¹ ‘Problems in Quintilian’, BICS Supplement 25 (1970)Google Scholar, henceforth cited as ‘Problems’. I also cite occasionally, as Watt, Watt, W. S., ‘Notes on Quintilian’, GB 15 (1988), 139–60.Google Scholar

² The situation which I describe applies to 5.14.12–8.3.64 and 8.6.16–67 and 9.3.2–9.4.135 and 12.11.22–31: therefore to most of 6–9 and parts of 5 and 12. Within 6–9, both AB are extant in 8.3.64–8.6.17 and 8.6.67–9.3.2, and neither A nor B is extant in 9.4.135–47 (the end of 9), where we are dependent on G (described below).

³ ‘A Problem in the Transmission of Quintilian's Institutio Oratorio’, CP 75 (1980), 312–20Google Scholar. This article is henceforth cited as ‘Transmission’.

⁴ In order, they were omissions of sed, et, sane, quam, ut aiunt, ex, et, nescire, Crescit oratio, in, in, in, nobilis et, ero, partem, admirationem, et recta, sententiae (the last three because of saut du même au même), et, et, non, ut, est, nos, pro, mira, nam, curiae.

⁵ See 8.4.1 (twenty letters omitted), 8.4.3 (thirty-one), 8.4.4 (four lines), 8.4.6 (one and one half lines), 8.4.7 (one line), 8.5.5–6 (five lines), 8.5.10 (fifteen letters).

⁶ See plate 174 in Chatelain, E., Paléographie des Classiques Latins (Paris, 1894–1900)Google Scholar, which shows thirty-three lines of two-column text, with lines in a column containing usually between twenty-two and thirty-two letters. According to Chatelain on this plate, folios 1–80 of A contain a two-column format, written by a single hand, and terminating at 4.3.6 fides. The rest of A (including the section with which we are concerned) is written by hands different from the first section in long lines across the page, usually thirty-one to thirty-three lines per page, but sometimes up to thirty-nine (see Chatelain, plate 175). The evidence suggests that the second part of A descends from a MS. which looked much like the first part of A. Since B is also written in a two-column format, the chances are good that even their common archetype had such a format. Therefore omissions which I below, in short-hand form, attribute to ‘A's exemplar’ are actually the product of a whole line of transmission, probably antedating the common source of AG (since the omissions and errors are shared by both codices).

⁷ Butler's Loeb (1921) and Cousin's Budé (1977) both print Propinquus, but share with Halm and Winterbottom a misinterpretation of the jokes as forms of insult. Cousin prints Contumeliis, obicienti aspere, ‘‹Men ex te› metiris?’, and legatario. Butler follows Halm in printing Contumeliis, obiicienti atrociora, ‘Me ex te metiris,’ and legato.

⁸ Both Hispo (Romanius Hispo of Tacitus Ann. 1.74; on the text there, see my ‘Tacitus Auctus’, CP 79 [1984], 314–26, pp. 319f.)Google Scholar and Fulvius are introduced as names we should recognize; therefore, though occasionally famous orators found themselves as defendants in prosecution or litigation, both, especially Fulvius, are more likely to be here patroni representing clients: for that is how most orators would have won their fame, and where they would have given most of their speeches. To put it another way, Quintilian identifies the speakers of jests either genetically, or by their proper names (with generic designation restricted to their opponent). Since he has identified Fulvius by name, and not as an anonymous ‘relative’, we should expect propincus either to define which Fulvius is meant (which is admirably accomplished by taking it as a proper name) or to have some point for the joke. But in the latter function it is simply misdirection. The point of the joke is not that he has called a relative ‘Sir’ (domine), but the joke turns in some way on uerus est.

