Progressive Focusing and Trustworthiness in Qualitative Research

Although content analysis has been used as a label for a variety of methods and analytical techniques, we regard it as occupying a very specific theoretical space within the more general domain of qualitative research. We view content analysis as ‘a class of methods at the intersection of the qualitative and quantitative traditions’ (Duriau et al. 2007, p. 5), which places much emphasis on inter-rater reliability (Neuendorf 2002; Strijbos et al. 2006; see also Welch et al. 2011) or the idea that ‘different people should code the same text in the same way’ (Weber 1990, p. 12). In contrast, the type of qualitative research that we focus on in this paper is more generalised. More specifically, our concept of progressive focusing is perhaps closest to the domain of qualitative research that Welch et al. (2011) term ‘interpretive sensemaking’. Nonetheless, parts of our discussion may be useful for researchers using other types of qualitative research such as content analysis and grounded theory, or mixed-methods research.

Sinkovics et al. (2008) point out that reliability and validity have a somewhat uncertain place in the repertoire of a qualitative methodologist (Armstrong et al. 1997), as these dimensions are grounded on a different paradigmatic view and therefore not directly applicable to qualitative research. This is why alternative terms and ways of assessing qualitative research have been proposed, such as credibility, transferability, dependability and confirmability (Denzin and Lincoln 1994; Guba and Lincoln 1989; Kirk and Miller 1986; LeCompte and Goetz 1982).

For similar concepts, see also cycles of deliberation (McGaughey 2004, 2007), systematic combining/abductive approach (Dubois and Gadde 2002), zipping (Orton 1997) and evolution of perspective (Peshkin 1985).

A preferable method would be to import the entire document where possible, which allows the coding of content as well as the recording of key attributes. Whilst this can easily be done in the case of Word files, the majority of journal articles are accessed online as PDF documents, which in our experience often poses practical problems. In principle, newer versions of NVivo (8 and 9) can handle PDF files, however, many PDF documents (especially older ones) tend to be very large files or lack text recognition, and as yet, NVivo does not appear to have sufficient processing power to manage these efficiently. Nonetheless, given that many PDF texts can already be highlighted and annotated in freely available software such as Adobe Reader X, we believe that this limitation is likely to diminish in the future as more powerful versions of NVivo are developed.

In particular, the development of visual models based on coding templates is facilitated by the modelling function in software such as NVivo: the researcher can work on a dynamic version of their model in a continuous manner, whilst also saving static versions of the model at different points in time, thus tracking the evolution of the research model.