¹ Adams, J.N., Bilingualism and the Latin Language (Cambridge, 2003), 71CrossRefGoogle Scholar.

² Swain, S.C.R., ‘Bilingualism in Cicero? The evidence of code-switching’, in Adams, J.N., Janse, M. and Swain, S.C.R. (edd.), Bilingualism in Ancient Society: Language Contact and the Written Text (Oxford, 2002), 128–67CrossRefGoogle Scholar, at 156–7, and Dubuisson, M., ‘Le grec da la correspondance de Cicéron: questions préliminaires sur un cas de bilinguisme’, La Linguistique 41 (2005), 69–86Google Scholar, at 71–2, both mention it briefly.

³ Adams (n. 1), 26 discusses some of the distinctions made between using a different language (code-switch) and using a word derived from it but now considered as part of the adopting language (loanword).

⁴ See e.g. Pelttari, A., ‘Approaches to the writing of Greek in late antique Latin texts’, GRBS 51 (2011), 461–82Google Scholar, who focussed on late antique texts, but suggested that ‘a full study of the manuscripts of, say, Cicero's letters could also yield interesting results’ (471 n. 32). This paper, which was begun before I was aware of Pelttari's suggestion, takes in all of Cicero's works, but restricts itself to those Greek words which appear in both alphabets in the corpus. The fullest collection of Cicero's Greek words is still Rose, H.J., ‘The Greek of Cicero’, JHS 41 (1921), 91–116CrossRefGoogle Scholar, which, despite its omissions, contains over nine-hundred words.

⁵ I adopt as a general if undogmatic principle in this study that, where there is disagreement, the older MSS are to be preferred, while accepting the principle's limitations: see Reynolds, L.D. and Wilson, N.G., Scribes and Scholars: A Guide to the Transmission of Greek and Latin Literature (Oxford, 2013 ⁴), 217–19Google Scholar.

⁶ Moretti, L. (ed.), Inscriptiones Graecae Urbis Romae (Rome, 1968–1990)Google Scholar, no. 707 (= CIL 6.2210); the fine monumental script can be appreciated through the photograph in Walser, G., Römische Inschrift-Kunst (Stuttgart, 1988), 169Google Scholar. Other inscriptions from Rome to combine Latin with Greek words written as Greek include IGUR 298, 397, 460, 501, 634, 638, 972 and 974; 308 presents the alternative approach of transcribing Greek words into the Latin alphabet.

⁷ paedagogus is already in Plaut. Bacch. 138, etc. TLL gives no indication that καθηγητής was ever admitted into Latin.

⁸ Pelttari (n. 4), 470 with further bibliography; see also Reynolds and Wilson (n. 5), 120–2, 272–3.

⁹ e.g. Loew, R., Quaestiones de Graecorum verborum, quae in epistulis Ciceronis exstant, fontibus, usu, condicionibus (Basle, 1889), 68–9Google Scholar; Tucker, T.G., ‘Emendations in Cicero's epistles’, Hermathena 15 (1909), 279–302Google Scholar, at 279–85; Rose, H.J., ‘Some difficulties in the letters of Cicero’, CR 34 (1920), 21–2Google Scholar; and Shackleton Bailey on Att. 1.16.12 for one instance. On the corruption of Greek words in other Latin texts, see Havet, L., Manuel de critique verbale appliquée aux textes latins (Paris, 1911)Google Scholar, §§873–5.

¹⁰ Kaczynski, B.M., Greek in the Carolingian Age. The St Gall Manuscripts (Cambridge, MA, 1988), 33Google Scholar.

¹¹ In several passages of Cicero's letters, late or inferior MSS show in Greek letters words which the stronger transmission has as Latin transliterations: πολιτείας for politias in Fam. 9.2.5; τυραννοκτόνοι/-οις for tyrannoctoni/-is in Fam. 12.22.2 and Att. 14.15.1; κωμικός (a late marginal correction in a single MS) for comicos at Fam. 2.13.2. For these cases, cf. Pelttari (n. 4), 470 n. 30 on a parallel in the text of Servius: ‘the question is whether it is more likely that a single manuscript preserved the original text or that a single copyist transcribed the word into Greek capitals.’ Admittedly, better-attested or even universal readings may also be unreliable (see Pelttari [n. 4], 469–71 and P.K. Marshall, ‘Gelliana Graeca’, CQ 10 [1960], 179–80, at 179), but one of the aims of this paper is to suggest a test of the reliability of alphabetic choices in Cicero's transmission.

