In the following, I will propose and/or defend new or previous emendations for a set of passages in Catullus, most of which are deemed corrupt or even beyond repair by many, if not all, philologists.Footnote ¹ For the sake of simplicity, I will first quote Mynors's OCT text, except for possible changes in punctuation that will be justified, either implicitly or in their own terms, in the ensuing discussion. Those sections I consider incorrect I will put between obeli (which, on several occasions, are also Mynors's ones). In addition, I will reproduce the relevant manuscript readings recorded in Mynors's apparatus, checked against Thomson's more complete collations. In each case, I will begin with summarizing the state of the question; nevertheless, owing to the vast amount of corrections or conjectures to be examined for many passages, I will concentrate on the most significant proposals.Footnote ² Next, I will try to show that the correction suggested conforms to the constraints of metre and language, and (in some cases at least) sheds some light on the symbolic or intertextual dimension of the poem at hand. Finally, I will provide an account of the corruption process that presumably operated, with the aim of establishing the palaeographical verisimilitude of my proposal.

The triple negative connection neque … nec … nec requires ipsis to convey, as a pronoun, either a metonymical anaphora pointing back to Bithynia or a deictic reference, in indirect discourse and in the plural, to the (singular) speaker of the poem.Footnote ³ As argued by Trappes-Lomax, ‘ipsis meaning “the natives” is a strange expression and all the stranger in that Catullus was hardly concerned with their well-being’.Footnote ⁴ On the other hand, the equation ipsis = nobis = mihi is linguistically awkward; different corrections—nec/neque ipsi (AvanciusFootnote ⁵ ), mihi nec/neque ipsi (Statius; see Gaisser, 415), mihi ipsi (Pleitner,Footnote ⁶ Heyworth)—may solve this problem, but it has frequently been noticed that any distinction between Catullus and the other members of the cohors will prove inconsistent with the content of vv. 12–13.Footnote ⁷ In order to salvage the emendation mihi ipsi, Heyworth replaces praetoribus with quaestoribus;Footnote ⁸ but, as pointed out by Friedrich and Syndikus, the topical association between praetoribus and cohorti seems to be echoed in Tib. 1.3.1–2: ibitis Aegaeas sine me, Messala, per undas, | o utinam memores ipse cohorsque mei!; the same objection, as noted by Kiss, applies to McKie's ipsis | exactoribus. The only solution left consists in making ipsis modify praetoribus (see ipse at Tib. 1.3.2, quoted above). This can be obtained by substituting hi(n)c or nunc for the first nec of v. 10;Footnote ⁹ but hi(n)c creates an odd deictic anchoring (Catullus and his addressees are in Rome) and nunc increases the incongruity of esse, which ‘is in the wrong tense as Catullus is recounting his own past experience’ (Trappes-Lomax).

Trappes-Lomax assumes v. 10 to have been fabricated by an interpolator who consequently had to replace fuisse in v. 9 with neque (or nec) ipsis in order to restore the syntax. Fortunately, one can easily dispense with this radical move if, elaborating on Westphal's intuition, one prints ipsis | iam praetoribus. This emendation was suggested by Schmidt, who convincingly traced back the corruption of vv. 9–10 to the trivial confusion between iã = iam and in/iu.Footnote ¹⁰ In addition, Lucretius provides two attestations of the same syntactic and metrical configuration (5.52–3: cum bene praesertim multa ac diuinitus ipsis | iam mortalibus e diuis dare dicta suerit [†inmortalibus†, see n. 10]; 6.1015–16: quod facit et sequitur, donec peruenit ad ipsum | iam lapidem). GooldFootnote ¹¹ and Trappes-Lomax claim that nihil … esse … cur should mean ‘there is no cause/reason why’—which, if true, would rule out the datives (of interest rather than possession) ipsis … praetoribus and cohorti. But cur may be construed with a ni(hi)l antecedent whose potential referents do not reduce to indeterminate causes or reasons; see Ov. Am. 2.19.56 nil facies cur te iure perisse uelim?, which can be compared with 14.4–5 nam quid feci ego quidue sum locutus, | cur me tot male perderes poetis?, and Apul. Fl. 6 qui nihil habet adferre cur prandeat, impransus ad opus foras extruditur.

Since iam combines with neque … nec, v. 10 alludes to the fact that war and Roman looting have exhausted Bithynia's wealth (Schmidt), and contextually implies that this annihilating process was already at work when Catullus first arrived in that country. Thus, while the situation referred to by esse remains in force at the moment of the poem's utterance by the speaker, its consequences for Catullus were triggered in the past. Notice the formal parallelism between iam Bithynia (v. 7) et iam praetoribus (v. 10): the first occurrence of the adverb may seem enigmatic (Bellandi) but the repetition elegantly suggests temporal continuity (Schmidt). Given that the destructive consequences of Roman policy are likely to persist, the plural praetoribus acquires a generic value by applying not only to Catullus’ own praetor, but also to all praetores in Bithynia, present or to come (see Cic. Verr. 2.1.131 pupillos et pupillas certissimam praedam esse praetoribus). Accordingly, the singular cohorti should be assigned a distributive reading: for each such praetor, there will exist one and only one cohors (Bellandi). Finally, one can reasonably envisage that Bithynia's condition is, and will be, constantly worsening, so that Catullus' own praetor, contrary to his present or future successors, still had the opportunity to fiddle some cash or goods without sharing them, obviously, with the men of his cohors. This crude picture of greed reminds us of other poems, notably 29, which has in line 22 (see below) a similar usage of unctus that also appears at Cic. Verr. 2.2.54.

In v. 26, commoda is ordinarily analysed (scanning the final syllable as short) as the dactylic variant of the imperative commodā.Footnote ¹² Such shortenings of cretic words occur in anapaestic lines of Old Comedy,Footnote ¹³ but in (pre-)classical verse they only apply to first-person verb forms, or proper names, ending in -ō (see, for instance, 85.2: nesciŏ, along with the frozen nescio quid/-s at 2.6, 6.4, 53.1, 80.5; Hor. Sat. 1.10.42, 1.10.85, Carm. 2.1.14 Polliŏ; Maecenas 3.2 [Courtney, FLP] diligŏ). In addition, it is highly implausible for such a process to affect a word-final -ā in a corpus that does not contain comparable examples with the weaker vowels -ē, -ī, -ū. Scaliger's commodo nam uolo ad Serapim | deferri runs against the same objections, while Friedrich's hypothesis of an adverbial commoda does not rest on any evidence. Burmannus' commodā enim and Hand's commodum enim …, which aim at improving on Scaliger's solution, imply unacceptable elisions.Footnote ¹⁴ In Catullus, all elisions, before a short syllable, of a cretic-final word ending in a nasal vowel (a graphic -m) belong to dactylic lines (64.211, 64.359, 64.366, 66.27, 67.31, 68.3, 68.4, 68.90, 73.6, 90.2, 91.9). Except for 97.6 (ploxeni habet), which features a very rare word (TLL s.v.), comparable examples with oral vowels involve words ending in -ō (6.11 argutatio inambulatioque; 85.2 sentio; 91.2 perdito amore) or -ē (45.3 perdite amo; 104.3 perdite amarem). This reflects the influence of dactylic verse, where all elided iambic-final words whose penultimate syllable corresponds to the first element of a pyrrhic end either in a nasal vowel or in one of the three vowels -ī, -ē and (possibly shortened) –ō, Lucil. 1071 (Marx) (delicā aitque) being certainly corrupt.Footnote ¹⁵ Moreover, since argutatio does not fall within strictly cretic words,Footnote ¹⁶ and sentio most probably had its last vowel shortened like nesciŏ in the same verse (see above), the examples with an oral vowel are three variations of the same stereotypic formula (see below, on 55.4). This also casts serious doubts on the plausibility of commodo/-e enim, proposed by BarthiusFootnote ¹⁷ and Mueller respectively.

