Qualitative Analysis of Stance
One of the principal findings of research on talk-in-interaction is that speakers design their utterances to perform specific social actions (Searle 1969; Atkinson & Heritage 1984; Levinson 2013). These actions, and the forms with which they are accomplished, do not arise haphazardly, but are instead sequentially organised in response to the speech of others and in keeping with the norms of particular speech events and activities (e.g., Labov 1971; Hymes 1974; Jefferson 1978; Schegloff 1982; Sacks 1987; Finegan & Biber 1994). In this chapter, I apply this basic precept to the analysis of a corpus of responses in online question-and-answer (Q+A) forums across three different topics (Family & Relationships, Politics & Government, and Society & Culture) and four world varieties of English (India, the Philippines, the UK, and the US). Like the other chapters in this volume, my central research question is how language use in these responses differs across topics and varieties. I approach this question through a close qualitative analysis of a small subset of the Q+A response corpus. The subset consists of 12 ‘texts’—one per topic for each of the four varieties-made up of the initial question posed and all of the responses provided. The texts themselves were automatically extracted from a larger corpus using ProtAnt, a tool that identifies the most (lexically) ‘typical’ text in a corpus by identifying texts that contain the most keywords when compared to a reference corpus1 (Anthony & Baker 2015). Table 10.1 lists the files that were identified as most typical in this way. Thus while the subcorpus I analyse in the following section is admittedly rather small (a total of 15,140 words with a mean value of 1,261.7 words per text), it is in a sense representative of the larger corpus from which it is drawn. The benefit of focusing on a smaller subcorpus is that doing so enables a detailed examination of certain aspects of linguistic form and content that would be difficult to explore in a larger sample. That said, it is nevertheless important to highlight that the findings to be discussed are based on a restricted empirical set, and it is therefore necessary to exercise caution when attempting to generalise from any patterns identified.