¹² The following Greek words are transliterated in all or at least the oldest MSS: ἁρμονία, διαλεκτικός, πραγματικός, σίλλυβος, σοφία, στρατήγημα, σφαῖρα, τοπικός, φυσιολογία, φυσικός, ψευδόμενος. ἐποχή usually appears as Greek and only at Luc. 148 is it universally Latinized, but the passage is corrupt and not discussed here.

¹³ For a comprehensive overview of Cicero's textual transmission, see Reynolds, L.D. (ed.), Texts and Transmission. A Survey of the Latin Classics (Oxford, 1983), 54–142Google Scholar; a list of his works preserved in ninth-century MSS is found on xxviii n. 105.

¹⁴ The Carolingian minuscule Laurentianus 49.9 (= M); see SB, Fam. vol. 1.3–20, Reynolds (n. 13), 138.

¹⁵ For discussion of some aspects see Kaczynski (n. 10), Herren, M.W. and Brown, S.A. (edd.), The Sacred Nectar of the Greeks. The Study of Greek in the West in the Early Middle Ages (London, 1998)Google Scholar and Berschin, W., Greek Letters and the Latin Middle Ages (Washington, 1988)Google Scholar.

¹⁶ Kaczynski (n. 10), 29.

¹⁷ Cf. Swain (n. 2), 157, who does not distinguish between Greek words in isolation and in phrases.

¹⁸ Plaut. Men. 248, etc.; Cato, hist. 83; Cic. Inv. rhet. 1.27, etc. The total figure for Cicero here, as with his other usages studied in this paper, is taken from a search on the Packard Humanities Institute's Classical Latin Texts (http://latin.packhum.org).

¹⁹ Cf. Plut. Cic. 5.2 and C.B. Bishop, ‘Greek scholarship and interpretation in the works of Cicero’ (Diss., University of Pennsylvania, 2011).

²⁰ e.g. schol. Aeschin. 2.31, schol. Ar. Lys. 785, schol. Eur. Andr. 107; cf. Sopat. Rh. 5.72.17 Walz, Alex. Aphr. in Top. 1.72.13 Wallies.

²¹ ἐν ἐπιτομῇ (Att. 5.20.1) vs epitome (Att. 12.5b [bis], 13.8); ἐπὶ σχολῆς (Att. 2.5.3) vs schola (twenty-one times: see n. 18 above); τοῦ ποιητοῦ (Fam. 9.10.1) vs poeta (more than two-hundred times: see n. 18 above).

²² Att. 6.5.2 (of Rome) vs Leg. 2.5 (already Latinized, but with reference to Athens, in Ter. Eun. 987).

²³ Att. 6.5.2 (ter s.v.l.) vs Tusc. 5.91, Leg. 2.68, etc.

²⁴ Att. 9.4.2 vs Nat. D. 1.18, 1.20, 1.22, 2.58, 2.73, 2.160; Pease's apparatus criticus on these passages shows that all the MSS have the word transliterated. Att. 13.8.1 has the word in Greek as part of a book title, a special category of Greek words which I plan to discuss elsewhere.

²⁵ Att. 2.17.1, 9.4.2 (bis), 7.11.1 (quotation) vs Att. 14.9.2, 14.14.2, Inv. rhet. 2.144, Tusc. 2.52, Nat. D. 3.84, Div. 1.111, Off. 3.90.

²⁶ Att. 6.4.3 vs Fam. 12.22.2 (the deteriores have Greek—see n. 11 above), Att. 14.6.2, 14.15.1 (one minor MS has Greek), 14.21.3, 16.15.3. The word is applied to Milo in a coded passage of Greek (50 b.c.), and elsewhere to Caesar's assassins in letters written in 44 b.c. It is first recorded in Cicero, and all the near-contemporary uses in Diodorus Siculus’ Book 16 (14.1, 65.6, 65.8) seem to date from after his death (see Stylianou, P.J., A Historical Commentary on Diodorus Siculus: Book 15 [Oxford, 1998], 20–1Google Scholar, a reference I owe to Dr Nathan Leber).

²⁷ Att. 2.12.4 vs Att. 1.18.3, 7.1.1, Fam. 9.26.1, De or. 1.60, etc. and passim. The word was well established in Latin long before Cicero (e.g. Ter. An. 57), and he uses it hundreds of times across his works.

²⁸ Att. 7.2.4 vs De or. 1.217 (bis—the MSS show both occurrences here in Roman letters), etc., Brut. 44, Orat. 15, etc., Acad. 1.6, etc. and passim.