According to Trappes-Lomax, commoda ‘may well be an interlinear gloss' that has replaced crastina.Footnote ¹⁸ But, as pointed out by Ellis and Nisbet, a simpler option consists in interpreting commoda as the plural neuter substantive attested at 23.24 (in the same metrical position), 68.21 and 84.1; see also incommoda at 14.23 and 68.11. Since commoda then substitutes, like commodum/-o/-e or crastina, for a fully expressed verbal predicate the absence of which would definitely blur the syntax (as also happens with Statius' nam uolo commode ad Serapim), Nisbet corrects quaeso into quaero, and istos into istaec (see 67.37) rather than ista so as to avoid a trochaic base.Footnote ¹⁹ Yet, quaero sounds far less natural than quaeso at the beginning of the girl's turn at speech.Footnote ²⁰

By proposing to substitute da modo for commoda or da istos for istos, Doering and AgarFootnote ²¹ have taken a first step toward an alternative solution, viz. da mi commoda: nam uolo ad Serapim (‘Give me these facilities [that is: lend me your litter-bearers]’). We find the same metrical and syntactic collocation of da mi and a dactylic neuter plural noun at 5.7: da mi basia mille, deinde centum. The introduction of dative mi requires mihi to be construed with inquit in v. 25—a punctuation that, pace Mynors and Nisbet, seems preferable on independent grounds.Footnote ²² The collocation da … commoda is a clear imitation of Plautine diction: at As. 445, si uelis, da commoda homini amico! should be translated as ‘if the fancy takes you, give facilities [that is, lend something] to a man who is your friend!’ (see 444: scyphos … utendos dedi).Footnote ²³ This parallel confirms that, in our verse, commoda means ‘loan’ (as assumed by Ellis; see Cic. Verr. 2.4.6) and not ‘perquisites’ (as assumed by Nisbet). Moreover, both Plautus and Terence use da mi: da mi, optuma femina, manum (Aul. 135; perhaps mihi); da mi operam parumper (Aul. 199); da mi igitur operam (Cur. 259); sed obsecro, da mi operam (Truc. 722); tum tu igitur paululum da mi operae (Eu. 281); mi gnate, da mi hanc ueniam (Hec. 605 [Bentley]). The comic intertext thus accounts for quaeso (see obsecro in Truc. 722), for paulum (see parumper in Aul. 199 and paululum in Eu. 281) and for the polyptoton mihi … mi … mi (see Hec. 605, if Bentley was right) that aptly mimics the seductive behaviour and talk of the wench. The corruption of the line stems from a mistake triggered by the repetition of mi or (-)da: a scribe wrote either †mi Catulle paulum | mi Catulle (paulum) commoda† or †commoda (mi) commoda†; the unmetrical line opening †mi Catulle (paulum)† or †commoda (mi)† was subsequently eliminated and the gap filled with a pronominal form derived from istud (v. 28).Footnote ²⁴

In v. 27, pyrrhic manĕ can be accounted for by iambic shortening or by prosodic hiatus; this uncertainty makes the text suspect. If we leave aside the corrupt sequences †tĕ in† and †modo ipse† of 55.4 and 114.2 (see below), the only prosodic hiatus that can be assigned with certainty to the Catullan corpus (97.1: ita me dĭ ament) belongs to a frozen expression inherited from comic diction (for instance, Pl. Am. 597; Ter. Ph. 883), and thus exhibits the usual shortening of a monosyllable before an iambic word.Footnote ²⁵ Trappes-Lomax argues in favour of minime! ‘no way!’ (ς), but such an abrupt rebuttal does not fit in with the speaker's otherwise dilatory reply; as for Bergk's mi anime,Footnote ²⁶ it sounds too (ironically) affectionate in this context. Furthermore, we may retain iambic shortening in mane by supplementing this imperative with tu (mane tum ς, me ne tu ς); see Pl. Mos. 885a: manĕ tu at- |que adsis-| te ili- |co (lecythium, again with iambic shortening in mane and elision of tu).Footnote ²⁷ Catullus quite frequently elides te or tu before prepositional or prefixal in(-): te incolumem (9.6), te inepte (12.4), te in (31.6), te in Circo … te in templo (55.4–5, see below), te in (64.149), tum te indomitam (68.118, dubious; tunc te ς, tum te Riese, indomitam Statius); in our poem, tu insulsa (v. 33).Footnote ²⁸ Once inquii had given rise to †inquit†, tu was replaced with †mi†, eventually misread as †me†; mi for me occurs at 16.3, and me for mi at 21.11 and 37.11 (see below, on 55.10).

All editors and commentators point out the difficulty created by die in v. 14. Friedrich quotes Pl. Capt. 333 (optumusque hominum est homo) and claims that Saturnalibus is an inserted apposition; but this grammatical figure does not occur in Latin poetry before VirgilFootnote ²⁹ and periret oddly separates die from optimo dierum. Watt's misti, hoc continuo ut die maintains the inelegant sequence die … optimo dierum; McKie's diu requires continuo to be construed with misti (as also assumed by Heyworth) and conveys a notion of graduality that proves inconsistent with the overall meaning of the passage (Kiss); Heyworth's ut periret ipsis, advocated by Trappes-Lomax, is palaeographically implausible. I propose to replace die with tibi: other examples of such ethical-possessive datives occur in Catullus (see below, on 55.10); the use of tibi after tuum Catullum (v. 13) recalls Cic. Fam. 9.2.1 (Caninius tuus … at tibi repente paucis post diebus, cum minime expectarem, uenit ad me Caninius mane); datives for the person(s) responsible or beneficiary are found with pereo, for instance at Sen. Ag. 524–5 quid hoc simul perire nobiscum iuuat | quibus perimus? and Phoen. 170–1 membratim tibi | uolui perire (TLL 10.1.1340.66–70). Haplography produced †ut ibi† (see †ut ibi† G for ut tibi at 66.24); under the influence of v. 15, metre was restored by substituting †die† (wrongly construed with continuo) for the temporal adverb.

Among the possible substitutes for †ac sulcis† or Schwabe's hapactic axulis, one should rule out Palmer's aesculis (‘oak piles’), given that there is no parallel example for such a use of aesculus and that, in its two attestations (Cic. Verr. 2.1.147: nullo lapide rediuiuo; Vitr. 7.1.3), adjectival rediuiuus (‘made of material recycled from demolished buildings’) qualifies (free)stones. McKie and Kiss adopt assulis (Statius), which they interpret as designating marble chips (see Vitr. 7.6.1 caementa marmorea, siue assulae dicuntur, quae marmorarii ex operibus deiciunt, pilis ferreis contunduntur cribrisque excernuntur); this hypothesis makes rediuiuis oddly pleonastic. Lafaye's referentially inadequate arcul(e)is can by improved on by editing arcubus (‘arches’; TLL 2.479.67–74). Nisbet proposed replacing rediuiuis by recidiuis (‘back-sliding’), but there is otherwise no parallel for recidiuus in this sense. Both McKie and Trappes-Lomax (who prefers insulis, ‘artificial islands’, without any parallel use attested) claim that crura makes any further mention of props, piles or arches superfluous. But, as shown by Rudd,Footnote ³⁰ the decrepitude of the bridge matches that of the man the speaker wants to humiliate; in such a context, the co-occurrence of crura and arcubus helps the reader to access the symbolism at work.Footnote ³¹ The corruption process probably started with the solecism †arcibus† (TLL 2.475.76–9); double metathesis (†acrubis†, which created a rhyme between the two subverses) and the usual confusion between b and li yielded †acruliis†, normalized as †ac sulcis†.Footnote ³²

