[...] the problem with both survey research and much ethnography is that they fail to topicalise their understandings. As Sacks says:

Now what I want to do is turn that around: to use what 'we' know, what any Member [i.e. member of society] knows, to pose us some problems. What activity is being done, for example. And then we can see whether we can build an apparatus which will give us those results. (1992a, 487)

---

As Maynard and Clayman (1991) argue:

Conversation analysts ...[are] concerned that using terms shuch as 'doctor's office', 'courtroom', 'police department', 'school room', and the like, to characterise settings ...can obscure much of what occurs within those settings...For this reason, conversation analysts rarely rely on ethnography data and instead examine if and how interactantss themselves reveal an oreintation to institutional or other contexts. (406-407)

---

This approach to data analysis is different from positivistic, survey research studies:

In more positivistic research designs, coder reliability is assessed in terms of agreement among coders. In qualitative research one is concerned with standardising interpretation of data. Rather, our goal in developing this complex cataloquing and retrieval system has benn to retain good access to the words of the subjects, without relying upon the memory of interviewers or data analysts.(27, my emphasis)

However, Glassner anfd Loughlin do suggest that their analysis fits conventional criteria of reliability (e.g. every finding was discovered independently by at least two analysts).

Source:

Silverman, D. (1993). Interpreting Qualitative Data. Methods for Analysing Talk, Text and Interaction. London: Sage Publications: 53 / 133 / 99