Identification of Latin Poets by Sound II

L.C. Curran: Identification of Latin Poets ? 45 Identification of Latin Poets by Sound II Leo C. Curran In a previous study,1 hereinafter called "Identification I," I established that many books of Latinpoetry could be assigned to their authors and to die works from which they were drawn solely on die basis of relative frequencies of sounds. My conclusion, with die texts dien available to me for analysis, was that it was possible, by sound alone, to distinguish one poet's unique voice from another poets and to measure die modulation of that voice from one of die author's works to another of diem. In addition, I presented in an appendix the results of some tentative exploration of die degree to which die differences in frequency of sounds in a book of Lucretius and a book of die Aeneid could be attributed to differences in subject matter radier than sound preference. However, die selection of texts used in die body of that article consisted, not of a coherent body of Latin poetry, but of what I happened to have in machine-readable form at die time and die appendix dealt with only a few sounds. The present study seeks to remedy die deficiencies of "Identification I." Part 1 expands die selection of texts so as to encompass a coherent whole, viz., almost all of Latin poetry of the Late Republican and Augustan periods. Part 2 does not deal with just a few sounds in die two texts chosen for sound-by-sound examination, but is a comprehensive treatment of all of die sounds that differ significantly in frequency in die two texts. The texts, Propertius 3 andAmores 1, were chosen because of their closeness in genre and in time. Parti: The New Texts I have now added 27 test books to die 48 analyzed in "Identification I," bringing the total number of texts to 75 and increasing die total number ofcharacters analyzed from just over two million to nearly 3.25 million (ca. 1600 pages). The new test books are: the works of Horace (die four books of Odes, die Carmen Saeculare, die Epodes, die two books of Satires, die two books of Epistles, and die 1 "Identification of Latin Poets by Sound," Syllecta Classica 2 (1990) 59-81, to which the reader is referred for certain general remarks and definitions not repeated in the current paperbut essential to its understanding. 46Syllecta Classica 3 (1991) Ars Poética), die 13 books of die Metamorphoses that I did not have dien (2-11, 1315 ), and die three books ofdie Tibullan corpus.2 I did not test die 27 new books in isolation. Instead I tested all 75 test books, old and new togedier, in order to see if any of die old books would be wrongly assigned to die new sources or any of die new books wrongly assigned to die old sources. In "Identification I," each test book was compared successively widi twelve sources. Widi die new sources added, die total of die sources is now sixteen: Eclogues, Georgics, Aeneid, Amores4ArsAmatoria, Metamorphoses, Fasti, Catullus 1-60, Catullus 61-68, Catullus 69-116, Lucretius, Propertius, Carmina, Epistulae, Sermones, and Tibullus (Books 1 and 2). This time I omitted die non-statistical proofs used in "Identification I" and used only die statistical proof, i.e., die chisquare test.3 It should be noted tiiat some statisticians would regard this metíiod as logical radier than statistical. The results of the new test are set out in Table 1.1. For each combination of test book and source I constructed a 31-by-2 contingency table (31 sounds and two texts); degrees of freedom = 30. In each row of Table 1.1, the chi-square scores for die comparison ofdie test book in that row widi each ofdie sources are listed.4 The lowest score in the row is underlined; die column that contains die underlined score for tiiat row indicates die source to which die test book is assigned. For example, die underlined score in die row labelled V. Aen. 1 falls in die Aen. (= Aeneid) column and dius dus book is assigned to die Aeneid. Of die 75 old...