⁹ ‘Names and Identities in Quintilian’, Acta Classica 28 (1985), 39–46Google Scholar (henceforth cited as Syme). Fulvius Propinquus is discussed in 40f. The inscription, in what CIL describes as litteris oplimis saeculi primi secundive (hence contemporary with Quintilian), reads: ‘M. FVLVIO PROPINQVO. P. ET. M. FVLVIO. PRISCO F. A. XIX XIX IVNIA. CROCALE S. P. F. C.’ The inscriptions from Spain in CIL 2 gives seven instances of the cognomen Propinquus (plus two of Propinqua), and over sixty of the nomen Fuluius or Fuluia.

¹⁰ Outside of Suetonius and legal texts, the TLL knows it only in Tertullian.

¹¹ Actually, while Goetz prints Chirographus from the eight-century Codex Vaticanus 3321, and lists cirografum as the reading of the later codices be, Lindsay, using the same codices, prints Chirographum and lists no variants. Lindsay also suggests Tob. 1.17 as the source of the lemma.

¹² Lewis and Short take the litigator's side by telling us to understand libellus or codex. Fulvius and Quintilian would respond that no libellus or codex is at issue, but a tabula.

¹³ Note that et uera would not have been a proper response, since not tabula but cheirographus/-um is the noun understood. Fulvius' response indicates that he took cheirographus to be a misused noun, not an adjective out of consonance.

¹⁴ The active form is also noted by the TLL as a variant in Vulg. Act. 5.3. But there mentire should be just a scribal slip for mentiri, since Jerome would hardly have translated the deponent ψεύσαθαι with an active form, even if he believed it to be acceptable.

¹⁵ Butler in his Loeb and Winterbottom in his ‘Problems’ interpreted the insult in et uerus below to be an implication that his opponent's documents were fake, But that would be ‘and mine are real’. There is no equivalent for ‘mine’ in the text (the joke could as easily be against himself, ‘and a real one this time’), and Quintilian, who elsewhere supplies the information necessary to interpret a joke (as 90 militi sine gladio decurrenti and de piscibus qui cum pridie ex parte adesi et uersati postera die positi essent), would here be a poor joke-teller for omitting information essential to understand the punch line (Butler, who read legato, supposed that the legate ‘had been suspected of forgery’).

¹⁶ Cf. Quint. 7.3.2, where for publicam (which would have been abbreviated pub.) codex G has p. ūb;, and the corrector of A has (over an erasure) its expansion as uerbis.

¹⁷ Cf. Quint. 7.2.8 bibit Gertz: uiuit A; 8.6.33 et βιοῖο Heraeus: etuio eo A; Ouidius] obidius A; 9.2.37 liuii B: libro A; iuuent Liu. iubent A.

¹⁸ I refer to Wackernagel's law, in which unemphatic pronouns go into enclitic position (normally second in clause or word-group, though the presence of other enclitics or of proclitics can affect the apparent position). See Wackernagel, J., ‘Ueber ein Gesetz der indogermanischen Wortstellung’, IF 1 (1892), 333–36.Google Scholar

¹⁹ Bailey, D. R. Shackleton, ‘Notes on Quintilian’, HSCP 87 (1983), 217–40Google Scholar (henceforth cited as Shackleton Bailey). The note on et uerum is in 222.

²⁰ If there is ambiguity, I think it would be more likely to lie in the ambiguity of both domum, ‘family’ or ‘house’, and colui, ‘cherished’, ‘courted’, ‘honoured’, ‘haunted’, ‘inhabited’. Our problem is that there are too many possible ways in which a simple credo could be considered funny. Afer could take the candidate (who is probably engaged in salutatio at Afer's home, seeking his support) to mean ‘I stayed in your house forever’, when he really meant ‘I always honoured your family’. The TLL 3.1682.15 lists this passage under the meaning ‘visitare’, implying that Afer took the meaning to be ‘I constantly visited your house’. Another opening is that, where the candidate might have said colo (‘I have always honoured your family, and still do’), he actually has said colui (‘In the past, I always honoured your family’): therefore the point of Afer's credo may be that he agrees with an implied ‘no longer do’ rather than the stated ‘always did’; but cf. Ov. Pont. 2.3.73 cum uestra domus teneris mihi semper ab annis / culta sit. Coming from a persistent candidate, even a meaning ‘I courted your house forever’ would be properly deflated by an ‘I believe it’. And even with no ambiguity, an ‘I take your word for it’ (implying that the candidate's assertion is the only way that Afer would know) would work (we would probably say in English, ‘I will have to take your word for it’). It should be kept in mind that Quintilian has introduced this retort under the category of ‘blaming less than one could’, not under the category of ambiguity (amphibolia), which is covered in 6.3.46–56, or of misinterpreting someone else's words, which, as Riggsby points out, is covered in 6.3.84–7.

