Validity discourse analysis by james paul gee

Page 94 14. What social goods e.g. status, power, aspects of gender, race, and class, or more narrowly defined social networks and identities are relevant and irrelevant in this situation? How are they made relevant and irrelevant, and in what ways? 15. How are these social goods connected to the cultural models and Discourses operative in the situation? Connection building 16. What sorts of connections – looking backward andor forward – are made within and across utterances and large stretches of the interaction? 17. What sorts of connections are made to previous or future interactions, to other people, ideas, texts, things, institutions, and Discourses outside the current situation this has to do with “intertextuality” and “inter- Discursivity”? 18. How do connections of both the sort in 16 and 17 help together with situated meanings and cultural models to constitute “coherence’’ – and what sort of “coherence” – in the situation?

5.8 Validity

Throughout this book I have held off discussing the question of what constitutes validity for a discourse analysis. This question could not be answered until enough of the “tools of inquiry” used in a discourse analysis had been laid out. However, now we are ready to deal with the issue of validity, an issue that has continually vexed so-called “qualitative research” but see my remarks in the Introduction to this book. Validity is not constituted by arguing that a discourse analysis “reflects reality” in any simple way Carspecken 1996; Mishler 1990. And this is so for two reasons: First, humans construct their realities, though what is “out there” beyond human control places serious constraints on this construction so “reality” is not “only” constructed. Second, just as language is always reflexively related to situations so that both make each other meaningful, so, too, a discourse analysis, being itself composed in language, is reflexively related to the “language-plus-situation” it is about. The analyst interprets his or her data in a certain way and that data so interpreted, in turn, renders the analysis meaningful in certain ways and not others. These two considerations do not mean that discourse analyses are “subjective,” that they are just the analyst’s “opinion.” I take validity to be something that different analyses can have more or less of, i.e. some analyses are more or less valid than others. Furthermore, validity is never “once and for all.” All analyses are open to further discussion and dispute, and their status can go up or down with time as work goes on in the field. Page 95 Validity for discourse analysis is based on the following four elements: 1. Convergence: A discourse analysis is more, rather than less valid i.e. “trustworthy”, the more the answers to the previous eighteen questions converge in the way they support the analysis or, to put the matter the other way round, the more the analysis offers compatible and convincing answers to many or all of them. 2. Agreement: Answers to the eighteen questions are more convincing the more “native speakers” of the social languages in the data and ‘‘members” of the Discourses implicated in the data agree that the analysis reflects how such social languages actually can function in such settings. The native speakers do not need to know why or how their social languages so function, just that they can. Answers to the eighteen questions are more convincing the more other discourse analysts who accept our basic theoretical assumptions and tools, or other sorts of research e.g. ethnographic research, tend to support our conclusions. 3. Coverage: The analysis is more valid the more it can be applied to related sorts of data. This includes being able to make sense of what has come before and after the situation being analyzed and being able to predict the sorts of things that might happen in related sorts of situations. 4. Linguistic details: The analysis is more valid the more it is tightly tied to details of linguistic structure. All human languages have evolved, biologically and culturally, to serve an array of different communicative functions. For this reason, the grammar of any social language is composed of specific forms that are “designed” to carry out specific functions, though any form can usually carry out more than one function. Part of what makes a discourse analysis valid, then, is that the analyst is able to argue that the communicative functions being uncovered in the analysis are linked to grammatical devices that manifestly can and do serve these functions, according to the judgments of “native speakers” of the social languages involved and the analyses of linguists. Why does this constitute validity? Because it is highly improbable that a good many answers to eighteen different questions, the perspectives of different “inside” and “outside” observers, additional data sets, and the judgments of “native speakers” andor linguists will converge unless there is good reason to trust the analysis. This, of course, does not mean the analysis is true or correct in every respect. Empirical science is social and accumulative in that investigators build on each other’s work in ways that, in the long run, we hope, improves it. It does mean, however, that a “valid” analysis explains things that any future investigation of the same data, or related data, will have to take seriously into account. Validity is social, not individual. A given piece of discourse work will have a major point or theme, or a small set of them. These are the work’s Page 96 hypotheses. Authors will normally argue for the validity of their analyses by arguing that some aspects of convergence, agreement, coverage, and linguistic details are met in their analysis. But no piece of work can, or should, ask all possible questions, seek all possible sources of agreement, cover all the data conceivably related to the data under analysis, or seek to deal with every possibly relevant linguistic detail. A discourse analysis argues that certain data support a given theme or point hypothesis. In many cases, for the individual piece of work, convergence and linguistic details are the most immediately important aspect of validity – that is, showing that answers to a number of questions like our eighteen questions pp. 92–4 and linguistic details converge to support the analysis. It is important, as well, that these questions come from a consideration of different building tasks, not just one, and that a number of different linguistic details support the conclusions drawn. It is important, too, that the researcher openly acknowledges if any answers to these questions or any linguistic details support opposing conclusions. Various aspects of agreement and coverage are also important in different ways in different sorts of studies sometimes through citations to, and discussion of, the literature. The individual piece of work is, then, of course juxtaposed to earlier and later work in the field. This juxtaposition allows further aspects of convergence, agreement, coverage, and linguistics to be socially judged and adjudicated. Validity is as much, or more, in those social judgments and adjudications, as it is in an individual piece of work.

5.9 Starting to do discourse analyses