²⁹ Steele, R.B., ‘The Greek in Cicero's epistles’, AJPh 21 (1900), 387–410Google Scholar, at 388–9, for instance, gives a list of words Cicero wrote as Greek but which appear in later authors as ‘fully naturalized Latin words’.

³⁰ Compare the treatments of the word in the first (1909) and the second editions (1989) of the Oxford English Dictionary.

³¹ For the chronology of the letters I have followed SB, who draws largely on the careful scholarship of predecessors. The chronology of the published works is, thanks to allusions in the letters and their own internal references, largely agreed on, and I have followed the dates in OCD ⁴. The fullest treatment of the chronology of Cicero's life and works is Marinone, N., Cronologia ciceroniana (Bologna, 2004²)Google Scholar.

³² ’Attested use in Latin’ in these tables is taken to mean not just written in that alphabet but also in a context which does not immediately explain the word as a Greek one (‘ut Graeci dicunt’ aut sim.). ‘First-attested use’ cannot, of course, guarantee that Cicero actually introduced the word into Latin.

³³ Cf. Laberius, fr. 12 Panayotakis Pythagoream dogmam, perhaps a direct parody of Cicero's introduction of the word: so Panayotakis, C., Decimus Laberius: The Fragments (Cambridge, 2010), 159CrossRefGoogle Scholar, 162–3.

³⁴ As Cicero tells us in Div. 2.1–2. On dating works in this year, see now Sedley, D., ‘Cicero and the Timaeus’, in Schofield, M. (ed.), Aristotle, Plato and Pythagoreanism in the First Century b.c. New Directions for Philosophy (Cambridge, 2013), 187–205Google Scholar, at 189.

³⁵ The transliteration in the published works is confirmed by Pease's apparatus criticus for Nat. D. and Giomini's for Div.

³⁶ Varro, Sat. Men. 409 would certainly predate sympathia in Cicero and probably συμπάθεια as well (Cic. Acad. 1.8 is the key text for dating Varro's work).

³⁷ The evidence for this is overwhelming but not unanimous: Shackleton Bailey reports no variation in his apparatus criticus, but TLL notes that the Latin form is found in two minor MSS and in a correction to a major one.

³⁸ Varro, Sat. Men. 164 nouam haeresim (on the date of which, see n. 36 above) already shows a Latinized inflection established, as does Laberius, fr. 22 Panayotakis ex Cynica haeresi, which presumably has the ablative in -ī of a Latin i-stem noun, rather than the unrelated -ει of the Greek dative (see Sihler, A.L., New Comparative Grammar of Greek and Latin [Oxford, 1995], 314, 317Google Scholar).

³⁹ J.F. McCall, ‘The syntax of Cicero's Greek in his letters’ (Diss., SUNY-Albany, 1980), 14–16; Adams (n. 1), 497–503.

⁴⁰ McCall (n. 39), 15, not acknowledging that this is an emendation. On the conjecture ex Dolabellae ἀριστείᾳ at Att. 14.19.1, see below.

⁴¹ Cicero's other uses of the word at Parad. 2 and Att. 14.14.1 also show the ablative form as haeresi; cf. n. 38 above.

⁴² On the approximate locations of these estates, see SB on Att. 1.6.2.

⁴³ Cf. Adams (n. 1), 551 on the use of Greek verses in a Latin inscription: ‘Greek is the language of literary pretension, in which humdrum information is not conveyed.’

⁴⁴ OLD s.v. palliatus 1 (‘a typically Greek form of dress’).

⁴⁵ The ablative of comparison in Latin is a ‘true’ ablative and so should be a genitive in Greek; the use of the dative here recalls the issues raised with the preposition ex taking a Greek dative instead of this genitive (see above on αἵρεσις and further below on ἀριστεία).

⁴⁶ Only in the second letter (Att. 14.9.1 sic enim tu ad me scripseras) is this hinted at, although the corrupt text (see below) makes certainty impossible.

⁴⁷ T&P and SB print as Greek, but the latter's apparatus criticus reports the tradition as aristian, or the even more Latinized aristiam, with a single minor MS showing ἀριστείαν (on the phenomenon of a single minor MS presenting Greek, see n. 11 above). Uniquely among the words studied in this paper, the Latin version is missing from the standard dictionaries of the language, presumably because editors have thought that the word should be written in Greek in our texts.

⁴⁸ Thus Adams (n. 1), 498–501 singles it out as extending the use of the Greek dative, without acknowledging that it is a conjecture (see the discussion of SB, Att. vol. 6.309–10).

