Alexander Fedchin, Tufts University
This paper provides the first systematic ranking of acrostics in Latin and Neo-Latin literature. The program performing the ranking examines all sequences of characters in a corpus and for each sequence estimates the syllabic similarity to Latin (consider the lorem ipsum placeholder text that is often used in publishing to fill empty space and which, although it is originally derived from Cicero, is nonsensical by design). The syllabic language model allows identifying acrostics that contain rare words or have been damaged in transmission, and I manually filter the top entries in this initial ranking to remove ungrammatical instances. The final result contains several previously unidentified acrostics, forms a basis for the classification of this type of wordplay, and casts doubt on the existing conjectures about acrostics in Vergil. The latter turn out to not only be unlikely from the statistical point of view but to also be dubious due to the function and location of these proposed acrostics in the text.
Qualitative analysis of the data gathered in this manner from the texts on the Wikisource, MusisQue DeoQue, and Poeti D’Italia databases highlights three functions that acrostics perform: signature, dedication, and key to a riddle. An example of the latter is a previously unidentified acrostic poem whose first ten letters spell ARSPOETICA (the poem appears in manuscript XIV.223 at Biblioteca Marciana; Padrin (1887) considers Mussato as one of the potential authors). All three functions listed above imply that an acrostic is meant to be discovered and interpreted by the reader. Indeed, the acrostics that the program locates are all positioned at the very beginning or the very end of the poem they appear in. This observation casts doubt on conjectures about acrostics located in the middle of a text unless such an acrostic is obviously intentional for other reasons.
These findings and the elaborate acrostics at the top of the ranking (e.g. MARCELLUSPALINGENIUSSTELLATUS - Marcellus’ signature at the beginning of Zodiacus Vitae) emphasize the fact that previous studies disproportionately favor Vergil and other golden age poets despite the lack of convincing examples where the possibility of accidental occurrence is statistically dismissible. The most well-known conjectures are, perhaps, the four-letter MARS acrostic defended by Fowler (1983) and the six-letter MAVEPU acrostic noticed by Brown (1963). I conclude the paper by challenging both conjectures and especially the latter, which involves three odd-numbered lines in Georgics I.429-33. To add to the existing critique of this proposal (Nisbet 1990), the arguments about positioning mentioned above, and the absence of statistical justification, I present another ‘signature acrostic’ structured in exactly the same way (three even-numbered lines contributing two letters each) in a work of a different author: Horace’s Odes III.3.12-16 spell PUVEMA, an observation that, to my knowledge, no one has made before. This new finding further exposes the MAVEPU acrostic as a random aberration and highlights the fact that there are probably no acrostics left where many have looked before - it is the less studied Neo-Latin texts that abound with possibilities.