The approach to text pragmatics presented here is based on research into the discourse structure of expository text. The initial corpus investigated was drawn from A Textbook of Economics (Hanson 1953 ), and a model of discourse analysis was designed (Tadros 1981) using the notion of Prediction. The corpus was later expanded to include other areas such as law, stylistics and linguistics.
The model of discourse analysis presented here is based on two basic assumptions. The first is that written text is interactive since two participants are involved: writer and reader, although, of course, ‘the exigencies of the medium oblige one of the participants to be only represented at the writing stage, thus complicating the process for both parties’ (Sinclair 1980:255). This means that the writer takes on the roles of both addresser and addressee and incorporates the interaction within the encoding process itself (see Widdowson 1978a: 21).
The second assumption is that the writer is in agreement with the propositions expressed in the text unless s/he specifically signals detachment. So, for instance, if the writer says, ‘Every commodity is nothing more than a bundle of services’, s/he will be taken to be in agreement with the proposition, but if s/he says ‘It has been pointed out by some economists that every commodity is nothing more than a bundle of services’, s/he is overtly detaching him/herself from the proposition and attributing it to some other entity. In this latter case, s/he will at a later point be expected to give an evaluation of the proposition expressed.
The term Prediction has previously been used in a generalized sense to refer to the activity of guessing or anticipating what will come in the text, an activity based on the reader’s common-sense knowledge of the world, of content and formal schemata (Carrell 1983; Swales 1986). As used here, however, the term is much more specific: it refers to an interactional