If a nominative, noui libri (v. 6) should normally come before cartae regiae in that it refers to the whole books, and not to any specific part of them; if a genitive, it deprives the enumeration of most of its rhetorical force (Kroll; Syndikus, 1.157 n. 12). Vindicating the manuscript reading nou(a)e, Nisbet proposed reading nouae bibli (adopted by Goold and Syndikus, 1.157), taken as a genitive (‘of new papyrus’) rather than a nominative (‘new papyri’), and parallel to the genitive membranae. In favour of his correction, Nisbet pointed out that v. 6 with libri is the only Catullan scazon where the penultimate position is filled by a syllable that, given the presence of an obstruent + liquid group, may remain short in other metrical environments, while ‘bl lengthens automatically at its rare occurrences in classical Latin, as it normally does in Greek’.Footnote ³³ But scazon-final lābris (Mat. 11.2 [Courtney, FLP]; mentioned by Nisbet) and fēbris ([Verg.] Cat. 2.3), as well as several comparable examples in Martial (1.96.13, 6.39.8 lābris; 3.93.12 mācrae; 7.7.10 Tīgris; 8.61.4 cēdro; 10.100.2 lībro, mentioned by Nisbet; 12.65.4 lībram), seriously weaken this argument.Footnote ³⁴ I suggest printing notae cedri, which designates the traces, stains or fragrances left by the cedar oil used for preserving books (Vitr. 2.9.13). Poets frequently evoke the cedrus: Hor. Ars P. 332; Ov. Tr. 1.1.7, 3.1.13; Pers. 1.42–3; Mart. 3.2.7, 5.6.14, 8.61.4 (where cēdro occurs in scazon-final position). For this use of plural nominative notae, see Lucr. 6.220–1 inusta uaporis | signa notaeque graues halantis sulpuris auras and Ov. Tr. 1.1.7 nec titulus minio, nec cedro carta notetur; see also nota and noto at 68.28, 68.148, 80.8, 107.6. Owing to the confusion between capital d and b Footnote ³⁵ and to the context, cedri was read as †libri†; not(a)e shifted to †nou(a)e† under the influence of following noui and, perhaps, 1.1 nouum libellum; see †nouo† for notat at Culex 71 (Salvatore), with Housman, 2.781.

The first i of conscribillent was undoubtedly long (Kroll). Pighi and Goold follow TurnebusFootnote ³⁶ in printing conscribilent flagella, but nothing, in the evidence available (TLL 4.473.71–6), allows us to envisage that, if the orthography with a simple l corresponds to phonetic reality, the second i did not undergo compensatory lengthening consecutive to the loss of the geminate consonant. Among the verb forms that have been proposed as substitutes for conscribil(l)ent, Trappes-Lomax's contribulent (before flagella) is the only one that may seem plausible, but the first attestations of this word occur in Christian texts (TLL s.v.). In addition, inusta (Calphurnius) should inspire suspicion, since this epithet always applies to the individual whipped or to the marks left by the whipping (TLL 7.2.270.10–12). In my view, we need a more radical solution: inurat acriter tibi, cum sibilet, flagellum. The adverb acriter regularly features in descriptions of burning, torture and similar (physical or mental) states or actions (Lucr. 3.289; Cic. Div. 1.66; Sall. Hist. frg. dub. uel fals. 4 [Maurenbrecher]; Livy 27.3.5, 39.43.2; Vitr. 2.3.2; Ov. Am. 2.19.3, Her. 19.15, Ars am. 2.354; Tac. Ann. 4.40.3, 15.57.1); see also Verg. Aen. 9.718 et stimulos acres sub pectore uertit and Tac. Germ. 6 ferro … acri. The subordinate clause cum sibilet (an improvement on consibillent, ς) aptly evokes the whistling sound of the whip: see 4.12 loquente saepe sibilum edidit coma; Pac. 334–5 (Ribbeck, TRF) flictus nauium | strepitus fremitus, clamor tonitruum et rudentum sibilus; Luc. 2.698 nec quatiunt ualidos, ne sibilet aura, rudentes; Sil. 17.256–7 stridorque inmite rudentum | sibilat. Given that sibilo is very naturally predicated of (objects shaken or tossed by) winds or tempests (4.12; Pac. 334–5; Sil. 17.256–7; [Quint.] Decl. 12.16 scissis rudentibus tempestas sibilat), a perceptual continuity relates our line to vv. 12–13, and thus to v. 4 Thalle, turbida rapacior procella, Thallus' announced misfortune being conceived of in the same metaphorical terms as his thefts. A mis-segmentation, coupled with the confusion between p and c ,Footnote ³⁷ with metathesis involving r (see above, on 17.3) and the a/u alternation, yielded †inura turpiter†, †inura† being aberrantly normalized as †insula† (see n. 28 on 10.33). The search for grammatical agreement, favoured by the constant analogy between whipping and burning, led to †inura … flagella†. As usual, s(e)i- shifted to †sci-†, hence †scĩ† = †scri†.Footnote ³⁸ Once cum sibilet had been altered into †conscribil(l)e(n)t†, a rhythmical imperative triggered the permutation: in this poem, no word of more than two syllables has its stress falling on the weak position of a foot; consequently, the accentual anomaly of †conscribíl(l)e(n)t flagéllum/-a† was eliminated by writing †flagéllum/-a conscribíl(l)e(n)t†.

Modern editions waver between the two plausible emendations ante (Statius) and uncti (G. Faernus, quoted by Statius; uncta ς, unctum Scaliger; see Gaisser, 376). Yet ante lacks relevance, in that it simply makes explicit a time reference already implied by the context, while uncti is a negatively oriented word that would aptly apply to Mamurra himself (see 10.11) or his possessions (see v. 22), but not to Gaul or Britain at the time they were still preserved from his destructive greed. I choose to read umquam so as to provide the free relative clause with an emphatically universal value; see, for instance, Pl. Poen. 449–50 di illum infelicent omnes qui post hunc diem | leno ullam Veneri umquam immolarit hostiam; Verg. Aen. 2.331 milia quot magnis umquam uenere Mycenis; Livy 4.6.9 quicumque aliquid seditiose dixerat aut fecerat umquam. This construction occurs in a mutilated distich that relates to poems 10 and 29: Bithynia quicquid | et pedicator Caesaris umquam habuit (Calv. 17 [Courtney, FLP] = Suet. Jul. 49.2); one may wonder who the imitator is. Imperfect habebat expresses habitualness (‘whatever long-haired Gaul and farthest Britain used to own’); compare with a similar use of the present tense at Ov. Met. 583–4 intumui quantusque, feror cum plurimus umquam, | tantus eram (‘and I was as great as I am whenever my flood is at its fullest’). This suggests, again, that Roman looting is a systemic process that does not take place in one fell swoop. I assume that umquam first shifted to its near-synonym †cumque† correlated with quod: dittography (†habebat tum-†) produced †cum-† and the common confusion between quom, (-)quam and -que took place. Then †cumque†, written †cumce†, evolved to †cum te†.Footnote ³⁹

Such a learned poet as Catullus could not resort to Adoneus/-ius,Footnote ⁴⁰ a native adaptation (attested at Pl. Men. 144) of the original Greek proper name.Footnote ⁴¹ Among the corrections proposed, Dioneus/-ius (Calphurnius et al.), Sidonius,Footnote ⁴² Thyoneus/-ius (Ellis [n. 40]) and Cydonius (Oksala) do not make sense (see Zicàri, 265 on Cydonius); Dodoneus (Herrmann) does not scan (Zicàri, 244); Aphrodisius (with elimination of aut; Trappes-Lomax) and ciconius/-eus (McKie) are unjustifiable hapaxes (see Kiss on ciconius). In addition, all those emendations create an awkward disjunction or asyndetic collision between albulus and the following epithet. I suggest editing ut albulus columbus aut olor nitens; olor probably occurs in Lucilius (268 [Marx] = Non. 200.20 [Lindsay])Footnote ⁴³ and nitens is a Catullan word (see 2.5, 61.186; also enitens at 61.21, nitor at 66.3). Doves and swans belong to Venus' attributes (for swans, see Hor. Carm. 3.28.15, 4.1.10; Sil. 7.441; Stat. Silv. 3.4.22), but the disjunctive structure of the noun phrase and the erotic overtones of the passage rather favour a reference to Leda's legend (Ov. Met. 6.109 fecit olorinis Ledam recubare sub alis; Man. 1.339–40 cum deus in niueum descendit uersus olorem | tergaque fidenti subiecit plumea Ledae; Mart. 1.53.8 inter Ledaeos ridetur coruus olores; Stat. Silv. 1.2.142 Amyclaeos … olores, Theb. 10.504–5 lubrica ripa Lacaenae | uirginis et falso gurges cantatus olori;Footnote ⁴⁴ [Sen.] Oct. 204–5 dominus caeli diuumque pater | … pennas sumpsit oloris); the same association between Venus' doves and Leda's swan occurs at Mart. 8.28.13 Spartanus tibi cedet olor Paphiaeque columbae. For nitens meaning ‘brilliantly white’, see Verg. Aen. 3.20–1, 6.895; Hor. Carm. 1.14.19–20, 2.5.18; Ov. Met. 1.610; Ciris 524–5; Sen. Tro. 184;Footnote ⁴⁵ at Germ. Arat. 278, the collocation Cycnum … nitentem designates the constellation, also called Olor (TLL 9.2.572.29–43). An intricate network of metrical and sound parallelisms relates columbus aut olor nitens to superbus et superfluens—repeated (-)ol- discretely echoing the wordplay on super-. Quite expectedly, olor was mixed up with †odor†,Footnote ⁴⁶ while nitẽs mechanically shifted to †neus† (see †uentus† for ueniens at Germ. Arat. 588; Housman, 2.514); the normalization of †odo(r)neus† as †idoneus† recalls that of †idcoeos† (from Idaeos) as †idoneos† at 64.178 (Friedrich, 360; see below, on 55.9).

