no error are 362,6 (ignoscim; A) and 362,8 (adnosctm; A). Note that reading ignoscim and adnoscim would produce impossible Latin, and that the context offers ample support for finding the right interpretation.

²⁹ The remaining five are: 9,23; 9,3; 13,12 (pertulim;); 15,4; 16,8. EF have all five correctly (9,23 suscipimus F). Note that reading suscepim, tenem, pertulim, postulam, would produce impossible Latin. As to interfuissem; (16,8), here the immediate context would probably force even the dimmest of scribes to read a plural.

³⁰ 3.17–110 are lacking in AEF, as indeed in all muttii.

³¹ Perhaps it was by a mere slip that the scribe of E forgot to add the apostrophe here, [si quaerim; A (prave Kuman.)]

³² Among the remaining ten, three instances of sum; (122,14; 123,8; 147,16) are preceded by nominative plural forms; viderim; (123,11) was perhaps saved by teramus, one line earlier. In the other cases (123,13–14; 126,13; 133,1; 137,18; 138,1; 138,10) neglecting thE‘semicolon’would have produced impossible Latin.

³³ 3.149–71 are omitted by A¹EF.

³⁴ I.e. the scribe of the lost intermediary between A and EF (a; see above, the introduction).

³⁵ Among the remaining 27, there are cases in which the immediate context offers little or no support for the correct interpretation of these forms (e.g. 181,13; 182,4; 185,19–20; 186,8; 186,9; 201,1; 210,18). As Professor Reeve (below, n. 64) points out to me, one could, if necessary, suppose that this section of a was written by a different scribe.

³⁶ Among the others are: 2.193,7 pater/num A: pater num E (patemum FH); 2.346,7 dici or/natissime A (with only a slight space between dici and or-): ditior natissime E¹ (did ornatissime Em pFH); 3.14,22 ii A (looking like n): n. E: om. F (ii H 1: hi H m p: ei edd.); 3.127,1 mane- Aa c: manu Ap c (the c made into a u, while copying): manci E¹ {manu FH). I have noted a few more.

³⁷ The -t of E's uolet is in my opinion added by a later hand, since (a) it looks different from the way E ¹ writes -t, (b) E ¹ always writes the verbal ending -el in ligature (-&), a n d (c) as it is now, there is n o space between uolet and t he following iudici, which is contrary to E's normal practice.

³⁸ Or perhaps I should say: nearly wrote, as one could quarrel about whether the b was ever finished. After seeing A itself I think it was, but I am not quite sure.

³⁹ Kumaniecki is wrong in attributing the correction to A²: considering the shape of the p involved and the hue of its ink, I am convinced that it was made by the original scribe.

⁴⁰ I infer this from the fact that elsewhere A always writes -us without ligature (hundreds of cases).

⁴¹ If the latter explanation is right, the corrections must indeed have been made by the original scribe, as there is no difference in ink.

⁴² See the reproduction of one of A's pages (f. 37^r) in Chatelain, É., Paléographie des classiques latins, 1 (Paris, 1894), PI. XIX. 1.Google Scholar

⁴³ Ströbel speaks of it as a manuscript ‘qui iam non tanta fide quanta A archetypi scripturas servaverat’ (p. 47).

⁴⁴ Cf. above, section B (Table 1). [2.24,9 scaeuola M (prave Kuman.)]

⁴⁵ As may be illustrated by the following examples. 2.164,8 (est²:) et A: est A²; 2.209,15 non gessit A: ‘(ue)l longe sit’ ss. A²; 2.231,11 si A: sit A²; 2.233,15 se A: esse A²; 3.113,5–6 mente in A (cryph. adh.): mentiri A² in mg.; 3.116,7 uerberibus A: -ue rebus A² in mg.; 3.121,10 pecus A (cryph. adh.): pectus A² in mg.; 3.132,13 cuius A: ‘(ue)l chous’ adscr. A² in mg.; 3.134,11 repereretur A: referretur A² in mg. In all these instances EF still have the corruption of A¹ in their text (‘emended’ at 3.121,10 penus F; 3.134,11 repperetur E: repeteretur F). For some exceptions, see below (with n. 60).

⁴⁶ 1.157–93, 2.13–18 (?), 2.234–87 and 3.149–71; now partly lost again, because of the mutilation of A. In EF, as in A, these passages were originally missing (they were c. 1470 supplemented in E, but F is still in its original state). Cf. Texts and Transmission, p. 103.

⁴⁷ Note that Ströbel did not see A itself: he had to rely on a collation made by Heerdegen in 1881 (see his statement to that effect on p. 2).

