* I thank the journal's referees for their helpful comments. Abbreviations for ancient and modern works follow the journal's normal style, with the following addenda:

¹ O'Sullivan, N., ‘Manuscript evidence for alphabet-switching in the works of Cicero: common nouns and adjectives’, CQ 68 (2018), 498–516CrossRefGoogle Scholar.

² An issue discussed was the occasional reported disagreement amongst the manuscripts themselves on the alphabet used for Greek words in particular passages (O'Sullivan [n. 1], 499–500), but that problem does not arise in the current study, as our editions do not indicate alphabetic variation in the manuscripts of the individual passages discussed here.

³ O'Sullivan (n. 1), 513–15.

⁴ So we do not discuss, for example, κατὰ Διόδωρον and κατὰ Χρύσιππον in Fam. 9.4. Cicero sometimes alludes to proverbs with single Greek words (e.g. Att. 1.20.3 Σπάρταν, Att. 8.11.3 malorum … Ἰλιάς: see SB ad locc.), and these have also been omitted from the list. On the other hand, the use of an original Greek title of a book did not require Greek presentation of the author's name also (see n. 11 below), and so such presentation can be seen as a deliberate choice.

⁵ E.g. δόγμα in Luc. 27 and 29, πολιτικός in Brut. 265 and Fin. 4.5 and 5.66.

⁶ The statistics are based on searches on the Packard Humanities Institute's Classical Latin Texts (http://latin.packhum.org). Some Latin forms are actually a little more frequent, but obvious homonyms are excluded from the totals, e.g. Plato the Epicurean, an otherwise unattested contemporary of Cicero (QFr. 1.2.14).

⁷ Adams, J.N., Bilingualism and the Latin Language (Cambridge, 2003), 342CrossRefGoogle Scholar. The connection is actually closer than Adams demonstrates, as we have inscriptions on bases recovered from Rome with sculptors’ names in the genitive: e.g. Λυσίππου ἔργον, Ἀπόλλων Μύρωνος: IGUR 1574, 1578. See further IGUR 1558, 1571, 1572, 1579, 1580, 1583, 1638. There are interesting recent discussions of such ‘signatures’ (or rather ‘attributions’ in many cases) by Squire, M., ‘Ars in their “I”s: authority and authorship in Graeco-Roman visual culture’, in Marmodoro, A. and Hill, J. (edd.), The Author's Voice in Classical and Late Antiquity (Oxford, 2013), 357–414Google Scholar and Hurwit, J., Artists and Signatures in Ancient Greece (Cambridge, 2015)CrossRefGoogle Scholar, especially ch. 10 on sculpture.

⁸ The context of the passage shows that the vividness has a rhetorical purpose. Cicero's argument depends on the inscribed Latin text, which he quotes in concise epigraphic form (‘cos.’ according to restorations by Malaespina and Purser), of two Roman statues near this Greek one; he appears to extend this exact reporting to the Greek inscription as well.

⁹ E.g. Cicero ‘portrays himself as possessed of certain forms of expertise and culture, and thus claims possession of the same cultural trappings as Atticus’ (Adams [n. 7], 318). Clearly this is not the place to address the larger issue of the applicability to Cicero and Atticus of the sociological linguistics underpinning much of Adams's book. This is a question which I hope to discuss elsewhere. Here I simply claim that Adams's insight into the underlying statue inscription offers a better explanation of the switch of alphabets in mentioning literary works than does a supposed desire to parade learning: as Philostratus (Dial. 1) succinctly said, φιλοτιμία ἐν ἐπιστολῇ μειρακιῶδες.

¹⁰ On Atticus as editor and publisher, see the overview of Buckley, M., ‘Atticus, man of letters, revisited’, in Sidwell, K. (ed.), Pleiades Setting: Essays for Pat Cronin on his 65th Birthday (Cork, 2002), 14–32, at 24–31Google Scholar.

¹¹ Note that, even when giving the title in Greek, Cicero sometimes refers to the author by his Latin name, as in Att. 2.1.8 dicit enim tamquam in Platonis πολιτείᾳ and Att. 13.32.2 Dicaearchi περὶ Ψυχῆς.

¹² The thorough monograph of Caroli, M., Il titolo iniziale nel rotolo librario Greco-Egizio (Bari, 2007)Google Scholar has superseded earlier studies on this subject, and his catalogue of testimonia (T) and papyrus fragments (P) is used in the following discussion.

