At Petron. Sat. 48.4 Trimalchio makes a famous boast about owning three libraries:

Because ‘one’ plus ‘another (of two)’ cannot equal three, editors since Tilebomenus (Jacques Mentel or ‘Mentelius’, in his edition published in Paris in 1664)Footnote ² have helped the numbers add up by supplying II where the MSS have tres. Thus one sees a fairly even split in the major recent editions of the ‘Cena Trimalchionis’ of the Satyrica between those who have made the maths work by way of emendation (Bücheler, Smith, Müller, Öberg) and those who have stuck with what is quite obviously the lectio difficilior (Ernout, Sage & Gilleland, Schönberger, Schmeling).Footnote ³

Those advocating tres have the MSS solidly on their side. But the stronger reading leaves them to make sense of a nonsensical assertion. It has been explained as ‘dumme Angeberei’ [‘stupid bravado’] on Trimalchio's part;Footnote ⁴ as a joke that we can no longer get;Footnote ⁵ ‘comic aposiopesis;’Footnote ⁶ an absurd assertion highlighting Trimalchio's stupidity and confusion about libraries coming only in pairs;Footnote ⁷ and as a reference to an actual third library (Atellanae in Oscan) that Trimalchio neglects to enumerate.Footnote ⁸ In his recent Oxford commentary, Gareth Schmeling has gathered up the main solutions that have been tried in support of keeping tres only to conclude, rather half-heartedly (pace Smith), that Trimalchio ‘might not mean that he has tres bibliothecas, “three libraries”, but rather that he has “three bookcases of books”’.Footnote ⁹ In a recent article not cited by Schmeling, Béla Adamik has sought to explain Trimalchio's three-library boast as a sudden flash of cultural inferiority on Trimalchio's part that finds him suppressing the third item in his list (more on this below).Footnote ¹⁰

Emending tres to II makes things easy, because it makes Trimalchio boast of having two libraries as if that were a remarkable achievement when, in fact, double libraries (Greek and Latin) were the norm for the educated elite in first-century Italy. If tres is emended to II, the numbers add up, and Trimalchio is caught out for proudly tallying as two libraries what most Romans would take to be the norm, and perhaps count as only one.Footnote ¹¹ To support emending tres to II editors have proposed a transmission history that has an original II (or an even more original duas subsequently rendered numerically as II) somewhere behind the fifteenth-century codex Traguriensis.Footnote ¹² At some point that original II was mistaken for III and mistranscribed, and the mistranscribed III was subsequently spelled out by yet another hand (likely by the copyist who produced the codex Traguriensis) as tres. To this textual-transmission argument they lend grammatical support by pointing out that the pronoun alter means ‘one or other (of two)’ in any given list of two, such as the one that Trimalchio provides.

But ‘the fact’ that the list's grammar is solidly on the side of II has recently been called into question by Bela Adamik, in a detailed analysis of the linguistic and grammatical habits of itemized explanatory lists in the Satyrica as a whole. Adamik has shown that in all two-item lists other than the one in question (Trimalchio's library boast) Petronius does not in fact use an unus–alter construction but an alter–alter construction. Even more significantly, Adamik demonstrates that, in laying out lists longer than two, Petronius always begins unus–alter, precisely the way Trimalchio begins (and concludes) the list of his libraries. The set of passages that Adamik has at his disposal as exact comparandi is too small to prove decisive,Footnote ¹³ but they are enough to suggest that the grammatical support for II is less certain than it has been made to seem, and that, within the microenvironment of the Satyrica itself, the stronger grammatical argument is in fact on the side of tres. In short: elsewhere in the Satyrica Petronius does not unfold two-item lists the way he does in the case of Trimalchio's three-library boast, but he always begins larger lists of three or more in exactly this way (unus–alter–tertius etc.).

The grammatical evidence leads Adamik to conclude that a third item has been suppressed from Trimalchio's list. Reviving an argument that was once put into play by Pellegrino,Footnote ¹⁴ Adamik argues that the missing item is a library of Semitic texts that Trimalchio sets out to name (thinking himself, as an owner of three libraries, that much more impressive than the emperor, who has only two), but then thinks better of naming because to do so would call attention to his humble origins as a slave from Asia Minor.Footnote ¹⁵ Adamik supports his argument by showing just how doggedly the Romans adhered to a strictly binary sense of what counted for literary learning and culture, refusing to acknowledge, and at times actively discounting, the cultural worth of anything that was not Greek and/or Latin.Footnote ¹⁶ The thesis is both provocative and well argued, but it presents a number of further difficulties that make it hard to credit in full. To my mind, the two most significant drawbacks are: (i) that it requires readers to supply an entire linked set of observations in the blink of an eye (Trimalchio failed to name a third library; he is Semitic; his third library is therefore probably Semitic; he must have had second thoughts about mentioning it), all based on the failure of a nonsensical blowhard to make sense. And (ii) that it catches in a moment of self-doubt and embarrassed self-reflection none other than Trimalchio, a man who is otherwise tone-deaf to the niceties of Roman high culture, and utterly incapable of correctly assessing how he might be coming off.