⁴⁸ See Chatelain (above, n. 42), PI. XIX.2.

⁴⁹ I have supplemented Ströbel's data with a number of cases I have found myself. For the sake of clarity cryphias, vacant spaces, and some irrelevant peculiarities are passed over in silence. Unless otherwise stated, the sigla ‘A¹’ and ‘H¹’ imply that a later hand has correctly repaired the error concerned. ‘E¹’ implies that a later hand (probably E², in most cases) ‘corrected’ the correct reading of E to the corrupt one of AH. Not listed are a few doubtful cases, notably 3.174,18 sum AH: suum F (E's reading uncertain, but perhaps sunt E¹: sum E²) and 3.219,5 colicandas AH: colligandas EF: coligandas edd. recc. (collocandas L, edd. vett., fort, recte).

⁵⁰ That E reads fletus instead offletum seems immaterial. Considering the testimony of F, the plural in E might be due to some mistake in copying.

⁵¹ How E's reading should be explained is another matter. Perhaps simply as a lucky guess. I will return to this when discussing the possibility of E's text being contaminated.

⁵² E actually reads uniuscu(us)q(ue), but the omission of the i seems only a slip of the pen.

⁵³ Cf. 2.193,1 quic AH¹: quicquid E¹: quid edd. (qui F), to illustrate that this method of emending may fail as well. Cf. above, n. 21.

⁵⁴ The same may hold true for, e.g., 2.357,5 (there are three words ending in s in the immediate context) and 3.190,17 (uobis and nobis are easily confused in the manuscripts).

⁵⁵ E.g. 2.39,17 (where he started writing n-, but immediately corrected to m).

⁵⁶ -mus for -mur is an old error in the M-tradition, emended in H but at the same time cautiously preserved in the margin as a ‘variant’ we have seen a similar case at 2.363,24 (above, with n. 24). -amur for -mur may be a copying-error of Lupus's.

⁵⁷ On contamination in descripti, see Reeve, ‘Eliminatio…’ (above, n. 1), pp. 23–5.

⁵⁸ Reeve, loc. cit. (p. 24, with n. 73).

⁵⁹ All those in which the siglum ‘A¹’ occurs.

⁶⁰ 2.163,4, 2.327,1, 3.119,13 and 3.136,23–4. Note that in these four instances the errors involved are easily emendable, and may well have been emended more than once independently. In total, I have found only some eighty corrections made by later hands in A. More than three quarters of these can be assigned to A², either certainly or probably. As we have seen above (with n. 45), they are as a rule ignored by E. The majority of those that are not by A² are adopted by E, and are listed in Tables 1 and 2. I have already discussed one of these (the addition at 2.24,9–10: above, with n. 44), the others are mentioned in what follows. The corrections at 2.99,20, 2.101,22 and 2.101,24 (possibly also 2.96,20) are by one and the same corrector, to be distinguished not only from A² but also from the hand active at 2.24,9–10. He writes a tiny but clear and precise Caroline minuscule in a fairly dark brown ink. It is likely that E owes its emendations to this hand, as seems to be borne out, paradoxically, by the one case in which it does not follow its lead: 2.101,24 tradi ciatis A¹ (tarditatis ss. A^mp): traditiatis E¹: tarditatis F (traditis E^mp: tradidatis H¹: tradi datis H^mp). In this case the corruption of A¹ is not deleted by our corrector, which may account for the divergence of E from F, as it may have induced a to copy both the corruption and the ‘variant’. Something similar may have happened at 2.159,7: here too the corruption of A¹ (ruina) is not deleted by the corrector (trutina simply being written above it), which may be the reason why we still find it in F, whereas E saw that the ‘variant’ trutina should be correct. I do not know to which hand I should attribute this correction, or the ones at 2.25,4 (perhaps A¹), 2.89,17–18 and 2.90,1 (perhaps the hand active in 2.99–101, discussed just now), and 2.146,25; but as far as I can tell, none of these is by A².

⁶¹ I give them here in more detail than I did above. The selection is of course somewhat subjective, but not, I hope, unreasonable.

⁶² Though not necessarily, of course. The corrections concerned may also be explained as deliberate conjectures.

⁶³ Apart from one or two remarkable readings.

⁶⁴ I am grateful to my colleagues in Amsterdam (to single out only a few: Se Lenssen, Rodie Risselada, Jaap Wisse) for discussing the contents of this article with me; also to Professor M. D. Reeve (Cambridge), who very kindly read an earlier typescript and returned it to me with numerous valuable annotations and corrections.