The not-so-simple art of imitation: Pastiche, literary style, and Raymond Chandler

This analysis extends the tools of statistical analysis to the challenging task of distinguishing between genuine works by an author, the preeminent American writer of mysteries, Raymond Chandler, and deliberate attempts by others to mimic the author's style. Rendering the task all the more challenging, the analysis focuses exclusively on the main elements of Chandler's style rather than on his minor but telling stylistic idiosyncrasies. Statistical analysis establishes that indicators of these stylistic elements can successfully detect the pastiches. Statistical stylistics (a.k.a. "stylometrics") was born in 1887, when Thomas Corwin Mendenhall, a physicist, decided to pursue a suggestion he had encountered in the writings of Augustus de Morgan, a mathematician. Pondering the question of how to settle disputes about the authorship of various books, poems, and plays, Morgan advanced the principle that "one man writing on two different subjects [should agree] more nearly with himself than two different men writing on the same subject". As Ellegard (1962, p. 8) later explained, this principle flowed naturally from the twin assumptions that "some features, or combinations of features, in a particular writer's style or language, remain reasonably constant, or change in a predictable manner, throughout his production" and that "some at least of * Lee Sigelman is Professor and Chair of Political Science at The George Washington University. His research interests range widely throughout the social sciences, including research methods, mass communication, political behaviour, and popular culture. With Ernest Yanarella, he co-edited Political Mythology and Popular Fiction, and has published several articles on political themes in popular literature. t William Jacoby is an associate professor in the Department of Government and International Studies at the University of South Carolina. His work has focused substantively on mass political attitudes and behavior. He has a strong interest in statistical methods, and has written extensively on dimensional analysis. these features are sufficiently rare to set the author apart from all or most of his contemporaries". Bycomparing distributions of word lengths in passages from several representative t xts, Mendenhall pioneered a method that stylisticians have used ever since to resolve authorship disputes (Lord, 1956; Williams, 1956). Over the years, this task has engaged not only students of language and literature, but also some of the leading statisticians of the day (e.g., Mosteller and Wallace, 1984; Yule, 1944). An author who deviates significantly from norms established for the context in which he or she is writing is said to possess a distinctive literary style (Osgood, 1960, p. 293). Style is a probabilistic phenomenon. Recognizing, for example, that even a writer who flaunts an abstruse vocabulary will also need to use many mundane words, stylisticians regard style as a general predisposition toward a particular mode of expression rather than an invariant habit or constant. In Dolezel's (1969, pp. 10-11) words, "The overall character of a style is called forth by the degree of presence (or absence) of a certain mode of expression, rather than by its exclusive use (or complete suppression)." More formally: Each text can be represented by a set of measurable, statistical characteristics: T = {C1, C2 ... Cz) in