¹³ Caroli (n. 12), 23–8.

¹⁴ A red-figure cyathus of 500–490 b.c. (T 2 = Tav. iv a and c Caroli).

¹⁵ P.Oxy. 2741 (P 7 Caroli = TM 59787).

¹⁶ Thus Caroli (n. 12), 28–40, accepting two possible forms for the word as current in Cicero's time, argues against Dorandi, T., ‘Sillyboi’, Scrittura e Civiltà 8 (1984), 185–99Google Scholar, who preferred the masculine forms σίλλυβος/σίττυβος. Elucidation depends heavily on evidence later than Cicero: the word, probably in different forms, seems to occur three times in the letters (see below), but in each case there are significant textual problems, and nowhere do the manuscripts preserve any of the forms in Greek characters.

¹⁷ Caroli (n. 12), 63.

¹⁸ Diog. Laert. 10.25.

¹⁹ Aly, Z., ‘Title of a lost play of Menander from Oxyrhynchus’, Études de Papyrologie 8 (1957), 163–7Google Scholar (= TM 61592).

²⁰ Caroli (n. 12), 66; the ancient justification was that the characters in a drama act and speak with apparent autonomy.

²¹ The size of Pindar's corpus means that this could not have been the label for just one roll, so it has been suggested that it belonged to a collection, perhaps to the capsa which contained it (Caroli [n. 12], 204): on these Rollenkasten, which are pictured with their own labels, see Birt, T., Die Buchrolle in der Kunst. Archäologisch-antiquarische Untersuchungen zum antiken Buchwesen (Leipzig, 1907), 248–55, 259–61Google Scholar. Exactly which sort of book label such a nominative represents does not matter for my purposes. Possibly the (now illegible) Εὔτυχος Χοριαμβικά written on the back of a roll in T 4 Caroli, a wall-painting from Herculaneum, may have represented a nominative and a title on a real book.

²² The one exception is Συνδείπνους Σοφοκλέους, where the reference to a physical book is not explicit and where the ipsissima uerba have, in accordance with general practice in Greek and Latin, been made to fit the grammar of the surrounding sentence: see e.g. Fraenkel, E. (ed.), Aeschylus Agamemnon (Oxford, 1950), 2.29Google Scholar (on Ag. 48) and Austin, R.G. (ed.), P. Vergili Maronis Aeneidos liber secundus (Oxford, 1964), 277Google Scholar (on Aen. 2.769).

²³ Our medieval codices of Plato differ slightly: τὰς δὲ τῶν τυράννων δεήσεις ἴσμεν ὅτι μεμειγμέναι ἀνάγκαις εἰσίν.

²⁴ ‘Semi-serious’ is SB's interpretation ad loc.

²⁵ Att. 9.18 of 28 March relates their meeting at Formiae.

²⁶ Att. 9.10.2 (18 March).

²⁷ For the nominative of the author on book labels, see above. Behrendt, A., Mit Zitaten kommunizieren: Untersuchungen zur Zitierweise in der Korrespondenz des Marcus Tullius Cicero (Rahden, 2013)Google Scholar shows how rarely Cicero quotes from Greek prose in his letters, for it is in the nature of prose to be less memorable and thus more likely to require reference to a written text for the purposes of quotation; cf. Steele, R.B., ‘The Greek in Cicero's epistles’, AJPh 21 (1900), 387–410, at 400Google Scholar.

²⁸ See n. 11 above.

²⁹ Cf. Dunkel, G., ‘Remarks on code-switching in Cicero's letters to Atticus’, MH 57 (2000), 122–9, at 123Google Scholar.

³⁰ Cf. Orat. 234 from two years earlier, using the same imagery to make the same argument.

³¹ Admittedly restored here: see SB's apparatus criticus.

³² See Rawson, E., ‘Cicero and the Areopagus’, Athenaeum 63 (1985), 44–66, at 48–9Google Scholar for the dating issue.

³³ SB ad loc. regards identification as reasonable, but would Cicero have inflicted on his own son a teacher of whom he clearly had such a low opinion?

³⁴ On Atticus’ enthusiasm for Dicaearchus, see SB ad loc.