²¹ Riggsby compares 6.3.45 Sed acutior est illa atque uelocior in urbanitate breuitas.

²² For other examples of this sort of intrusion over distances, see my ‘Lucilius, Fragment 3 (Marx)’, TAPA 101 (1970), 379–86, p. 382Google Scholar; see also Winterbottom in BICS 21 (1974), 29Google Scholar. In 6.3.94, the two passages would be separated in A's immediate exemplar by a single page, if that exemplar contained here the same amount of text as A did in the page published as plate 174 by Chatelain (1.5.5 [11 W] tractandum erit to 1.5.12 [20 W] nam Cicero Cano-). But the mistake is probably still earlier than that (since it is shared by G). Possibly a single column of A's exemplar corresponded to a single page of an earlier majuscule codex, and the marginal variant showed through a hole or missing corner of that codex, or was copied when a scribe mistakenly turned two pages at once. The intrusion at 8.6.26 also suggests turning two pages at once, but there the folio (two pages) would have contained the equivalent of about six more lines of OCT text (which is within the range of scribal variation in the same codex). But it is usually impossible to determine how such intrusions occur; it suffices to point out that they actually do occur. Winterbottom in his review of Cousin's edition (Gnomon 52 [1980], p. 786)Google Scholar points out that codex H (a codex descriptus) has repeated at 9.2.54 a line which occurs two sections before.

²³ Ergo often introduces a question: cf. e.g. Quint. 9.2.83; TLL 5.2.764.32–66. Cicero frequently used ergo to introduce double, contrasting questions in a construction which the TLL calls argumentatio ex contrariis. This follows the pattern of Phil. 2.30 (Ergo ego sceleratus appellor a te … Me … a te honoris causa nominatur?) where the emphases are on the contrasting ego and ille. In 6.3.72 we have essentially the first part of such a double question with the second part understood: ‘Did I never win, (my opponent always win)?’ In such a construction, ergo is proclitic, increasing the emphasis on the following ego.

²⁴ Note the expanded meaning which Quintilian gives to adverbs. Here ridicule means ’by means of ridicule’, just as in 6.3.100 barbare meant ‘in language marred by barbarisms’.

²⁵ ‘What? Are you the one who thinks I am a Centaur?’ I suppose that that is barely possible, but only barely. I see no basis for an interpretation ‘you too’, or ‘you (rather than some third person)’, while ‘you (rather than I)’ seems pointless. The point should be not ‘You think I am a Centaur, while I do not’, but, ‘You think I am a Centaur, while I actually am a man riding a horse, and not responsible for what the horse does.’ So Hippocentaurum not tu should be emphatic.

²⁶ In my reading, the contrasting, emphatic words are nescioqui and Stoicos, not nescio and puto, which hardly fits the category of transtulit crimen.

²⁷ Butler translates: ‘An Epicurean and a friend of Caesar.’ But there is no ‘and’.

²⁸ In the charge of the praetor, Caesaris is the emphatic word, and so must start the word group. That is, the praetor would not say that Varius was an Epicurean friend of Caesar (as if Epicureanism rendered him sacrosanct), but that he was ‘Caesar's friend’ (therefore not to be insulted). But it is very doubtful that a descriptive Epicurio could be uttered in Quintilian's voice within a quod clause with the subjunctive (which implies that the contents represent the words of the praetor). It follows that Epicur(i)o must be part of the man's name, and not a descriptive adjective.

