Skip to main content

How did the Romans write a million? Or five thousand? Or a hundred thousand? Most people know the Roman numerals up to a thousand, but even professional classicists struggle to go higher than that. Editors are no different, and where Latin texts include high figures our supposedly critical editions fill up with unreal numerals produced by error or guesswork. For example, all modern texts of Cicero’s letters, including that of Shackleton Bailey, the scourge of inherited nonsense, print a style of numeral shown to be unclassical already in the nineteenth century (Ad Q. Fr. 3.1.3, cf. Mommsen 1885, n. 1). Elsewhere a simple typographical error in an Oxford Classical Text, producing nonsense instead of a number, has survived all reprints, and quotation in the Oxford Latin Dictionary, and is even ‘translated’ in parallel editions based on the OCT (at Cicero, Rosc. com. 23, as noted by Axer 1980, p. 63, though he is unaware that the typo is not a real numeral at all). The list of such errors and nonsense forms in the numerals of modern editions could be extended at rather depressing length.

Yet numerals need not just be a pitfall for unwary editors. Properly understood, they provide valuable help in establishing the text, thanks to the distinctive history of Roman numeral palaeography. It developed differently from the palaeography of words and letters, and this enables us to pinpoint key stages in the transmission of a text in much the same way as distinctive abbreviations have been used to trace the geographical and historical paths of textual traditions.

In my paper I will focus on what I call the ‘early classical numerals’—the unwieldy term centum/decem milia vetusto more (per propria signa) expressa in Dessau, vol. 3.2, p. 797, is the only name they receive in the research literature—that is Iᴐᴐ = 5000, ccIᴐᴐ = 10,000, Iᴐᴐᴐ = 50,000, cccIᴐᴐᴐ = 100,000, and their combinations and multiples. For these, the oldest attested Roman numerals for large sums, I will present classical examples from inscriptions, wax tablets, and ancient manuscripts, as well as their survivals and mutations in mediaeval manuscripts. The datable instances imply that these numerals fell from active use in the earlier Roman empire, though they were still copied accurately in late antique texts of older works, prompting an antiquarian treatment by the sixth-century grammarian Priscian, as I will discuss.

Taking two sample texts full of large sums—Cicero’s speeches Pro Roscio comoedo and Verrine 2.3 (De Frumento)—I will show that analysis of the numeral palaeography yields some counterintuitive insights about the transmitted numbers in these works. There is general consensus that the numerals are hopelessly corrupt in Rosc. com., which survives only in mutilated form via a single humanist copy, whereas in the Verrines, with their wider and older manuscript basis, the numbers are assumed to be sound. By tracing the different processes of corruption of the early classical numerals in the two traditions it will be shown that we should reverse both these assessments: the transmitted numbers in Rosc. com. are secure, those in the Verrines dubious. This has serious consequences for editions, and readers, of the two texts and for those, such as economic historians, who build on the figures they contain, but I hope it may also illustrate more generally how numeral palaeography can be a help, not a hazard, for editors.