Among the editors of Catullus, only Bardon and Della Corte follow Deroux in printing quid est? alit … in v. 15.Footnote ⁴⁷ In favour of this conservative reading, Deroux mentions the pun it creates on est (‘What is happening?’ or ‘What does he eat?’) and a similar collocation: pu]eri et puellae qui ex liberalitate eius [Imperatoris] aluntur.Footnote ⁴⁸ But, as shown by the translations in Bardon (‘Eh bien! c'est une largesse de voleur qui l'alimente’) and Della Corte (‘Come no? Lo mantiene la prodigalità dei ladrocinii’), the statement obtained abruptly interrupts a sequence of interrogative lines (vv. 14 and 16) expressing indignation (Zicàri, 270); more importantly, it entails that Caesar's (and Pompey's) connivance with fraud and looting is ‘feeding up’ Mamurra, whereas the overall content of the poem precisely points to the impossibility of putting an end to his greed.Footnote ⁴⁹ Yet, alid = aliud (Statius; Gaisser, 414) does not yield better sense, nor even correct syntax (Syndikus, 1.180 n. 21): in all parallel examples, the grammatical subject that directly combines with the collocation quid est aliud…? is an infinitival clause (Cic. Div. 2.78, Off. 2.83, Phil. 1.22, 2.7, 3.21, 5.5, 10.5, 12.13, Pis. 47, S. Rosc. 54, Verr. 2.1.128, 2.3.71; Livy 6.40.8; Ov. Her. 18.181–2) or an equivalent relative clause with a subjunctive verb (Cic. Cael. 32). Faernus' quid abstulit…? (reported by Statius), though failing to solve our problem, suggests the most attractive emendation quid extulit…? Such uses of effero meaning ‘generate, produce’ are well-documented (TLL 5.2.143.46–81); see, for instance, Cic. Sen. 9 artes exercitationes uirtutum, quae … mirificos efferunt fructus, Verg. G. 2.167–9 haec [that is Italia] genus acre uirum … extulit, Petr. 120.98–9 horrida tellus | extulit in lucem nutritas sanguine fruges. Moreover, Phaed. 2.9.15–16 sin autem rabulis doctus occurrit labor, | sinistra quos in lucem natura extulit may well imitate Catullus. The morphophonological parallelism between extulit and expatrauit (v. 16) favours the word-play relating the second form to paterna (v. 17) and patrimonia (for the semantic affinity between [h]elluor and patrimonium, see Cic. Sest. 111, [Verg.] Cat. 13.11, Gell. NA 2.24.11). A mis-segmentation †ext ulit† was spontaneously read as †est alit† (see, for instance, †testa† X for texta at 64.10) under the influence of the context.

In v. 19, scit hardly makes sense. The comparable corrections proposed by Ezra de Clercq van Jever (qua nitescit)Footnote ⁵⁰ and McKie (qua tumescit) require the elimination of amnis. Wiman'sFootnote ⁵¹ unxit proves unsatisfactory, in that ‘the verb ung(u)ere (as opposed to its past participle unctus – cf. 29.22) is never used for the idea of metaphorically anointing or oiling, i.e. enriching, and seems inappropriate for the more liquid action of a river’ (so Harrison, who opts for auxit, already proposed by MarciliusFootnote ⁵² ). I think egit is definitely preferable, given the frequence of the collocation praedam ago in military contexts (TLL 1.1368.81–1369.7, 10.2.528.33–6); in Livy, one can find 31 examples of it (for instance, 25.31.13 praedamque omnis generis retro ad naues egit), even without taking prefixed derivatives from ago into account. This technical expression immediately makes sense in a poem dealing with Caesar's campaigns and with Mamurra's constant looting; by combining it with aurifer (ago and fero are synonymous in this context), Catullus makes Tagus appear both as another Pactolus and as a soldier who brings booty to his commander. Owing to the confusion between capital g and c ,Footnote ⁵³ egit shifted to †ecit†, subsequently normalized as †scit†; see †eciet† for sciet at Priap. 52.12 (Clairmont).

In order to filter out implausible candidates among the innumerable emendations that have been proposed for v. 20, I will adopt three general criteria. First, we should dismiss the hypothesis put forth by Minyard and Deroux that the first foot might be spondaic owing to the presence of line-initial nunc or another prosodically equivalent word;Footnote ⁵⁴ the deviation with respect to a pure iambic metre that occurs in v. 3 is licensed by the proper name Māmurram (for similar phenomena, see trisyllabic Camēri̯um at 55.10, Au- | runculeia at 61.82–3 and perhaps the syncopated or hypermetric Rauide at 40.1). Secondly, any correction that posits interrogative -ne at the third (short) position produces an utterance that lacks the rhetorical force needed in the context at hand, since non-focussing ‘yes-no’ questions with -ne have an argumentative orientation that assimilates them to weaker variants of the corresponding negative statements.Footnote ⁵⁵ Thirdly, all reconstructions that maintain some form of timeo run against the definitive objection that neither Gaul nor Britain had anything left to fear about, that nobody could feel fear for them anymore, and that the whole poem deals with Mamurra's past and actual plunder, not with his future actions (Syndikus, 1.180–1 n. 22). Two interesting conclusions emerge from this systematic examination. At the third (short) position, enclitic -ce or -que are good substitutes for -ne.Footnote ⁵⁶ Instead of … timet, one may envisage (et) ultima(m) (et) or ultimae,Footnote ⁵⁷ which provide a good parallelism with v. 4; notice, however, that the correction ultima et would exclude the possibility of construing ultima with Britannia, since Catullus never postposes et (Housman, 2.626). Given that the only conjecture that meets all the conditions formulated (sciuntque Galliae ultimam [praedam] et Britanniae; Birt [n. 56]) loses all plausibility after the elimination of scit from v. 19, I suggest caduntque Gallia ultimae et Britannia, which resorts to the same military lexicon as [praedam] egit; for a similar collocation, see 11.22–3 cecidit uelut prati | ultimi flos Footnote ⁵⁸ and Sen. Tro. 243 tum saeua Amazon ultimus cecidit timor. The corruption of the line involved different processes. First, a mis-segmentation †Gallia ul tima e et† produced †Gallia ut tima et et†; confusions between e(x), es(t) and et abound in manuscripts.Footnote ⁵⁹ Secondly, †ut† shifted to †et†, as constantly happens, and †tima et† to †timet†.Footnote ⁶⁰ Thirdly, under the influence of v. 21, caduntque was replaced by †quid hun(c)que† = †quid hun(c)ce†.Footnote ⁶¹ At that stage, the line read as follows: †quid hun(c)ce Gallia et timet et Britannia† (metrical, except for the resolution on †timet†). Fourthly, the rare form †hun(c)ce†, only (possibly) attested at Laber. 21 and Apul. Apol. 99, reduced to †hunc† and †Gallia et† shifted to †Galliae†.Footnote ⁶² This triggered the adaptation to †Britanniae†, perhaps on the model of 45.22 Syrias Britanniasque or Plin. HN 17.42–3 alia est ratio, quam Britanniae et Galliae inuenuere … Gallias Britanniasque. Finally, the unmetrical line-opening †quid hunc Galliae…† was improved on by eliminating †quid†.