In deference to the scholars who have produced so many possible (but ultimately unprovable) solutions on both sides of the II vs tres divide, I would like to propose a rather mundane syntactical solution that will allow one both to stick with the MSS reading of tres, and to make Trimalchio's ‘one’ plus ‘another’ equal it, without actually adding a third library. I propose it as yet another possible (but unprovable) solution to be added to an already overburdened list. The solution is suggested in the fine print of OLD s.v. alter ² 6. There we are informed that phrases of the unus et alter type mean ‘one or two persons or things’ in the sense of ‘a few’ or ‘a mere handful’. Generally not a literal ‘two’, in other words, but ‘several’. Giving more room for explanation and examples, Lewis and Short s.v. alter 2d α list a number of cases where unus–alter is used to mean exactly ‘two’, but then follows at 2d β a longer list of loci where the phrase is used to mean ‘several’ or ‘a few’ (‘more freq. of an indef. number’). The phrase has a literal sense of ‘two’, in other words, but its preponderant usage is in the sense of ‘one or another’, or ‘a handful’, or ‘a few’.

Examples of both types are to be had from literary sources of all types. One sees clearly that unus–alter means exactly ‘two’ at Cic. Leg. 2.5, where Cicero has Marcus claim that ‘I think that all citizens of a municipium have two fatherlands, one of nature, the other of the state’ (omnibus municipibus duas esse censeo patrias, unam naturae, alteram ciuitatis). The same ‘two = one plus another’ phraseology is found at Cic. Top. 6, where he says that ‘every systematic reckoning of argumentation has two parts, one of invention, the other of judgement’ (omnis ratio diligens disserendi duas habeat partis, unam inueniendi alteram iudicandi). Trimalchio matches the Ciceronian examples in his phraseology, but his counting is mysteriously off by one (‘three = one plus another’).

As the Lewis and Short entry makes clear, phrases of the unus (et) alter type are used far more often in Classical Latin to imply an indefinite small number that exceeds the number two. Examples abound in well-known sources. At Epist. 2.1.74-5 Horace says of stiffly wrought archaic poems that ‘lines here and there may strike us as better crafted than the rest (uersus paulo concinnior unus et alter, 74), but they ought not to carry and sell the entire work’, and at Sat. 1.6.100-2 he complains that, if he were to run for public office, ‘I should immediately have to acquire a large establishment, greet more visitors, take one or two companions with me (unus | et comes alter, 101-2) to avoid being on my own when going off to the country’ (trans. Rudd). At Rem. am. 363-4 Ovid says ‘as long as I am sung throughout the entire city, let a handful of naysayers impugn my work as much as they like’ (quamlibet impugnent unus et alter opus, 364). In all such cases the phrase implies not a literal ‘two’ but an indefinite ‘handful’, and this is by far the predominant use of phrases of the type.Footnote ¹⁷ Most examples of ‘one and another’ meaning ‘several’ use et, aut, atque or –ue to link some form of unus to a matching form of alter. But Plin. Ep. 9.10.3 uses the words in asyndeton to produce exactly the same sense of ‘some small number’. Writing to Tacitus, Pliny says oratiunculam unam alteram retractaui ‘I have revised one or two short speeches’ (trans. Walsh).

Back to Trimalchio. One sees what he is (quite possibly) up to: he is straining the potentials of the unus (et) alter phrase to say ‘two’ and ‘a handful’ at the same time (double entendre). He says ‘I have three libraries’, then he proceeds to count them out in a punning doubly-resonant way that finds ‘one plus one’ adding up to a literal ‘two’ (unam Graecam, alteram Latinam), but implying a loose smattering of ‘several’ at the same time. In terms of basic syntax, the specific phraseology that Trimalchio uses (‘one this, another that ’) has to mean ‘two’, not ‘several’, because it assigns one adjective to each part (as in the Ciceronian examples cited above). I am not arguing that Trimalchio's words say and mean ‘several’, but that he is playing with the phrase so as to imply that it does. Taking Adamik's linguistic data in a different direction, I argue that Petronius deviates from the phraseology that he uses elsewhere for two-item lists not because he would have us surmise that a third item is being suppressed from the list, but because the same double effect is not available to phrases of the alter–alter type (unus et alter can be used to mean ‘several’, alter et alter cannot).

As he does so often in the course of his cena, Trimalchio is engaging in linguistic play, striving to impress, but not quite managing to pull it off. The joke is both dull and un-clever, but that is fully in keeping with the rest of Trimalchio's jokes. Seen this way, the harder reading tres stands up to hard scrutiny, and nothing outside the potentials of the phraseology itself is needed to defend it. That said, translating Trimalchio's strained linguistic joke into reasonably funny and self-obvious English is another matter altogether. The best I can manage is something along the lines of the following: ‘I have three libraries, Greek and Latin … one of each, a handful in a pair!’ Or ‘I have a handful of libraries: a Greek one, and a second Latin one to go with the first [sc. Latin one]. That makes three!’.