Timothy B. ALLISON Stylistic Variation in Aeschylus: the case of an ignored linguistic variable

In traditional philology, the language of Attic tragedy has generally been considered a Kunstsprache in which there exists "free variation" among alternative linguistic forms; that is to say, the individual choice of an Attic form versus a non-Attic form has no stylistic effect and is made only for metrical convenience. However, William Labov, the pioneer of modern sociolinguistics, has questioned the validity of "free variation" and has contributed two important insights into the nature of linguistic variables. First, linguistic variables, even very small phonemes and morphemes, can have large social and stylistic significance. Second, linguistic variation is not absolute but rather a matter of ratios, of "more or less" rather than of absolutes "always, optionally or never" (Labov, Sociolinguistic Patterns, 1972, xiv). In this paper, I draw from these insights to show how the seemingly insignificant distribution of alternative forms of a morpheme in Aeschylus reveals the stylistic force of that morpheme and helps to shed light on Aeschylus' changing methods of characterization.

 Inscriptions reveal that in the Attic language contemporary with Aeschylus the o-stem dative plural form, -ois, was in the process of superceding the older form, -oisi. As Leslie Threatte (The Grammar of Attic Inscriptions, 1996) has shown, the long form mostly yields to the short form around 450 B.C.E. in Attic prose inscriptions, and it thus seems extremely likely that the variants of this morpheme were available as stylistically significant linguistic variables at least towards the end of Aeschylus' career.

 In his earlier extant works, Pers., Sept. and Supp., Aeschylus does not appear to make any stylistic use of this variable. All characters, regardless of metre, use the long form at a rate of roughly 34%. However, in the Oresteia, while most characters follow this usage, the nurse (7%) and Apollo (57%) emerge as statistical outliers. By comparing the nurse and Apollo to all of Aeschylus' other characters and applying two different statistical tests, one finds that the chances of such deviations from the mean are miniscule (.2% and .6% respectively by one measure, and 2.4% and 1% by another) if the system is truly random, i.e. driven only by metrical necessity. While the other lower status characters, the guard in Ag., the doorkeeper in Choe. and Aegisthus, for example, do not offer enough data to determine statistical significance, it is suggestive at least that they, too, avoid the long form.

 These findings correlate well with three other studies. First, Richard Janko (Homer, Hesiod and the Hymns, 1982) found that the high rates of -oisi in the Shield of Heracles, the Hymn to Pythian Apollo and the Hymn to Hermes strongly suggest false archaism; this points to a conscious, stylistic manipulation of this variable in a genre outside of but related to Attic tragedy. Second, within Aeschylus, Stevens ("Colloquial Expressions in Aeschylus and Sophocles," CQ, 1945) and West ("Colloquialism and Naïve Style in Aeschylus" in Owls to Athens, 1990) also observe how the Oresteia goes to greater lengths to differentiate the characters. They argue that the Oresteia represents a development of Aeschylus' methods of representation, and my findings support this.

 In summary, I offer a case study of how the careful application of sociolinguistic insight and statistical methods to the study of a seemingly insignificant morpheme in Aeschylus can reveal Aeschylus' changing methods of characterization. Further, this study helps to recover a small part of how different characters in the Oresteia may have sounded to a contemporary audience.