In v. 23, the resolution and the spondaic foot of opulentissime/-i cannot be maintained (again pace Minyard and Deroux [n. 54]). The less implausible emendation compatible with urbis (o potissimi, Mueller) runs against palaeographical objections, and one may reasonably hesitate between the referentially adequate but useless urbis and Haupt's hyperbolic orbis.Footnote ⁶³ Trappes-Lomax advocates eone nomine, imperator unice (Scaliger), but this repeated insult (see v. 11) does not lead naturally to the vocative that follows. We will obtain an excellent text if we successively adopt Maude Thompson's o bis and Giri's optimi uiri: eone nomine, o bis optimi uiri, | socer generque, perdidistis omnia? Footnote ⁶⁴ For bis meaning ‘both, doubly’ and not ‘twice (at different times)’, see Eleg. Maec. 2.29 = 173 et tibi succrescant iuuenes bis Caesare digni, about Gaius and Lucius Caesar (Schoonhoven ad loc.) and Stat. Silv. 3.2.64–6 nec enim temeraria uirtus | illa magis, summae gelidum quae Pelion Ossae | iunxit anhelantemque iugis bis pressit Olympum (Vollmer ad loc.).Footnote ⁶⁵ The shift from †obis† to †õbis† = †orbis† is palaeographically trivial, and manuscripts systematically waver between †orbis† and †urbis†. Owing to repeated mistakes on a series of minims, optimiuiri was read as †oplitissimi† = †oplẽtissime†, expectedly normalized as opulentissime; this was favoured by the semantic affinity of optimus and opulentus (see Enn. scen. 259 [Vahlen] = Cic. Fam. 7.6.1 matronae opulentae optimates).

Commentators have shown anxious to associate with Venus all the place-names occurring in this passage. Ancona's and Durrachium's cult and temple of VenusFootnote ⁶⁶ can perhaps be put on a par with the legendary Cnidus, Amathus and Golgoe (on this form, see below); but the enigmatic Vrios, approximately identifiable with the name of one or another obscure port or town,Footnote ⁶⁷ definitely looks like a non-existent word produced by textual corruption.

Since there is no reason for assuming v. 15 to be interpolated (see n. 66), one should envisage that, by evoking the Adriatic and some of its ports, Catullus aimed not only at establishing a loose symbolic link to sea-born Venus, but also at alluding to some referentially relevant detail. Williams (n. 66), Wiseman (n. 67) and Goold think of Catullus' route home from the East in 56. But Volus(i)us lived in, was native from, or related to some place in the delta of the Po (see below, on 68.157–8) and he may have been a Picene, so that it would have been quite natural for him to situate an episode of his annals in the Adriatic space.Footnote ⁶⁸

Few attempts have been made to emend Vriosque apertos: Vriosque portus (Heinsius, dubitanter)Footnote ⁶⁹ does not solve the problem, but interestingly suggests that the participle form might be corrupt too; Chytrosque apertos (Bergk),Footnote ⁷⁰ from Gk. Χύτροι designating a town in Cyprus (see, for instance, Xenagoras, fr. 10, FHG 4.527), makes good sense, but remains devoid of any mythical overtones; Trappes-Lomax, after considering the palaeographically implausible Paphonque celsam, prefers to obelize the manuscript reading. I propose to print quae sanctum Idalium, quae Hydrunta portum. Feminine Hydrus, Hydruntis is the ancient name of Otranto, the nearest Italian port to the Epirean coast and thus a natural stopping place for people who crossed the Adriatic (Cic. Att. 15.21.3, 16.5.3, Fam. 16.9.1; Livy 36.21.5; Plin. HN. 3.100–1; Mart. Cap. 6.650). Though other Latin writers only use the accusative Hydruntem (Cic., Livy) or, in later times, Hydruntum (Mart. Cap., Paul.-Nol. Carm. 17.85),Footnote ⁷¹ Hydrunta here imitates Gk. Ὑδροῦντα (Str. 6.3.5). Similarly, Catullus resorts to Ancona instead of Anconem (Cic. Att. 7.11.1) or Anconam (Caes. BCiv. 1.11.4; Cic. Fam. 16.12.2, Phil. 12.23; Livy 41.1.2; Mela 2.64; Plin. HN 6.218; Tac. Ann. 3.9.1), and to the exceptional Amathunta, never found elsewhere, except at Ov. Met. 10.220, 10.531. In addition, Golgos does not belong to the Latin declension of an otherwise unattested Golgi, but adapts the Dorian accusative Γολγώς (nominative Γολγοί; see Golgoe at Plin. HN 5.130) used by Theocritus, Id. 15.100–1 Δέσποιν᾿, ἃ Γολγώς τε καὶ Ἰδάλιον ἐφίλησας | αἰπεινάν τ᾿ Ἔρυκα, χρυσῷ παίζοισ᾿ Ἀφροδίτα. Since 64.96 quaeque regis Golgos quaeque Idalium frondosum confirms that Theocritus' verses are one of the main sources of our passage, we can conjecture that the accusative Ἔρυκα (with lengthening of the final vowel by the following obstruent + liquid group) induced Catullus to accumulate comparable forms in his verses. In view of the fact that words beginning in hydr- always have an initial long syllable in Latin poetry (see, for instance, Hydrochoi at 66.94 and spondaic Hydrus at Luc. 5.375),Footnote ⁷² the amphibrachic scansion of Hydrunta may seem unacceptable. But this exception takes its roots in Greek prosody, where words beginning in ὑδρ- and etymologically linked to ὕδωρ frequenly tolerate Attic correption (see Appendix A); moreover, H y̆drunta may find an echo in Hădriaticus (see n. 68). The postposition of portum obeys to the norms of topographical language (see, for instance, Mela 3.11 Ebora portum; Plin. HN 3.34 Citharista portus). The presence of Hydrunta portum increases the density of sound effects triggered by the parallelism Hydrunta-harundinosam-Amathunta and by the repetition of the nasal vowel -um (ten occurrences with portum); notice the strict phonological, metrical and even orthographical correspondance between -dalium quae Hydrunta and -na Cnidumque harundi-. This might allude to some characteristic feature of Volus(i)us' diction.

A mis-segmentation †Hydrunt aportum† gave rise to †apertum†. The relative quae was read as †-que†Footnote ⁷³ and moved after †Hydrunt†. †Hydrunt†, written †idrũt†, shifted to †utrus†, †utrios† that reduced to †urios†; a similar confusion between ıt and d occurs at Ov. Met. 8.806 where, owing to metathesis involving r (see above, on 17.3 and 25.11), †ordĩe† = †ordine† stems from crate read as †oride† (Housman, 1.133). Finally, grammatical agreement yielded †apertos†.

Contrary to what is claimed by Trappes-Lomax (57), the prosodic hiatus tĕ in (v. 4) does not conform to any well-established constraint of archaic or (pre-)classical metre. In archaic (essentially comic) diction, monosyllables shortened by prosodic hiatus normally precede an iambic word (see above, on 10.27). In the far less frequent configuration where the second word itself is a short monosyllable (as happens here),Footnote ⁷⁴ the first word either (i) is the accusative pronoun me placed as an enclitic just after its verb and before a syntactic boundary, or (ii), more commonly, belongs to a closed set of grammatical words (relative / interrogative pronouns, subordinating conjunctions, non-lexical adverbs); case (ii) is also attested in dactylic verse (see Appendix B). Neither can tĕ in gain plausibility from a parallel with Verg. Aen. 6.507 nomen et arma locum seruant; tĕ, amice, nequiui. Firstly, Virgil's example belongs to a corpus where prosodic hiatuses often testify to a Greek influence (see Ecl. 2.65, 3.79, 6.44; G. 1.281, 1.437, 4.461; Aen. 3.211, 5.261). Secondly, the etymology that links amicus to amo Footnote ⁷⁵ entails that amice should be put on a par with the trisyllabic forms of either word that license shortenings like oscu-|lum … sat est| oscu-|li mihi| uostri. | quĭ ama-|bo, pa-|ter? (tro⁷, Pl. St. 91) or quin tuos | inimi-|cos poti-|us quăm a-|micos | uocas? (ia⁶, Pl. Ps. 880) by virtue of an analogy with ita me dĭ ament (see above, on 10.27), quĭ amat (Pl. Trin. 242a), tĕ amo (Ter. Ad. 680), etc. (Drexler [n. 25], 2.310; Lindsay [n. 13], 247); see also, in dactylic verse, quĭ amant (Verg. Ecl. 8.109) and mĕ amas (Hor. Sat. 1.9.38). The anomalous elisions perdito amore, perdite amo, perdite amarem (see above, on 10.26) and Pollio amat (Verg. Ecl. 3.84, 3.88) illustrate this peculiar status of amo too (Soubiran [n. 15], 207–18), while at contra quăe amara atque aspera cumque uidentur (Lucr. 2.404) originates from the wordplay relating love to bitterness.Footnote ⁷⁶ Thirdly, the shortening in tĕ amice is made transparent by a parallel elision at Aen. 6.509 ad quae Priamides: ‘nihil o tibi, amice, relictum’,Footnote ⁷⁷ whereas the two indisputable occurrences of te in of poem 55 (4 te in Circo; 5 te in templo), as well as those at 31.6 (te in tuto) and at 64.149 (te in medio), have a metrical value that conflicts with the pyrrhic scansion required by the prosodic hiatus. Such a disharmony proves all the more implausible since, in poems 55 and 58b (both about the elusive Camerius), Catullus takes the experimental liberty of freely contracting the pyrrhic section of the standard phalaecian line.Footnote ⁷⁸ Notice, to conclude on this technical point, that 58b.8 defessus tamen omnibus medullis, while recalling 55.4, conforms to classical norms and that, at 58b.10 essem te mihi, amice, quaeritando, Catullus resorts to the same prosodic trick as Virgil at Aen. 6.509.Footnote ⁷⁹