³⁵ For a discussion of these works—and of the suggestion that the ‘Aristotelian’ Ἀθηναίων πολιτεία may be referred to here—see Huby, P.M., ‘The controversia between Dicaearchus and Theophrastus about the best life’, in Fortenbaugh, W.W. and Schütrumpf, E. (edd.), Dicaearchus of Messana. Text, Translation, and Discussion (New Brunswick and London, 2001), 311–28, at 324–8Google Scholar.

³⁶ Att. 13.12.3 (at 13.16.1 the full phrase Ἀκαδημικὴ σύνταξις occurs), in contrast with the frequent Academicus, first attested in Cicero. But the Greek form of the adjective is usually Ἀκαδημαϊκός, and it has long been noticed (Steele [n. 27], 406) that Cicero's Ἀκαδημικός appears to be a back-formation from his Latin.

³⁷ Att. 13.19.4, 13.12.3; QFr. 2.13.1. On πολιτικός/politicus, see O'Sullivan (n. 1), 510–12. Caelius, who hardly ever uses Greek in his letters to Cicero and usually transliterates it when he does according to the manuscripts (embaenetica in Fam. 8.1.4, didascalia in Fam. 8.3.3), Latinizes Cicero's Πολιτικά into politici libri in Fam. 8.1.4.

³⁸ Cf. Att. 15.13.3 on Varro's (Latin) Ἡρακλείδειον.

³⁹ So Adams (n. 7), 342: ‘stressing the essential Greekness of such a treatise’.

⁴⁰ Amalthea or the Amaltheum on Atticus’ property is mentioned in Att. 1.13.1, 1.16.15, 2.20.2 and Leg. 2.7. It seems to have been a decorated garden rather than a building (so F.G. Moore, ‘Cicero's Amaltheum’, CPh 1 [1906], 121–6). Cicero plans to build his own at Att. 1.16.18, and seems to allude to its existence at Att. 2.1.11 and 2.7.5.

⁴¹ Alternatively, Atticus might have already written to Cicero in Greek mentioning the Amaltheum, and Cicero is quoting Atticus’ Greek back to him, as he sometimes does (on the whole subject, see now N. O'Sullivan, ‘In search of Atticus’ Greek’, JHS 139 [2019], 94–109). Cicero's is the first surviving Roman mention of Amalthea, and his desire for more information on her is understandable, given that the ancients could not even agree whether she was a goat or a nymph (see LIMC s.v.).

⁴² Brut. 38, 59; Att. 12.6a.1; admittedly that letter is a special case which corrects a passage in the original text of Orat. 29.

⁴³ On this legend of Eupolis’ death, see Storey, I.C., Eupolis Poet of Old Comedy (Oxford, 2003), 101–3CrossRefGoogle Scholar.

⁴⁴ As reported by Pease and Pohlenz respectively.

⁴⁵ Some take cum here as causal, but the adversative sense brings out the contradiction in the passage better: Cadmus was only a mortal (the explicit point of Hes. Theog. 975; cf. 967), but he was still the father of a goddess, while the children of the Sun himself were not regarded as divine.

⁴⁶ The final passage, like the first here, is explicit about her transformation into a divinity, and presents her in Latin characters: multos habent ex hominibus deos … Leucotheam (quae fuit Ino) et eius Palaemonem filium cuncta Graecia (Nat. D. 3.39).

⁴⁷ Interestingly, our sole source for this part of the Republic, the famous Vatican palimpsest, still shows the original reading, not the alteration (in contrast with Orat. 29, which now contains a correction Cicero requested at Att. 12.6a.1); for a discussion of the thinking behind the change in the Republic, see Gurd, S., Work in Progress: Literary Revision as Social Performance in Ancient Rome (Oxford, 2012), 55–6Google Scholar. The Greeks themselves used both Φλιούντιος and Φλιάσιος (Hdn. Gr. 2.882 Lentz; cf. Diod. Sic. 15.19.3).

⁴⁸ Tusc. 5.8; Ver. 2.2.109.

⁴⁹ Arch. 24 describes him as scriptor rerum; cf. FGrHist 188.

⁵⁰ So elsewhere does Cicero cast a negative light on his ethnic background: Att. 5.11.3 and 9.11.3. It has long been noted that Cicero's limitless admiration for Greeks of an earlier time is not matched by his respect for contemporaries (see e.g. H. Guite, ‘Cicero's attitude to the Greeks’, G&R 9 [1962], 142–59).

⁵¹ BNP s.v. Theophanes 1.

⁵² All other instances (including Fam. 15.16.1) are Latinized in the manuscripts.

