New Research on the Stylometry of Latin Prose

Organizer's Statement
As the 1995 conference of the ALLC-ACH in Santa Barbara showed, there is growing interest in exploiting the recently digitized corpus of Latin texts, particularly by scholars of style and authorship. At the 1996 conference we propose a three-paper session at which representatives of research teams will report on exciting new discoveries and work in progress in different sub-fields. Besides the focus on Latin prose, the three papers share a common methodology, which is based on statistical and stylometrical analysis.

(1) Bernard Frischer will report his team's discovery of a new distinguishing feature of Latin vs. Greek prose style. Cluster analysis reveals that whereas Greek prose authors place the direct object before or after the verb with equal frequency, Roman authors have a preference for placement before the verb. Frischer suggests that a current theory from Applied Linguistics can account for the fact that when Roman authors wrote in Greek, they often retained their native word order, and vice versa with Greek authors writing in Latin. The theory can also explain the related (and unexpected) observation that when Romans wrote in Latin utilizing a Greek source (and vice versa), they tended to shift their word order to reflect that of their source. For example, as cluster analysis proves, when the Greek writer Plutarch wrote biographies of Romans based on Latin sources, his Greek word order resembles the Latin pattern. But when he wrote about Greeks, his word order reverted to its normal Greek pattern. Frischer concludes by applying these insights to some problems of the native language of several ancient authors (including Plotinus and Charisius); and to the much-debated question of whether or not Book 3 of Cicero's De Officiis had a Greek source.

(2) Emily Tse then will report on her project to determine the validity of Dessau's thesis (first proposed in 1889) that the important late-antique collection of imperial biographies known as the Historia Augusta was written--not by the six authors mentioned in the manuscripts--but by a single author. She presents a summary and critique of recent work by Meissner, which attacked the Dessau thesis (made popular, after decades of controversy, in a pioneering computer study by Marriott in 1979) by focusing on rates of variance in single vs. multi-author corpora. Tse shows that although Meissner's study is flawed, it may well have put us on the right track to settling the question once and for all, while also providing a useful tool (variance of function words) for helping us to distinguish single from multiple-author corpora in Latin and, very possibly, in other languages as well.

(3) Fiona Tweedie then will take us forward a millennium to the Neolatin text known as the "De Doctrina Christiana", which some scholars have attributed to Milton. Debate about this question is fierce. Tweedie's work promises to make an important contribution. Working with digitized versions of this text and others by Milton and his contemporaries, Tweedie and her team identify discriminators that can separate the known Milton samples from the control texts to see where "De Doctrina Christiana" falls. In this paper, Tweedie will concentrate on the frequency of function words. First, Tweedie shows by a principal components analysis that study of the one hundred most frequently occurring words does successfully discriminate the known Miltonic and control texts. Then she studies where the "De Doctrina" falls among Milton's works. The results on each of her various tests differ, but in all cases the "De Doctrina" is seen to overlap with other works of Milton. The conclusion reached is that the author of the "De Doctrina Christiana" may be Milton, but literary genre is also a strong factor.