In v. 4, libellis should be corrected into tabernis (ς, probably under the influence of Hor. Sat. 1.4.71); contrary to what is assumed by HandFootnote ⁸⁰ and Trappes-Lomax, this word does not refer here to bookshops or inns, but to stalls in the Circus (see Oksala, who quotes Cic. Mur. 73 quid statuent in uiros primarios qui in Circo totas tabernas tribulium causa compararunt?). It thus turns out that Catullus enumerates three places, one of which should be described by means of a conjoined phrase.Footnote ⁸¹ I propose to edit teque omnibus (which improves on the unmetrical te que in omnibus, ς). At the same time, I retain Mynors's text in v. 3 (where many editors, following Sillig, insert V's in after te), so as to create a regular alternation between the absence and presence of the preposition; notice that, in v. 14, in is superfluous and unidiomatic before fastu (Schuster, Heyworth, Trappes-Lomax). As frequently happens, te et substituted for teque; next, the second et was read te and dropped accordingly.Footnote ⁸² This gave rise to a metrical anomaly; someone added a marginal in that was subsequently introduced at two different places in vv. 3–4, and perhaps also, contagiously, in v. 14. Trappes-Lomax writes that ‘libellis might be no more than a metrical filler’; but tabernis was more probably altered into †tabellis†, subsequently replaced with †libellis†.

Among the solutions proposed for making sense out of v. 9, one should first reject ah uel te (ς) and all similar corrections that postulate an otherwise unattested use of flagito with an accusative designating the person looked for; at Cic. Clu. 83 cur … Staienum … non requirebant? Oppianicus querebatur; Quinctius flagitabat, the implicit objects of querebatur and flagitabat do not refer to Staienus himself, but to his absence from, and presence at the court, respectively. Like ah uel te, etc. or at certe,Footnote ⁸³ Thomson's audite en (audite, ς) does not provide vv. 9–10 with an explicit verbal predicate; moreover, like Harrison's efferte en, it creates an unwelcome repeated use of the interjection in contiguous turns at speech. Camps's a, cette huc avoids both objections, but one can reasonably doubt whether Catullus would use such an archaism.Footnote ⁸⁴ Third-person auellent (Ellis) does not fit in with the dialogical structure of the passage; auelli sinite ipse (Avancius [n. 5]) fares better, but the expression is unnatural and the reduction of sinite to sic palaeographically problematic. Foster's reddatis,Footnote ⁸⁵ while sounding spontaneous, requires uultu (v. 8) to have played a crucial (and rather implausible) role in the corruption process. Goold's aufertis is in the wrong tense (Harrison); auertistis, saepe (Riese) proves definitely preferable in this respect, but saepe is pleonastic (in contrast with ipse, which underlines the fact that, contrary to the norms of Old Comedy, the speaker personally takes in charge a task he should have entrusted to a slave). I thus propose nouistis, sic ipse flagitabam, | Camerium mihi, pessimae puellae? (see Pl. Poen. 591–2 hunc uos lenonem Lycum | nouistis?), which here conveys obvious sexual overtones.Footnote ⁸⁶ Similar uses of a dative substituting for an unmetrical possessive adjective occur at 21.10–11 and 37.11–14.Footnote ⁸⁷ Haplography produced nouisti, which gave †auelte† owing to the confusion between ıo-, with one minim lost, and a- Footnote ⁸⁸ and the usual shifts from i to e, and s to l.

When he conjectured nudum sinum reducens in v. 11, Avancius (in ed. Ald. 1502) took a first step towards what I think is the right solution. But, by maintaining nudum, he also led future critics on a wrong track. In spite of the fact that, in its anatomical acceptation, sinus always refers to the woman's womb, a major trend in Catullus scholarship, influenced by v. 12 and by the meaning of the Italian or French reflexes of the word, enduringly entertained the absurd notion that the girl might uncover her breast (an intolerably provocative gesture in Roman society).Footnote ⁸⁹ Since reducens did not make sense under such an interpretation, it was replaced with the palaeographically close recludens (Riese).Footnote ⁹⁰ Almost simultaneously, pectus (Ellis) was substituted for sinum; this semantically sound move did not leave any metrical slot for reducens/recludens: nudum reducta/reduce pectus (Ellis, Lenchantin de Gubernatis) restored metre by turning back to previous nonsense (pace Housman, 2.623); nudum reclusa/reclude pectus (Riese dubitanter, Friedrich) produced new nonsense in that elsewhere pectus recludere means ‘open the breast’ (Fordyce).Footnote ⁹¹ If we opt for the referentially more plausible hypothesis that the girl shows her cleavage, niueum retecta pectus (Harrison) looks like the best choice (see Prop. 2.22.8: candida non tecto pectore si qua sedet). But a simpler solution consists in assuming that sinum designates the girl's dress (as in 61.52–3, see n. 89): by tying her girdle (reducens) or having it tied (reducta) under her bosom, she makes the dress cling to her breast (an erotic but socially more tolerable posture that also motivates latet in v. 12). As a consequence, we have to emend (away) nudum. Some of the corrections available saturate the text with occurrences of the same interjection: ‘en’, inquit quaedam, sinu/-um reducto/-ens (Goold), quaedam ‘en, en’ inquit sinum reducens (KokoszkiewiczFootnote ⁹² ). Froehlich's tunicae sinum and Hermes' (dubitanter) dextram ad sinum Footnote ⁹³ do not run into this objection, but can account for the paradosis only with difficulty. I suggest reading nodo sinum reducta. In conformity with attested usage (see, for instance, Verg. Aen. 6.301, 11.775–6 tum croceam chlamydemque sinusque crepantes | carbaseos fuluo in nodum collegerat auro; Curt. 6.5.27; Sen. Herc. F. 766–7; Stat. Theb. 4.266), nodo refers to the knot that fastens the girdle.Footnote ⁹⁴ My preference for reducta against the equally acceptable reducens stems from the fact that Catullus has the same use of a past participle with an accusative and an ablative at 64.64–5 non contecta leui uelatum pectus amictu, | non tereti strophio lactentes uincta papillas, 64.122 deuinctam lumina somno and 64.296 restrictus membra catena; at Aen. 1.320 nuda genu nodoque sinus collecta fluentes, Virgil probably imitates our line as well as Laev. 12a et reuocet uolucres in pectore balteus sinus (see Courtney, FLP ad loc.). At the first stage of the corruption process, someone jumped from nod- to -um, all the more easily since the mistake involved two bisyllabic words and the resulting text remained meaningful (‘with the knot tied’); next, †nodum† was altered into †nudum† (see †nudus† for nodus at Prop. 2.29.10 [Heyworth]); finally, a scribe stopped writing after reduc- because he did not understand the construction anymore.

