BICS-59-2  2016 © 2016 Institute of Classical Studies University of London 140 THE AUTHORSHIP OF THE HISTORIA AUGUSTA: TWO NEW COMPUTATIONAL STUDIES JUSTIN A. STOVER AND MIKE KESTEMONT The Historia Augusta (henceforth HA) is a collection of biographies of Roman emperors stretching from Hadrian (AD 117–138) to Carus (AD 282–283) and his sons Carinus (AD 283–285) and Numerian (AD 283–284). 1 The lives purport to be written by six different authors, Aelius Spartianus, Julius Capitolinus, Vulcacius Gallicanus, Aelius Lampridius, Trebellius Pollio, and Flavius Vopiscus, working under the Emperors Diocletian (AD 284–305) and Constantine (AD 306–337). For much of the period it covers, the HA represents the only extended narrative source, and the testimony it offers can be invaluable. Unfortunately, the HA is also famous for its bizarre details and puzzling omissions, as well as its lurid focus on emperors’ peccadilloes and personal habits to the detriment of their political accomplishments. It also notoriously includes documents — speeches, letters, laws — which are almost certainly fabricated by the author(s), and cites a whole host of authors nowhere else attested, and probably invented. 2 But the problem of the HA is not only its unreliability as an historical source: it also includes throughout troubling anachronisms, mentions of offce and titles that only came into being in the middle of the fourth century, decades after the supposed dates of its composition. In 1889, Hermann Dessau put forth the provocative thesis that the HA was in fact the work of a single author working under the reign of Theodosius (AD 379–395), and that division of the lives between six authors and their dedications to Diocletian and Constantine were merely a literary ploy. 3 Ronald Syme — the most infuential exponent of the Dessau thesis — would famously term the author ‘a rogue grammaticus’. 4 1. A computational solution? As early as the late 1970s, it was realized that this question of single or multiple authorship in a corpus offered a perfect test case for statistical methods of authorship attribution. Ian Marriott conducted a groundbreaking analysis, published in the Journal of Roman Studies in 1979, which suggested that computational analysis indicated single authorship of the corpus. 5 This was a seminal application of forensic stylometry, as developed by Mosteller 1 Justin Stover would like to thank George Woudhuysen for helpful suggestions. We are both grateful to the editors for accepting this paper and the anonymous referees for many helpful suggestions. The code and texts for this paper can be found in the following repository: https://github.com/mikekestemont/ruzicka. 2 See L. Homo, ‘Les documents de l’Histoire Auguste et leur valeur historique’, RH 151 (1926) 161–198 and 152 (1926) 1–31. 3 H. Dessau, ‘Über Zeit und Persönlichkeit der Scriptores Historiae Augustae’, Hermes 24 (1889) 337–92. 4 R. Syme, Ammianus and the Historia Augusta (Oxford 1968) 207. 5 I. Marriott, ‘The authorship of the Historia Augusta: two computer studies’, JRS 69 (1979) 65–77.