²⁹ Winterbottom evidently construed Epicurio as a cognomen, to judge by the listing in his index, L. Varus Epicurius. But t he form in -ius should belong to a nomen at best. Cousin (Paris, 1977) prints L. Varo Epicurio but translates ‘L. Varius Epicurius’. His ‘Varius’ must be a slip, since in the notes he claims that ‘Lucius Varus Epicurius’ was ‘parent de Quintilius’, that is, of Publius Quinctilius Varus. To be a member of Varus' immediate family, one should be a Quinctilius. Epicurus is attested as a cognomen in four inscriptions from Rome (CIL 6.16493 M. Cosconius Epicurus, 21104 = 9.4821, 6.20839 M. Iunius Epicurus, 25376 L. Rasinius Epicurus), and one from Africa (CIL 8.344 L. Caelius Epicurus), to ignore examples where it is clearly a slave's name.

³⁰ The banker Sex. Clodius Phormio who was a witness against Caecina is presumably different from the henchman of Publius Clodius, Sex. Cloelius, whose name is often miswritten or misedited as Clodius: on the spelling of the latter's name, see Bailey, D. R. Shackleton, ‘Sex. Clodius–Sex. Cloelius’, CQ N.S. 10 [1960], 41fCrossRefGoogle Scholar. Shackleton Bailey suggests that the cognomen of Cloelius was Siculus: see Cicero's Letters to Atticus, i (Cambridge, 1965), p. 376 on Alhenione.Google Scholar

³¹ See Butler's Loeb: ‘I do not know who they were who insulted him, I suppose they were Stoics.’ Butler conveniently omits a translation of et.

³² I do not know whether the sequence nescio inquit quis (with the meaning ‘someone’) is elsewhere attested, nor is the one attestation of nescio quis inquit known to me secure (Curt. 9.2.31; see below). But inquit often behaves like an enclitic, occupying the second position in a word group. It is characteristic of enclitics that they can interrupt the normal connections of words: e.g. cf. Sen. Clem. 1.9.10 ‘quo’, inquit, ‘hoc animo facis,’ where the status of both inquit and hoc as enclitics permits the interruption of quo animo. In competition with other enclitics, inquit more often falls into third position in the grouping, but it sometimes (as in the example just cited) takes precedence over other enclitics: for examples of inquit in second position (often immediately following proclitics or preceding other enclitics) see TLL 7.1.1788.7–70, to which many other examples could be added. I suppose that, in nescioquis, nescio is probably proclitic. Plautus separated nescio quis with the intensifying pol or edepol: Aul. 71 nescio pol quae illunc hominem intemperiae tenent (‘Surely some madness or other has got hold of him’): Ep. 61 nescio edepol quid [tu] timidu's, trepidas, Epidice (‘ Surely you are in a fright and flurry over something. Epidicus’). Since enclisis is a natural part of ordinary speech, it may be that such separation of nescio quis has a colloquial flavour, suited to the telling of a joke. Therefore I do not choose to conjecture a transposition to nescio qui inquit, though an easy explanation could be provided for corruption of such a reading to nescio inquit qui (saut du même au même would result in omission of inquit, which might then be reinserted by a corrector after the apparent first word of the quotation). A similar process may have occurred at Curt. 9.2.31 (a citation for which I am indebted to the services of the TLL, which checked entries of nescio quis down to the time of Apuleius): where the later MSS. have nescio quid inquit, the earliest and best codex, P, is alleged by Bardon, H. in his Budé edition (Paris, 1948)Google Scholar to have nescio quit, and its corrector nescio inquid quit (that is, nescio inquit quid). But since the inquid is presented as written in rasura, it may be that P's original reading was nescio quid inquit. Much depends on the size of the erasure: a short erasure would imply an original nescio inquit (a saut from nescio inquit quid).