⁴⁹ Adams (n. 1), 330–5.

⁵⁰ So SB ad loc.

⁵¹ Swain (n. 2), 160.

⁵² Fam. 13.53 can only be dated to some time during Cicero's governorship of Cilicia, but it surely was written after Att. 5.15, the first letter to Atticus from the province, and sent only three days after the proconsul's arrival.

⁵³ e.g. from the current study Luc. 27 (δόγμα) and Fin. 3.55 (ποιητικός).

⁵⁴ Note that the apparatus in Kumaniecki's 1969 Teubner records that the consensus of the oldest (IX–X) MSS at De or. 2.270 is ironia, and only later (XV) exemplars have the Greek lettering which he prints.

⁵⁵ Following the information in Westman's 1980 Teubner. In the two earlier passages a single late (XV) MS has the word in Greek letters.

⁵⁶ Quint. Inst. 2.10.12 (printed as Latin by Winterbottom, but even here with a Greek ending: epidicticon) vs 3.4.13 & 3.4.14; Fronto, Aur. 3.17.2 has one Latinized form amongst several in Greek.

⁵⁷ See now N. Leber, ‘Portraits of character: depictions of Cassius and Brutus in the correspondence of Cicero’ (Diss., University of Western Australia, 2015), 137–67 (http://research-repository.uwa.edu.au).

⁵⁸ It seems unlikely that Catullus 64 predates Att. 1.17: see NP. Ant. s.v. Catullus [1].

⁵⁹ All instances are from Att. (I have not tried to make sense of the corrupt fragmentary letter to Octavian in Nonius p. 288M): thus heros (of Milo 4.3.5 and of Caesar's assassins 14.6.1, 14.11.1) and ἥρως (of Labienus 7.13.1 [s.v.l.], of Celer 9.18.2 [s.v.l.], where he ‘is called a hero, not without sarcasm, simply as participating in the νέκυια’ [SB ad loc.] and again of the assassins 14.4.2, 15.12.2). I cannot offer any explanation for the change of alphabet in these circumstances, especially when reference is made to the same men.

⁶⁰ I have excluded from Table 2 the superlative ἱστορικώτατος in Att. 6.2.3, as this could not be Latinized by a simple transliteration.

⁶¹ Varro's use of the word in Rust. must be after Cicero's first use (Mur. 16) in 63 (see Rust. 1.1.1 and, in general, Cornell, T.J. et al. [edd.], The Fragments of the Roman Historians [Oxford, 2013], 1.415Google Scholar).

⁶² TLL 5.2.559.73–7 cites only a scholiast on Horace and a conjecture in Julius Victor.

⁶³ ap. Gell. NA 11.2.5 (= Mor. 2).

⁶⁴ poeticus always means ‘poetic’ in the literary sense in Latin (the only exception in TLL is from Boethius’ translation of Aristotle, where ποιητική in the philosophical sense is rendered poetica). I have excluded from the list of poeticus (see above) Cicero's use of poetica [sc. ars] as a virtual noun.

⁶⁵ LSJ s.v.

⁶⁶ The text of the Censors’ edict is preserved in Suet. Gram. et rhet. 25 and Gell. NA 15.11.2.

⁶⁷ Cicero uses rhetor as a Latin word fifty-eight times, with around thirty-seven in the sense ‘instructor in rhetoric’, although the context is not always enough to distinguish the meaning (for lists of passages see n. 18 above).

⁶⁸ Mar. Victorin. Rhet. 1.1 (156.23–5 Halm, Rhet. Lat. Min.).

⁶⁹ OLD s.v. b.

⁷⁰ Swain (n. 2), 156 uses this term of Cicero's replacement of a Greek word already integrated into Latin with ‘a form closer to that of the original’.

⁷¹ It is the agent noun from εἴρω: see Beekes, R.S.P., Etymological Dictionary of Greek (Leiden, 2010), 2.1284Google Scholar.

⁷² Cf. the restoration in Enn. inc. 42.

⁷³ Despite the traditional numbering, Att. 13.28 appears to predate 13.12: see Schmidt, O.E., Der Briefwechsel des M. Tullius Cicero von seinem Prokonsulat in Cilicien bis zu Caesars Ermordung (Leipzig, 1893), 293, 318Google Scholar.

