Why interview? (By David Silverman)*

Daivd Silverman has kindly allowed me to re-post a blog entry he wrote for SAGE last year.

Why Interview?
David Silverman – Goldsmiths College, University of London, UK

Instead of looking, listening and reading, the majority of contemporary qualitative researchers prefer to select a small group of individuals to interview or to place in focus groups. In this sense, by assembling a specific research sample, linked only by the fact that they have been selected to answer a pre-determined research question, such researchers prefer to ