Movie conversation: a reflection of face-to-face conversation and a source for teaching spoken language.

Face-to-face and movie conversation are claimed to differ in terms of spontaneity (Taylor 1999, Sinclair 2004a). Face-to-face conversation is generally considered the quintessence of spoken language as it is totally spontaneous (Chafe 1982, Tannen 1982, Halliday 2005, Biber et al 1999, McCarthy 2003, Miller 2006), while movie conversation is usually described as non-spontaneous in that it is artificially written-to-be-spoken, prefabricated speech (Pavesi 2005) and consequently not likely to present the features that characterize conversation (Sinclair 2004a). The present paper investigates these two conversational domains in empirical data drawn from the Longman Spoken American Corpus, representing face-to-face conversation, and the American Movie Corpus (AMC), a corpus purposely built and manually transcribed to study movie language. Investigations employ corpus-driven criteria (Francis 1993, Tognini-Bonelli 2001) and Biber's (1988) Multi-Dimensional analysis approach, which applies multivariate statistical techniques, in order to study more than one statistical variable at a time. Preliminary results demonstrate that the linguistic features underlying involved production, non-narrative concern and situation dependent reference, typical of face-to-face conversation also characterize movie conversation. The resemblance between these two conversational domains has interesting implications for the teaching of spoken discourse as movies may be effectively used as a potentially valid source of material