Patrick J. Burns
Frederick Ahl’s Metaformations argues for pervasive wordplay based on the “anagrammatic rearrangements” of words and subwords (specifically syllables) in Ovid’s Metamorphoses among other works of Latin poetry. Some contemporary reviewers saw the book as “an extremely important, new critical tool” (Gross 1987). Yet others questioned the basis for such ludic literary criticism, with one reviewer writing: “Even as a modern structuralist system his method requires more rules about which syllables count and greater general rigour” (Harrison 1986). In this paper, I set Ahl’s readings of syllabic wordplay against a systematic study of the frequency and distribution of trigrams (that is, three-letter subword combinations) within Ovid's poem in order to set a baseline for determining “which syllables count.” As a first step, each book of the Metamorphoses is transformed into its constituent trigrams (e.g. Met. 1.1 In noua fert... ~ ['inn', 'nno', 'nou', 'oua', 'uaf', 'afe', 'fer', 'ert', …]) and anagrammatic permutations of these trigrams are plotted and clustered (here, using k-means clustering). We can then compare the text sections with the smallest differences between clusters against the sections of text where Ahl has identified examples of concentrated wordplay. (See Figure 1 [https://bit.ly/scs-2021-metaformalism-abstract-img-1], for example, where tighter clustering of ‘sol’ anagrams occurs at the beginning of Book 2 as compared to Book 1; cf. Ahl 1985: 180-181 on “solar wordplays.”) As a second step, since Ahl describes anagrammatic wordplay as a localized effect, dispersion measures (such as Juilland’s D) are used to confirm an uneven distribution of anagrammatic trigrams in different books of the poem. So, for example, the ‘sol’ trigram shows a lower, i.e. less even, dispersion measure in Book 2 than in Book 1 (.751 vs. .837, where 1 indicates a completely even distribution; for comparison, note that ‘que’ shows .931), suggesting that is concentrated in specific passages of this book. The advantage of measuring, modeling, and mapping anagrammatic effects in these ways is not unlike that argued for in recent computational analyses of intertextuality in Latin poetry: that is, these methods help explicate and contextualize what readers have previously experienced as a “network of latent patterns” (Forstall and Scheirer 2019). The contribution of this study to Latin literary criticism is twofold: 1. it provides a baseline measurement of anagrammatic play in a major work of Latin literature against which qualitative arguments can be better situated; and 2. it repurposes a corpus-linguistic measure to make a quantitative argument not based solely on the frequency of features in texts but also on dispersion within them.