³³ CQ 27 (1977), 187ff.Google Scholar

³⁴ Bauman, R. A., Impietas in Principem (Munich, 1974), pp. 29–31Google Scholar, takes the opposite tack, placing the exile in 8 because it was a time of turmoil. The difference of four years is hardly critical for our purposes, but experience has taught me to distrust Jerome when it comes to dating. The date of 8 is suspicious precisely because it is a good date to guess if you had no other evidence. Since the organization of Jerome's work was based on dates, he was not permitted not to guess when he had no information. Although it would also be possible that Jerome is right in placing the death in 32, but wrong in making the exile last twenty-five years, it is more likely that a twenty-five year exile was the sole information that Jerome had, since Jerome's usual source, Suetonius (for Cassius Severus, the lost De Oratoribus), generally does not provide specific dates in his extant works.

³⁵ PIR ¹ vol. 1, p. 165 (on Asinius Pollio).

³⁶ Quint. 10.1.22. Asprenas was consul in a.d. 6, proconsul of Africa in 14, and reappears again in a.d. 20 speaking in the Senate in Tac. Ann. 3.18.

³⁷ See Heikki, Solin, ‘Beitraege zur Kenntnis der griechischen Personnennamen in Rom’, Commentationes Humanarum Litterarum 48 (1971), especially 138–45Google Scholar for evidence of Greek cognomina among the nobility. Our evidence that the cognomen of the poet Varius was Rufus is limited to a marginal note in an early Beneventan codex, Paris. Lat. 7530, dated by its calendar and paschal tables to 779–97. No extant ancient writer mentions his cognomen, and it is conceivable that Varius, because of his Epicurean interest, could have acquired an extra cognomen Epicurus (or Epicurius), and as a L. Varius Rufus Epicurus could have passed the cognomen on with no violation of tradition.

³⁸ The notable substitutions for tamen include amens Radermacher, amentem Hunt, talem Gertz, patrem Halm, amicus Kennedy, iuuenis Gemoll. I am indebted to the researches of David Silverman for collecting these citations. The conjectures of J. M. Hunt and G. Kennedy were made in their reviews of Winterbottom, CP 67 (1972), 56–9Google Scholar and JRS 61 (1971), 309 respectively. Kennedy did not specify the position in which he would insert amicus, but offered it as an alternative to Radermacher's amens, to which he objected as not producing a climax. The position at the beginning of the sentence is an acceptable position of emphasis, but the parallel at 7.4.21 cited below suggests that Quintilian would prefer to emphasize the subjects by positioning them in front of the verb: the repetition of occidit with mulier and uxor serves to give to each repetition the weight of a colon, with each subject (including the one which has fallen out from our MSS.) starting each respective colon (the position of emphasis).

³⁹ Senex is otherwise acceptable and certainly possible, since it can describe not only an old man, but an old woman.

⁴⁰ Silverman points out that many of the occurrences in prose before the fourth century refer to a tree or plant: Sen. Ep. Mor. 86.20, Plin. N.H. 16.130 and 24.2, Columella 6.28. There are two other occurrences in Pliny, 4.89 (gens … annoso degit aeuo of the lifespan of the Hyperboreans) and 23.40 aliud minus annosum; Apuleius uses it of a goat (Met. 7.11 hircum annosum). Aug. Conf. 1.7.11 in aliquo annosiore does have the requisite sense and application to a person. Augustine (who uses also the positive and superlative in C.D. 21.11 and 8.15) here uses a true comparative, but a reading annosior or aliquis annosior in Quint. 7.1.28 would have equal chance with annosus of omission after tyrannum and would have the advantage of greater sexual ambiguity (annosus would have to be interpreted as ‘an aged person’, with masculine standing for either sex, if it refers to t he same person as mulier and uxor). See below.

