1. A note on terminology used throughout this paper: both ‘code-switch’ and ‘Grecism’ may be used to refer to a borrowed term (as argued by Mayer [1999]), but there remains a subtle distinction between the terms, as the former emphasizes procedure while the latter emphasizes only difference in language base. For this discussion, where I will use both words, it should be assumed that all Grecisms are code-switches (though, naturally, not all code-switches are Grecisms).

2. Emilio Bosazzi (2000) has produced a very useful index of the Grecisms used by Vitruvius, sorted into categories based on usage history (e.g., neologisms, or previously rendered forms) and subject matter (e.g., architectural terms, astronomical terms, medical terms, et cetera). By his account, there are over four-hundred Greek borrowings in De architectura, of which over sixty percent are either vocabulary introduced into the Latin lexicon by Vitruvius for the first time or are previously borrowed words with new semantic meanings (such as coruus as a type of battering ram, rather than a type of raptor). This index, though a helpful resource for inquiries into Vitruvian Grecisms, offers very little interpretation of the words themselves.

3. In fact, the fluidity with which Vitruvius can approach both languages in his use of technical jargon complements the looseness with which he relates the very concept of proportionality throughout his text (the truest execution of which is impossible outside of theoretical schematics and designs). See Riggsby (2016).

4. There are, however, probabilities of code-switching that determine what units are most likely to cross linguistic barriers. For example, a noun is the most common of all grammar units borrowed across languages, whereas conjunctions (and other functional words) are rarely transported across language barriers (Field [2002]).

5. In this respect, code-switching is distinct from the phenomenon of ‘language interference’, where grammar or syntax of a secondary language is compromised by rules of the first, or vice versa. Whereas language borrowing can be performed from a position of relative ignorance, language interference is not, and reveals bilingual capacity by subtle errors (Adams [2003], 27f.). There are of course other terms used to describe similar phenomena, some of which overlap. The psycholinguist Grosjean, for example, used the phrase ‘language borrowing’ to describe the deliberate actions of monolinguals, which he contrasts with the more spontaneous ‘speech borrowing’, or ‘nonce borrowing’ (see Field [2002]).

6. Dickey (2018).

7. Gardner-Chloros and Weston (2015a), 189.

8. Gardner-Chloros and Weston (2015a) and (2015b).

9. Thus, Greek authors, especially those of the Second Sophistic, preferred Atticized Greek, even after koine had spread throughout the Mediterranean. Replicating Homeric or Platonic Greek is effectively analogous to a modern English speaker opting to write in the style of Shakespeare, or even Chaucer, rather than a contemporary author.

10. Adams (2003), 13f., in fact, concludes that the ‘extent and quality of elite Roman bilingualism … cannot be determined’, because, on the one hand, the basis of comparison is not preserved and, on the other, because knowledge of Greek was a marker of status, and, as such, was prone to being exaggerated either way, whether to the credit or mockery of the alleged bilingual individual. Such exaggeration occurs, for example, in Lucilius’ second book of Satires, where the satirist hilariously presents the court case of Albucius and Scaevola as a battle of philhellenism, with knowledge of the Greek language wielded like weapons (Persyn [2019]).

11. Myers-Scotton (1993) and (1998).

12. Jorma Kaimio's seminal work on this subject (1979) remains an excellent study. Frederique Biville and Bruno Rochette have also contributed vastly to the study of bilingualism in the ancient world, the power dynamics reflected by what patterns may be observed, and the apparatus of language use represented in the ancient epigraphical corpora; see, for example, Biville (2002); Rochette (2010) and (2011).

13. Feeney (2016); see also Wallace-Hadrill (2013). Siobhan McElduff (2013) has also contributed helpfully to our understanding of the attitude of one-upmanship that forms the foundation for many ancient Latin translation projects, and Claudia Moatti's related work (2015) likewise remains seminal.

14. Glauthier (2020).

15. Despite Latin's supplanting of the native languages of the Italian peninsula (Penney [1988]).

16. It is thus significant that Vitruvius, despite his status as an author of an intact work from the Augustan age, is more or less absent from recent scholarship on Greco-Latin bilingualism, with both J.N. Adams (2003) and Gregory Hutchinson (2013) deferring to previous studies of Vitruvius’ use of Greek.

17. Lendle (1992) regards Vitruvius as a translator, an Übersetzer, of Greek, while Schiefsky (2005) is primarily concerned with the overall inconsistency of Vitruvius’ methodology of introducing ballistic terminology.

18. See especially Callebat (2013).

19. Roby (2016), 82, italics original.

20. As one might tolerate unwanted raisins in a disappointing oatmeal cookie.

21. The preface to book 7, for example, provides a lengthy list of Vitruvius’ perceived Greek forebears, which includes architects alongside philosophers and literary figures. Vitruvius crafts a similar list of experts in book 1, concluding this opening passage with an appeal to Caesar and a seemingly disingenuous request for patience with his own lack of literary style. The disingenuity, I would argue, is revealed in the phrasing that follows: namque non uti summus philosophus nec rhetor disertus nec grammaticus summis rationibus artis exercitatus, sed ut architectus his litteris inbutus haec nisus sum scribere (‘For not as the greatest philosopher, nor as an eloquent orator, nor as a grammarian disciplined in the greatest grammatical skills, but as an architect trained in literature I rely on these things to write’, 1.1.18)—an ‘architect trained in literature’ would, indeed, aptly describe the figure Vitruvius aims to cut, and is hardly a humble aspiration (pace Mayer [2005]).