Verse 157 has been corrected in multifarious ways (see the Catullus Online website for more details). One reason for this emendatory abundance lies in the fact that the line exhibits a rare metrical profile, with an unelided polysyllable ending at the ennehemimeral position before an unelided pyrrhic word.Footnote ⁹⁵ Catullus elsewhere has two such hexameters, at 64.58 and 98.3 (notice also the pyrrhic words datur, Themis and tua at the same position in 68.147/153/155); Ennius in Ann. (Vahlen) ten or eleven, at 14, 20, 32, 40, 52, 110, 385, 517, 518, 529 (?), 543 (notice terram in the same position), with a higher frequency in Book 1 (six cases); Lucretius thirteen, at 1.69, 1.155, 1.184, 1.250, 2.184, 2.305, 2.592, 3.874, 4.267, 4.1255, 5.293, 5.1339, 6.1131; Cicero nine, at Arat. 33.174, 33.187, 33.309, 33.325, 33.343, 33.376, 33.454, 33.468 and Homer. 9.1 (Soubiran); Virgil sixteen, at G. 1.80, 2.153, 4.251, Aen. 3.695, 5.731, 8.382, 10.302, 10.400, 10.440, 10.442, 10.471, 10.772, 10.849, 11.143, 11.170, 11.562 (the concentration in the last books of the Aeneid confirming the Ennian archaism of this pattern). The distribution of dactyls and spondees increases the exceptionality of v. 157; but similar examples are found at Lucil. (Marx) 1372 quare da te homini: complectetur, mihi crede; Lucr. 4.1255 uxores parere, inuentast illis quoque compar and 5.1339 ut nunc saepe boues Lucae ferro male mactae; Cic. Arat. 33.309 sed tantum supera terras semper tenet ille and Homer. 9.1 tales sunt hominum mentes, quali pater ipse; Verg. Aen. 10.302 omnes innocuae; sed non puppis tua, Tarchon!; Hor. Sat. 1.3.36 natura aut etiam consuetudo mala namque.

Though dismissed in passing by Fordyce, auctor (ς, Rossbach) for †aufert† is a palmary correction. Indeed, this substantive can designate the founder of a people or lineage (see, for instance, Verg. Aen. 4.365, Stat. Theb. 1.224; TLL 2.1204.30–66), that is the individual whose natural mission consists in giving a land to his descendants, generally after some tribulations on the sea across different parts of the world; see, for a similar use, Ov. Tr. 2.194: ulterior nulli, quam mihi, terra data est (Baehrens, Friedrich and Lenchantin de Gubernatis also mention Pl. Merc. 195–7; Cic. Sen. 71; Ov. Tr. 4.5.5, Pont. 1.2.60, 1.10.39, 2.8.68, 2.9.9). Once we opt for auctor, v. 158 makes perfect sense: for the members of the family, all good things originate from (a quo) their first (primo) common ancestor; see, for instance, Ov. Met. 13.142 nostri quoque sanguinis auctor, Stat. Theb. 2.463 primus sanguinis auctor. But the hiatus primo omnia is an undisputable sign of corruption. Nisbet suggests plurima (or tot mihi) nata bona; Trappes-Lomax, tanta parata bona (he assumes that tãta parata bona was reduced to tata bona owing to the homoeoteleuton; ‘a scribe then altered tata to nata to provide the necessary participle, and filled up the line with omnia’). A simpler solution consists in reading a quo primo sunt (ς, Scaliger, Oksala), but primo sounds pleonastic after principio. I propose a quo patria sunt omnia nata bona. The collocation patria bona frequently occurs in Republican texts: Ter. Eu. 235 hominem haud impurum, itidem patria qui abligurrierat bona; Acc. 164 (Ribbeck, TRF) = Non. 471.21 (Lindsay) qui nostra per uim patria populauit bona; Afran. 50 (Ribbeck, CRF) = Non. 475.21 (Lindsay) cum testamento patria partisset bona; Cic. Verr. 2.1.113 ut liberis eius bona patria—uoluntate patris, iure, legibus tradita—eriperes, 2.2.30, 2.2.59, 2.4.37, S. Rosc. 145; Sall. Cat. 14.2 quicumque … bona patria lacerauerat; see also paterna … bona (29.17), patriam (63.49), patria … creatrix … patria … genetrix (63.50), patria (63.55), patria, bonis (63.59), patriis … aris (64.132).

Even if any such hypothesis must remain highly speculative, two reasons induce me to identify Catullus' auctor with Volesus, the mythical founder of the gens Valeria whom Silius claims to be of Spartan descent (see Livy 1.58.6, 2.18.6, 2.30.5, 3.25.2; Asc. Pis. 12; Sil. 2.8; Juv. 8.182; [Aur. Vict.] De vir. ill. 15.1; Dion. Hal. Ant. Rom. 2.46.3; Plut. Vit. Num. 7). First, Ovid was perhaps recalling our distich when he evoked Volesus in a poem addressed to a member of the gens Valeria: quos Volesus patrii cognoscat nominis auctor (Pont. 3.2.105). Secondly, given the equivalence between the etymological Volesus and its vernacular variant Volusus,Footnote ⁹⁶ I am inclined to believe that Catullus' acrimony against Volusius (or perhaps Volusus, since both forms are compatible with genitive Volusi) and his annals (36, 95.7–8) stemmed from the fact that this bad poet—related, in some way, to the north of Italy (95.7 Paduam … ad ipsam, that is somewhere in the delta of the Po)—belonged (or claimed to belong) to the gens Valeria Footnote ⁹⁷ and perhaps endeavoured to celebrate the deeds of that family (see above, on 36.9–17). In Virgil (Aen. 11.463), the elided tribrachic vocative Voluse occurs just a few lines after two verses that mention the same (or another) branch of the Po and its raucous swans (11.457–8 piscosoue amne Padusae | dant sonitum rauci per stagna loquacia cycni).Footnote ⁹⁸ In Catullus, the diaeresis peruoluent of 95.6 not only creates a wordplay with Volus(i)us' name (Friedrich), but also echoes a famous Ennian line (var. 18 [Vahlen] = Cic. Tusc. 1.34: uolito uiuos per ora uirum) subsequently imitated by Virgil (G. 3.9; Aen. 10.790, 11.296, 12.235), Propertius (2.1.2, 3.1.24, 3.9.32) and Ovid (Met. 15.878–9). This formal and intertextual network underlines the contrast between the talentless annalist and his literary model.

One could find it strange that Catullus might indiscriminately address good wishes to Allius/Manlius, his wife or girlfriend and his house, to Volesus and, in conclusion, to Lesbia herself. But, as shown by the legend of Laodamia and Protesilaus, military expeditions make people abandon, like Catullus' brother, their home and native land, and thus make them inverse the mythical itinerary of their auctor. Similarly, Catullus has left Rome for Verona, and thus inverted the itinerary that led him to literature and love. The chaotic world of poem 68 symbolically reflects the impossibility of any synthesis of those different components: love and poetry will never find a home because experiencing them amounts to constantly alternating between peace and war, or life and death.

The corruption of vv. 157–8 involved two different processes. On the one hand, auctor shifted to †auector†, a non-existent but plainly intelligible word that someone glossed with aufert in the margin; subsequently, †aufert† ousted the graphically similar, but unmetrical, †auector† from the main text.Footnote ⁹⁹ On the other hand, p̃r̃ia = patria was altered into †prĩa = prima† (see the alternation between †primos† and †patrios† at Ciris 45 [Knecht, lxxxi]), which was moved after sunt in order to restore the metre; this produced first the false grammatical agreement †primā … bona†, with hemistich-final lengthening (see above, on 17.3), and subsequently †primo … bono†, which allowed coreference with a quo.