⁵³ Syme, R., The Roman Revolution (Oxford, 1939), 45Google Scholar. Cicero is assuring Pulcher that his friendship is not based on self-interest (as Epicurus supposedly held about friendship in general). The recipient can only have been hostile to Epicurus: on his religiosity—which extended to writing a treatise on augural law dedicated to Cicero—see RE 3.2853.

⁵⁴ SB on Att. 1.12.1.

⁵⁵ Most clearly and extensively in Att. 6.4 and 6.5 (cf. 6.7.1), where again the subject matter is financial, but elsewhere too a Greek phrase occurs apparently to hide a confidential observation: e.g. Att. 2.17.1. On Greek for security reasons in the correspondence, see Adams (n. 7), 329–30 and Nicholson, J., ‘The delivery and confidentiality of Cicero's letters’, CJ 90 (1994), 33–63, at 43–7Google Scholar.

⁵⁶ E.g. Butler, S., ‘Cicero's capita’, in Jansen, L. (ed.), The Roman Paratext: Frame, Texts, Readers (Oxford, 2014), 73–111CrossRefGoogle Scholar and, more generally, R. McCutcheon, ‘An archaeology of Cicero's Letters: a study of Late Republican textual culture’ (Diss., University of Toronto, 2013).

⁵⁷ Dunkel (n. 29), 123 cites Att. 4.17.1, 5.19.1, 10.17.2, 11.24.2 (and note SB on redeo here: ‘perhaps simply as implying that this is the natural way to write a letter, from which dictation is a departure’). The collection of Cicero's Greek words made by Rose, H.J., ‘The Greek of Cicero’, JHS 41 (1921), 91–116CrossRefGoogle Scholar is still the fullest we have and shows that around 700 of his somewhat more than 900 Greek words are from the letters to Atticus.

⁵⁸ Att. 6.9.1; contrast 7.2.3 and 16.15.1, in which a secretary's hand is commented on, and in the former taken as a sign of Atticus’ ill-health.

⁵⁹ See the examples collected by McDonnell, M., ‘Writing, copying, and autograph manuscripts in ancient Rome’, CQ 46 (1996), 469–91, at 474–5CrossRefGoogle Scholar, involving letters to Quintus as well as to Atticus. But we must not overstate our awareness of which letters were dictated: in at least two cases we know that dictation was being used only because Cicero unapologetically draws attention to his own handwriting beginning at a certain point (Att. 11.24.2, 13.28.4).

⁶⁰ But the practice was not restricted to close friends: Cicero wrote sua manu to Appius Pulcher (Fam. 3.6.2) and received such letters from Trebatius (Fam. 7.18.2) and from Decimus Brutus (Fam. 11.23.2), although the latter may just refer to an autograph subscript, as in the examples cited by McDonnell (n. 59), 474 n. 25.

⁶¹ Tiro is expected to help them decipher the script, some of which Cicero himself found hard to read! The work in question is not certain, but SB ad loc. argues that it was the lost Laus Catonis.

⁶² McDonnell (n. 59), 473, 476.

⁶³ Cf. Elder, O. and Mullen, A., The Language of Roman Letters: Bilingual Epistolography from Cicero to Fronto (Cambridge, 2019), 121 n. 29CrossRefGoogle Scholar.

⁶⁴ QFr. 2.16, 3.3; Att. 4.16, 5.14, 5.17, 13.28.

⁶⁵ So Cicero does not hesitate to use Greek in his letters to Tiro, using words not just from medicine (Fam. 16.18.1) but also from scholarship (Fam. 16.10.2).

⁶⁶ I exclude the thoroughly Latinized rhetor in QFr. 3.3.4 (see O'Sullivan [n. 1], 509) and philologus from the autograph postscript to Att. 13.28.

⁶⁷ Att. 4.16.3, 4.16.6, 4.16.2. On the words, see O'Sullivan (n. 1), 502, 504 and 510.

⁶⁸ These would be the exact forms of the Latinized words, with the ablative politia corresponding to the locative dative of the Greek (in πολιτείᾳ) as preserved.

⁶⁹ O'Sullivan (n. 1), 502–3, 510–11.

⁷⁰ A Latin genitive in -us for a Greek personal noun ending in -ης has left no trace in the extensive collection of Neue, F., Formenlehre der lateinischen Sprache (Leipzig, 1892–1905³)Google Scholar.