22. For a modern example: if you are not sure what an ‘adze’ is, my French parenthetical reference that I mean an ‘herminette’ will probably not enlighten you to the form or type of tool to which I refer.

23. These frequencies are approximations. Uncertainty is due both to the state of attrition from the Greek and Latin corpora (from which we cannot determine how much material has been lost), and to the myriad foibles of digital and traditional search tools, dictionaries, and concordances. I have drawn my figures from cross-referencing TLG, TLL, Brepols Online Latin Databases A and B, and online corpus tools such as the Packard Humanities Institute, Perseus, and Logeion. As many of these tools include works that far post-date classical antiquity, I have used 500 c.e. as a cut-off point for source material.

24. Rose (1899), Krohn (1912), and Gros (1990) are united across their critical editions in this intervention, though Granger (1931) in his Loeb offers these terms in Roman letters. As Rowland, Howe, and Dewar (1999), 20, note, Vitruvius’ choices for translating, transliterating, and retaining Greek orthography and spelling vary vastly throughout the corpus, ‘reflect[ing] the eternal dilemma of any writer who works between two languages’. Apparently, the dilemma persists.

25. All of the Latinized words in this passage are nouns, a consistent feature that adheres to patterns of linguistic assimilation-probability outlined by Field (2002). On the other hand, many of the unassimilated, Greek-scripted borrowings are either adjectives or substantivized adjectives.

26. Note that both instances of drachma and obolos in Vitruvius’ work belong to this passage (3.1).

27. analogia, the exception, is borrowed with relative frequency into Latin: there are over three-hundred uses of the term from antiquity, almost one-hundred fifty of which are found in the Varronian corpus alone.

28. The marked consistency and delicate patterning of De architectura 3.1 becomes apparent when one contrasts this passage with what follows. In the sentence immediately subsequent, as Vitruvius enumerates the types of temple facades, he not only uses a phrase-long code-switch (quod Graece ναὸς ἐν παράστασιν dicitur, ‘which in Greek is called: a temple in pilasters’, 3.2), but also provides a list of Greek terms without definition (prostylos, amphiprostylos, peripteros, pseudodipteros, <dipteros>, hypaethros). This dramatically divergent manner of code-switching offers a uariatio that accentuates his prose style.

29. This does not include the unassimilated borrowing of the Greek term, architectōn, which is also in evidence in Latin literature from before Vitruvius’ time and is likewise found elsewhere in De architectura.

30. Likewise, ἐπιδίμοιρον (though formed logically according to the same linguistic patterns as the other ratios) occurs only three times in the Greek corpus, in the works of Clement of Alexandria and Vettius Valens, both active much later than Vitruvius. Neither πεντέμοιρον nor ἐπιδίμοιρον is found elsewhere in Latin literature. We thus have no evidence of Vitruvius’ sources for these terms.

31. Among these borrowings is Aulus Gellius, who describes this term, and hemiolios, as ‘lacking Latin equivalents’ (uocabula in lingua Latina non habent, NA 18.14). Vitruvius, however, does manage to supply each with a Latin equivalent (or paraphrase); strikingly, however, Vitruvius’ Latin terms (sesquialter and tertiarium alterum) are themselves less commonly utilized in Roman literature than the Grecisms epitritos and hemiolios, thus vindicating Gellius’ observation. Vitruvius does make a similar observation to that of Gellius, however, in book 5 of De architectura when writing about musical notation: harmonice autem est musica litteratura obscura et difficilis, maxime quidem quibus graecae litterae non sunt notae. quam si uolumus explicare, necesse est etiam graecis uerbis uti, quod nonnulla eorum latinas non habent appellationes (‘Harmonics, moreover, is a musical notation that is obscure and challenging, especially for those to whom Greek letters are not known. If we wish to explain this, it is necessary also to use Greek words, because some of them do not have Latin terminologies’, 5.4.1).

32. While individually these verbs may be encountered elsewhere in technical writing to introduce unmarked paraphrases and quotations (as one may encounter in Cicero or Varro), the stacking of three distinct verbs of speaking has a collective effect that is marked and is indicative of vocality rather than of writing or recording.

33. Nichols (2017), 26. See also the discussions of Wallace-Hadrill (2008) on how Vitruvius frames the integration of Greek and Roman cultures while simultaneously positioning himself to be an authoritative arbiter and teacher of their merging.

34. That the opening code-switch is ἀναλογία, a word particularly familiar to the heir of Julius Caesar, further emphasizes this function of crossing linguistic boundaries in order to form an intimate circle of communication.