Philologists analyze nihil either as the modifier of a verb omitted in the paradosis (prodest in Puccius's marginalia, see the Catullus Online website for more details; iuuerit Baehrens; iam iuuat Munro) or as the attribute of a copula sentence with est implicit or to be restored (ed. Ald. 1502, Friedrich)—which requires a different addition in v. 3: for instance, immo etiam taedet, taedet obestque magis (ed. Ald. 1502). None of those proposals accounts for the omission of the first foot and for verse-final †magisque magis†. I thus prefer to print nihil fecisse benigne | malis: immo etiam taedet obestque magis (‘You had better not be kind: contrary to what you may expect, it will be tiresome and harmful rather’). For malo (attested at 24.4, 45.22, 70.1, 79.1) governing a perfect infinitive, see Livy 42.11.1 plurium annales, et quibus credidisse malis, ipsum Eumenem uenisse tradunt; Ov. Fast. 6.70–1 blandior et partes paene rogantis ago | remque mei iuris malim tenuisse precando; Plin. HN 28.179 aliqui murinorum capitum cinerem miscuisse malunt, 28.203 aliqui et nitrum et aquam adiecisse malunt. Archaic legal language combined perfect infinitives with nolo or negated uolo in prohibitions, and poets frequently imitated and extended this use (see, for instance, 69.1–2 noli admirari quare tibi femina nulla, | Rufe, uelit tenerum supposuisse femur; Hor. Sat. 1.2.28; Prop. 2.19.32; Ov. Am. 1.4.38); nihil, though to be construed with fecisse here, surely favoured the analogy (Kühner-Stegmann, 1.133–5; Hofmann-Leumann-Szantyr, 2.351–3). Moreover, the verse-enclosing pair malis … magis creates an easily accessible etymological wordplay (Cic. Orat. 154 ‘malle’ pro ‘magis uelle’ … dicimus). Translators (for instance, Goold, Lee) usually assume that immo (etiam) requires there to be an opposition between the two sentences it connects; but this particle may introduce an utterance that corroborates a previous speech act when the justification provided contradicts common expectations (see, for instance, Petr. 43.7–8, about a septuagenarian: et adhuc salax erat. non mehercules illum puto domo canem reliquisse. immo etiam puellarius erat, omnis Mineruae homo; ‘contrary to what you may expect of such an old man’). For the use of immo (uero) with magis meaning ‘rather’, see Cic. Mil. 34 ‘obstabat in spe consulatus Miloni Clodius’. at eo repugnante fiebat, immo uero eo fiebat magis. Owing to the confusion between capital g and c , hence l, initial malis was read †magis†.Footnote ¹⁰⁰ Then, a scribe who attempted to restore the metre moved †magis† to the end of the line so as to fabricate an additional dactyl; see the manuscript reading †immo etiam taedet stetque magisque magis† (ς), probably derived from †immo etiam taedēt estque magisque magis† with lengthening in hemistich-final position.

In v. 1, the collocation of the adjectival and participial forms creates a hiatus and a pleonasm. By editing si quoi quid cupido atque optanti contigit umquam (not recorded in the Catullus Online website; quoi quid Ribbeck;Footnote ¹⁰¹ contigit Puccius, marginalia, see the Catullus Online website for more details), Trappes-Lomax eliminates the hiatus, but not the pleonasm (compare with v. 5); in addition, optigit is undoubtedly idiomatic in this use (TLL s.v.) and one may want to conserve the enriched alliteration opt- … opt … . In Froehlich's si quicquid quoiquam cupide optanti obtigit unquam, where quicquam (ς) should substitute for quicquid, the adverb cupide, also attested at 63.2 and 64.267, slightly mitigates the pleonasm; but the repetition of cupido in vv. 4–5 suggests that this epithet must be maintained here. I propose si cuiquam cupido optatum quicquam obtigit umquam. This emendation combines two metrical idiosyncrasies. First, it exhibits, like v. 5, the elision of a polysyllabic word at the ennehemimeral position (Soubiran [n. 15], 536–7). We find sixteen such lines in other poems (62.50, 64.44, 64.56, 64.180, 64.184, 64.248, 64.255, 64.322, 64.363, 68.89, 71.1, 80.1, 81.1, 83.5, 91.5, 99.3); in four cases, the fitfh foot begins with a verbal prefix, as happens here (62.50 educat, 64.56 excita, 64.255 inflectentes, 71.1 obstitit). Secondly, the lack of a central (penthemimeral or trochaic) caesura is compensated by two breaks at the trihemimeral and hepthemimeral position. This pattern occurs at 64.18, 64.128, 64.193, 64.196, 64.248 and 108.1; in all cases except 64.193, metrical prodelision (erasure of the second vowel in contact) and/or morphological division after a (pseudo-)prefix would permit a penthemimeral caesura, as happens here;Footnote ¹⁰² 64.248 obtulerat mente immemori, talem ipse recepti shows the same combination of both features, while the strict parallelism between 107.1 and 108.1 si, Comini, populi arbitrio tua cana senectus rules out Trappes-Lomax's hypothesis that 108.1 might be metrically corrupt. Repeated scribal confusions between qu- words or morphemes (see above, on 29.4 and 36.12) and case contagion in †optanti† account for the paradosis.

In v. 3, the lexical meaning of quoque conflicts with the unequal force of the two predications (Fordyce); Avancius's et carius auro does not really improve on the paradosis. Maehly's quare hoc est nobis quoque gratum et carius auro,Footnote ¹⁰³ not recorded in the Catullus Online website and advocated by Trappes-Lomax, solves the problem but confers an oddly syllogistic flavour to vv. 1–4. I suggest editing quare hoc est gratum nobis, hoc carius auro (Froehlich); the anaphora allows vv. 3–6 to form one single sentence closed by a climactic exclamation. By writing †quoque†, a scribe expressed the semantics of the two occurrences of hoc without capturing the rhetorical impact of the structure.

In vv. 7–8, I suggest reading ac te | optandam esse unam (ac me Schwabe in apparatu; optandam uitam Puccius, marginalia; see the Catullus Online website for more details). For this use of (prae)opto in an erotic or epithalamic context, see 62.30, 62.42, 62.44, 64.22, 64.31, 64.120, 64.141, 64.328, 64.372, 66.79; Verg. Aen. 11.270; Ov. Her. 16.36, 16.173, 17.101, 17.111, Met. 8.324–5, 10.622, Fast. 1.417 (TLL 9.2.829.17–41); for the construction of the comparative, see 22.16 and 61.169 illi non minus ac tibi. The substitution of unam for uita not only eliminates a reading that, in my view, has led most scholars on the wrong track (Syndikus, 3.118–19; see the Catullus Online website for more details),Footnote ¹⁰⁴ but also puts both lovers on a par: the speaker's happiness is unique and no woman is so uniquely desirable to any man as Lesbia (see 45.14, 45.21, 45.23, 58.2, 68.135, 68.147). For a similar collocation, see Cic. Att. 11.19.1 est autem unum quod mihi sit optandum, si quid agi de pace possit, Fam. 2.6.4 qui si ex omnibus unus optandus esset, quem tecum conferre possemus non haberemus; Stat. Theb. 2.689 huic una fides optanda labori. In v. 7, te was replaced with †et†, hence †est†; scribes attempted to make sense out of the paradosis by writing †hac† or by repeating ablative †me†.Footnote ¹⁰⁵ In v. 8, optandã esse reduced to optandã est (probably owing to a confusion between the ẽ and ẽe abbreviations) and thus shifted to †optandus†;Footnote ¹⁰⁶ a trivial mistake on a series of minims altered unã into †uita†.

As pointed out by Kroll, the metrical anomaly in v. 6 cannot be justified by iambic shortening or prosodic hiatus operating on an underlying modō: this adverbial form does not occur in Catullus (modo scurra at 22.12 is no example of it, pace FordyceFootnote ¹⁰⁷ ) and ipse is no iambic word (see above, on 10.27). Most of the emendations proposed are equally unmetrical (dum domŏ/domĭ/bonŏ ipse Lachmann dubitanter, Ellis dubitanter, Riese), unidiomatic (for instance, dum tamen, dum modo et ς; dum modio RichmondFootnote ¹⁰⁸ ), or contextually inadequate (for instance, dummodo te ipse egeas Froehlich, dum domino Peiper dubitanter Footnote ¹⁰⁹ ). Trappes-Lomax's dum omnibus (not recorded in the Catullus Online website) creates a formal redundancy with v. 5 and requires a complex palaeographical drift. I suggest editing dum modo ope ipse egeat. Catullus uses ablative ope at 34.24 and 67.2; the graphical similarity of ope and ipse led a scribe to drop the first word.Footnote ¹¹⁰