⁴¹ 3.1.14, 5.6.6, 12.6.4. The first of these (eoque iam seniore) refers to Isocrates when he was 98. The use of senior as a noun is frequent, as Apuleius Met. 2.27 ille senior, Cic. Off. 1.122 aliaque sum iuuenum, alia seniorum.

⁴² The conjecture of W. Gemoll was made in his review of Meister's 1886–7 Teubner edition: Wochenschrift für klassische Philologie 37 (1887), 1138Google Scholar. It was based on comparison of 7.4.21, which Gemoll knew from Halm.

⁴³ Silverman points to the pattern of 7.1.24, where Quintilian illustrates descending from the genus to the last species by a suasoria where Numa deliberates whether he should accept the kingship if the Romans offer it. The genus is an regnandum, the species are successively an in ciuitate aliena, an Romae, an laturi sint Romani talem regem. Roma is the ciuitas aliena further specified, while talem regem gives a different type of specification applied to t he same case. In 7.1.28 the introducing clause, ut a communibus ad propria ueniamus, suggests a gradual movement from commune to increasing propria in the same case at issue. If we could be sure that such a progression was strict and consistent, we would prefer tyrannum ‹femina› occidit (a female, married woman, his wife), though that would produce different categories of propria from 7.4.21. But femina and mulier are usually mere equivalents for each other, and the progression in the suasoria of Numa (introduced by a genere ad extremam speciem descenderem, which similarly implies a progression) is strict in only two of its three parts. In a broad sense, mulier is a type of older person, since the term would not be used of a very young person (though it does not serve to distinguish between iuuenis and senex). In 7.4.21 the implied topic of the declamation would go: ‘There is a law that whoever kills a tyrant can claim a certain reward. Each of two different individuals kills a tyrant, and both claim the same reward. Which is more worthy?’ Old age, female sex, and close relationship to the tyrant all render the act more exceptional and more worthy of merit, and they are arranged in proper, climatic order of increasing merit. Since I doubt that any special merit attaches to a tyrannicide for being a mulier rather than a virgo, a progression femina, mulier, uxor, by making mulier a word of status rather than sex, fails to satisfy.

⁴⁴ E.g. sacra Riese, tantum Timpanaro (the latter printed by Shackleton Bailey).

⁴⁵ The main reason for suspecting that the correction is a transmitted one is that the act of conjectural correction would be easier if the errors communem and occidi had not already occurred. Marginal and interlinear corrections are often transmitted as such from codex to codex. Possibly the failure of Winterbottom to note A's omission may indicate that the supplement was made by the first hand, which would suggest that the supplement was transmitted. For an example of a conjectural correction transmitted differently to A and G, see my explanation of 5.14.21 in ‘Transmission’ 313. To the explanations there offered may be added the possibility that the common source of AG had nisut (a corruption of the correct nisi et), and that above this was superscribed a conjectural correction cum eìus: this produced in ‘a’ a text cum eius (over an erasure), but in G a conflation ṇịṣ cum eius ut.

⁴⁶ Modern scholars often act as if they believed that paleographic motives for errors were modern discoveries. Scribes met them in their every day existence, and found themselves committing them every day. Many medieval corrections operated on paleographic principles. The corrector known as ‘a’ is a particularly competent one, willing to attempt conjectural correction even of Greek words: see my ‘Transmission’. But in this instance the supplement may have been made at a still earlier point in the transmission, and been passed down as a marginal or interlinear correction. The deletion of the final occidit could also have been transmitted as dots under or above the letters, which subsequent scribes were able to observe (by omitting the word, or by transmitting both the word and the dots) or ignore.

⁴⁷ Winterbottom rejected in rem on the grounds that in after adplico, rather than a dative, is exceptional. It should have been a reason for rejecting the whole in finitionem.

