There is considerable variation in the use of the term discourse Jm_a_ly__sis, so I had better insert a brief note on what I mean by it. First, I do not intend to draw any important distinction between text and discourse. As these terms are normally used, they often imply only a difference in emphasis, but nothing of theoretical importance need hang on the distinction. A distinction is sometimes implied between written text and spoken discourse. Alternatively, discourse refers to interactive language versus text as non- interactive monologue, whether spoken or written. For example, one can talk of the text of a speech. Another distinction is that discourse implies length whereas a text may be very short. Halliday and Hasan (1976) define a text as a semantic unit and point out that complete texts include Exit and No smoking. Some scholars have used the two terms to label theoretically important distinctions, but since I will not be concerned with those distinctions here, I will ignore them.
- Similarly, I will favour the term discourse analysis over other terms for reasons of convenience rather than theory. The term text analysis would do equally well except that it usually implies a particular European tradition of text linguistics. The term conversational analysis might also serve, except that it almost always implies work which derives from ethnomethodology. It is also too narrow in that it implies a restriction to conversation, and the exclusion of more formal discourse, although there are problems in the analysis of discourse which are common to formal and informal, written and spoken language - for example, the analysis of lexical and grammatical cohesion.
- By discourse analysis I mean therefore the linguistic or sociolinguistic analysis of naturally occurring discourse or text, spoken or written. This does not deny the validity of other approaches: a full understanding of discourse is necessarily interdisciplinary. It merely restricts my topic here to manageable limits. (This definition of discourse analysis is developed at length by Stubbs, 1983b).