Using corpora in discourse analysis.

In the linguistics literature, ‘discourse’ is often defined in two, not mutually exclusive, ways, structurally, for instance, ‘language above the sentence or above the clause’ (Stubbs, 1983: 1) and functionally, for example, ‘language that is doing some job in some context’ (Halliday, 1985: 10). We shall privilege the functional viewpoint here, though analysing the structures of discourses are important in shedding light on the jobs being done. It has to be stressed that discourse is not a special form of language, but a perspective upon it, language described not only as a set of interacting units and systems, but also precisely that implied by Halliday, as an instrument put to work, and the work which it does is the attempt by one participant or set of participants to influence the ideas, opinions and behaviour of other participants. Most forms of traditional non corpus-assisted discourse analysis have practised the close-reading (that is, ‘qualitative analysis’) of single texts or a small number of texts in the attempt to highlight both textual structures and also how meanings are conveyed. Some types, such as much work in critical discourse analysis (CDA), use few concepts from linguistics proper, tending to rely on the analyst’s knowledge and experience (and prejudices) of similar texts, in a manner reminiscent of literary analysis (though with a politically-driven purpose). Other traditional discourse analysis is more linguistically-grounded. Thompson (1996a: 108-112), for instance, demonstrates the power of functional grammar, in particular transitivity analysis, in displaying how meanings, including what we might call non-obvious meanings, are communicated. In what follows we will attempt to outline ways in which corpus-assisted discourse studies (CADS) can help build upon traditional qualitative linguistic analysis, what ‘added value’ it can bring. We contend that it can contribute in two ways, firstly by combining close reading with statistical ‘overview’ analysis, very generally of a large number of tokens of the discourse type under scrutiny and secondly by integrating into the analysis of discourses a number of insights into how discourses function which have developed within the field of corpus linguistics.