Discourse analysis is a generic name for a range of different methodologies and techniques for investigating texts (in the broadest sense).
There have been various attempts to identify the core of 'discourse analysis' but in most cases there are exceptions within the broad gamut of approaches that fall under the discourse analysis umbrella.
It is suggested that, although the different forms of discourse analysis explore different aspects of language use, they all view language as social interaction and most address the social context of discourse.
Antaki (undated), for example, states that, with the exception of conversation analysis, one thing all forms of discourse analysis agree on is:
that the analyst's first focus must be on language, and what it does in the world...The next thing they agree on is that the analyst must 'go beyond' the data itself. The analyst has to interpret by appeal to a theory (e.g., a theory about society, or power, or culture).
Discourse analysts (of whatever kind) look for how things are constituted by what they call "discursive practices". If you set out on an investigation into a certain social phenomenon, you will find an identifiable set of things that go together....
e.g., particular words, phrases, terms of reference, metaphors, rhetorical styles, systematisations of knowledge (e.g., rule books, catechisms, manuals, style guides....)
....which, together, construct that pheonomenon as a certain kind of social object (e.g., 'homosexuality', 'Science' 'Muslims' etcetera).
In each of those cases, the social object is being constructed by the discourse's choice of description, and the associations it implicitly makes.
e.g., the choice between: Muslim vs Islamic
fundamentalist vs devout
and the association between "Muslim" and...terrorism vs insurgency vs freedom vs...?
Whichever choice you make, and whatever associations you imply, you will help to construct (or 'constitute') a certain social object. For DA [discourse analysis] (in common with many theories of language in general) the choice of one description over another, and the association of one description with another, is significant. The categories of the world are not ready-made, nor is any use of them neutral.
Hodges et al. (2008), in the context of medicine, provide a view of the generic nature of discourse analysis:
Discourse analysis is an effective method to approach a wide range of research questions in health care and the health professions. What underpins all variants of discourse analysis is the idea of examining segments, or frames of communication, and using this to understand meaning at a "meta" level, rather than simply at the level of actual semantic meaning. In this way, all of the various methods of discourse analysis provide rigorous and powerful approaches to understanding complex phenomena, ranging from the nature of on-the-ground human communication to the inner workings of systems of power that construct what is "true" about health and health care.
Hodges et al. (2008) also suggest that the multifarious approaches require a careful approach to empirical sources. Referring toRogers et al. (2005) they concur that 'some discourse analysis research suffers from scanty explanation of the analytical method used'. It is important, they argue, that data sources are explicitly described, including how interviewees or focus groups were chosen, and a clear statement of the context of the study.
They add, in a statement that suggests a tendency to a critical discourse approach:
The method of analysis should be clearly explained, including assumptions made and methods used to code and synthesise data. Finally, given that the goal of critical discourse analysis is to illuminate and critique structures of power, it is especially important that researchers describe the ways in which their own individual sociocultural roles may influence their perspectives. (Hodges et al., 2008, p. 572)