⁴⁸ For attestation of both abbreviations, simil and simiia, see Lindsay, W. M., Notae Latinae (Hildesheim 1963), pp. 290, 336–9, Supp. 55.Google Scholar

⁴⁹ The verb ‘to be’ is often omitted in both halves of a division, but is also expressed in both halves, as in 7.1.24 Vltima species est an optare possil alienam uxorem. Generate est an quidquid optarit accipere debeat.

⁵⁰ For t he form with alter, see Quint. 8.3.83 eius duae sum species: altera quae plus significat quam dicit, altera quae etiam idquodnon dicit. For the use of aut … aut with wide separation, see earlier in the same chapter 7.7.2–3 (6–9 W) Colliduntur autem aut pares inter se, ut … conlalio est: aut secum ipsae, ut…

⁵¹ When Quintilian means ‘two types’, he says duae species, as in 8.3.83, quoted in the preceding note.

⁵² I would read Dousa's quotque: modo then means ‘just now’.

⁵³ For the expression, cf. Scaev. Dig. 28.2.29.14 nisi quod, licet audenter, possis dicere, Porph. in Hor. Carm. 1.12.1 ‘auritas quercus’ audenter dictum.

⁵⁴ The situation differs from the preceding sentence, where aerumnas can be explained as the expected object of exanclo. Neither reor nor autumo could govern prolem.

⁵⁵ Lindsay, W. M., Notae Latinae (Hildesheim, 1963), 49Google Scholar and Supp. 11, records di s for dicens, by analogy with which di i could be interpreted as dicenti. Abbreviations of forms of dico tend to be arbitrary, and to vary from codex to codex.

⁵⁶ ‘Ein makaronisches Ovidfragment bei Quintilian’, RhMus. 79 (1930), 253–78.Google Scholar

⁵⁷ The topic under discussion is coinage of new words.

⁵⁸ Since throughout 8.6 Quintilian quotes complete units of poetry, Badius' supplement is preferable to the simple deletion of ogra putant which Winterbottom entertains. In his text, Winterbottom prints ogra putant between daggers.

⁵⁹ CP 67 (1972), p. 57Google Scholar. Hunt, after entertaining non ‹minus quam›, preferred Spalding's et.

⁶⁰ For the contrast of quid and quo modo, cf. 5.13.4 Plurimum autem refert et quid protulerit aduersarius et quo modo. Polysyndeton goes well with a group of two, but asyndeton suggests a grouping of three.

⁶¹ The conflation would also have arisen if aliter extum was superscribed over ut specta, but specta is a word more likely to have arisen by conjecture in an attempt to make sense of exta (the corruption of exta to extū presumably followed the invention of specta, and may even have followed the act of conflation in an attempt to provide textum as an object for specta). For another example of conflation of a superscribed correction, see my explanation of 5.14.21 in ‘Transmission’ 313 (and above, n. 45).

⁶² It is unlikely that G's reading is misreported, since the variant anumerauimus is recorded in editions published well before Freund solved the crux.

⁶³ See note 2 of Winterbottom's discussion of the passage in ‘Problems’. Winterbottom translates: ‘Though a man is an enemy, he is yet a man.’ He perhaps has merely translated freely, since the TLL (6.3.2885.70) is surely right in classifying homo hostis under the category of adnominal modification: ‘Although an enemy man, yet a man.’ Cf. the other examples of adnominal modification with homo in TLL 6.3.2885.60–84.

⁶⁴ I have translated freely to preserve the pun. For uehementer conuenit cf. 6.2.19 maxime conuenit, 12.8.14 minime conuenit. I have retained the transmitted contigit, since it can be construed, but I would have no serious quarrel with any scholar who would prefer to take it as a simple slip for contingit (internal -n- is weakly pronounced, and often omitted).

⁶⁵ For such extension, cf. e.g. 11.2.32 et uelut oculis intuetur non paginas modo sed uersus prope ipsos, of someone recalling in his memory what he has written.