⁷⁴ There are 271 instances of it in the Latin form (see n. 18 above). Stang, N., ‘Philosophia, philosophus bei Cicero’, SO 11 (1932), 82–93CrossRefGoogle Scholar, at 83 thought that for the single Greek occurrence at De or. 1.9 we should read a Latinized form, but the MSS evidence seems unambiguously against this.

⁷⁵ Acad. 1.25, Div. 2.11, Fin. 3.5. But other evidence for pre-Ciceronian usage is surprisingly sparse: see TLL 10.1.2024.37–40, 46.

⁷⁶ I rely on Madvig's third (1876) edition, as more recent editors seem less interested in reporting this information. He notes that A (Vaticanus Pal. lat. 1513), much the oldest of the MSS (see Reynolds [n. 13], 113), has the Latinized forms at 3.51 (similarly TLL s.v. proegmenon).

⁷⁷ LSJ s.v. II.

⁷⁸ ὑπόστασις is also an untranslatable pun: ‘plan, purpose’ and ‘reality’ as a philosophical technical term (LSJ s.v. B II 3 & III 2).

⁷⁹ Swain (n. 2), 157.

⁸⁰ Att. 1.13.4, 1.18.6, 2.12.4, 7.8.3, 7.9.2, 9.4.2 (bis), 10.1.3, 10.13.1, 13.30.2, 16.15.3. Att. 7.9.2 simul hoc διευκρινήσεις πρόβλημα sane πολιτικόν breaks up the phrase, but πρόβλημα πολιτικόν is a term of Peripatetic philosophy: see Mutschmann, H. (ed.), Divisiones quae vulgo dicuntur Aristoteleae (Leipzig, 1906), 55Google Scholar and Diog. Laert. 5.47 (in a list of the works of Theophrastus).

⁸¹ Att. 2.1.3, 5.12.2, 13.10.2, 14.6.2, 14.14.1, Fam. 2.8.1.

⁸² Att. 4.8a.4, 7.8.4, 7.13.1.

⁸³ This letter appears to be the one work we have of Cicero which contains more Greek than Latin words: the context hardly favours an assumption that the transmission has changed Cicero's Greek into more recognizable forms.

⁸⁴ Fin. 4.5, 5.66; Quint. Inst. 1.10.15; cf. Apul. De dog. Plat. 2.8.

⁸⁵ A PHI search (see n. 18 above) for the string ciuil in Cicero yields 364 matches. For the general principle of a preference for Latin words, see Off. 1.111.

⁸⁶ The context of both passages suggests that he may be indirectly quoting Greeks’ designations of the documents.

⁸⁷ Editors print this last example as Greek, but the better MSS have it in Roman letters. Shortly after this was written, it seems, Cicero's son (Fam. 16.21.8) used the thoroughly Latinized hypomnematis (abl. pl.) with reference to his Greek lecture notes. These are the only two occurrences of the word in Latin.

⁸⁸ Around seven hundred of the (more than) nine-hundred Greek words collected by Rose (n. 4) are from the letters to Atticus.

⁸⁹ Admittedly, ‘publication’ was something less distinct in the ancient world than in ours (so Phillips, J.J., ‘Atticus and the publication of Cicero's works’, CW 79 [1986], 227–37Google Scholar, at 228), and Cicero did talk about publishing his letters towards the end of his life (SB, Att. vol. 1.59–60). But the crucial distinction was explicitly made shortly after Cicero's death by Cornelius Nepos, Att. 16.3, contrasting the letters to Atticus with those books qui in uulgus sunt editi. There is much of interest about the eventual publication of the letters by others in Beard, M., ‘Ciceronian correspondences: making a book out of letters’, in Wiseman, T.P. (ed.), Classics in Progress: Essays on Ancient Greece and Rome (Oxford, 2006) 103–44Google Scholar, who remarks (perhaps a little too scrupulously) that they were ‘not (for the most part at least) specifically written for publication’ (123).

⁹⁰ See Baldwin, B., ‘Greek in Cicero's letters’, AClass 35 (1992), 1–17Google Scholar, for some examples of the differences.

⁹¹ So Pelttari (n. 4), 464 n. 8, on Augustine's apparent preference for writing Greek in Latin characters, ‘influenced by his intended audience's ignorance of Greek’.

⁹² e.g. εἰρωνεία Luc. 15 & 74, δόγμα Luc. 27 & 106, etc.

⁹³ See Nep. Att. 4.1 on Atticus’ ‘near native’ knowledge of Greek.

⁹⁴ There are dozens of these, some of which are otherwise unattested in either language, e.g. gymnasiode in Att. 1.6.2 and 1.9.2.