⁶⁶ It follows that I cannot accept the conjecture rector of Rutland, L. W., ‘Institutio Oratorio 10.3.25 a Suggestion’, RhM 128 (1985), 191–4Google Scholar. The silence of the night et al. do not function ‘as a tutor or guide’, but operate by not distracting. And the elimination of the predicate adjective with teneat produces an awkwardness: derexeris below is a variatio for rectos teneat.

⁶⁷ In his ‘Problems’, Winterbottom suggested aut ‹facile› deerit, and translated: ‘The only time we must devote to work is time that is superfluous to sleep—or at least can without difficulty do without it.’ This is neither good English nor a credible interpretation of the Latin text (it is not time, but the orator who would have to do without sleep; but if time is easily lacking to sleep, that would mean that it was sleep that was doing without time). Butler read haud deerit, and translated: ‘From these hours we must take only such time as is superfluous for sleep, and will not be missed.’ This gives to deerit a unique meaning, and one which cannot be justified as a contrast to supererit. For the meanings that can be attached to superest and deest in contrast with one another, see the examples which I cite below for such play in Quintilian, especially 4.2.44.

⁶⁸ 1.1.4 and 10.1.85. In the last instance, G has aut. In Tacitus, we find haud defutura in Hist. 4.39.5, and haud defuere in Ann. 12.7.1.

⁶⁹ See Watkins, C., ‘An Indo-European Construction in Greek and Latin’, HSCP 71 (1960), 115–19Google Scholar; Renehan, R., Greek Textual Criticism (Cambridge, Mass. 1969), 77–84CrossRefGoogle Scholar, and ‘Compound-Simplex Verbal Iteration in Plautus’, CP 72 (1977), 243–8Google Scholar. Most of the examples cited in Latin are from poetry, but that is only because poetry is where those who have written on the topic have looked.

⁷⁰ The TLL 5.2.1803.9 lists the participle as an equivalent for ‘immitis, asper, probrosus (opp. dulcis)’, but gives this as the sole instance. Butler translated it ‘harsh’.

⁷¹ On the practice of the Romans to speak and write in units which we call cola, see Fraenkel, E., Leseproben aus Reden Ciceros und Catos, Sussidi Eruditi 22 (Rome, 1968)Google Scholar, Kleine Beitraege zur klassischen Philologie (Rome, 1964), 1.73–92Google Scholar (‘Kolon und Satz’) and 1.93.119 (‘Kolon und Satz II’), Noch einmal Kolon und Satz (Munich, 1965)Google Scholar, Habinek, T. N., The Colometry of Latin Prose (U. C. Publ.: Classical Studies 25; Berkeley 1985)Google Scholar. In the sentence in question, the second colon is marked by the enclitic eum as starting with malle. Therefore the first colon ends with dicendi, not with fuit. Since the verb does not end its colon (one of the positions open to it), nor is it emphatic and initial in its colon, it must be in the only remaining position, the common one for a copula, enclitically second in its word group, here right after the first word of its predicate.

⁷² Cf. also p. 92.6, where Ammonius distinguishes the first four categories from the other six, as does Quintilian.

⁷³ This paper has benefited from the helpful criticism of W. S. Anderson, as well as of the students mentioned above. E. Courtney supplied useful bibliographic information. The first note, on 6.3.100, was presented in briefer form at the American Philological Association Convention in December, 1989. On 6.3.78, Courtney points out to me that Syme was anticipated in arguing for Vario (and in identifying him with L. Varius Rufus) by Koerte, A., ‘Augusteer bei Philodem’, RhM 45 (1890), 172–7Google Scholar, who himself (p. 173) gives credit to Teuffel, W. S., RLG ⁴ 51.1. But the fourth edition (Leipzig, 1882)Google Scholar was revised by Schwabe, who was probably responsible for lecting the reading. To judge from Halm's report, all were anticipated by the codex descriptus M (which Syme did not acknowledge).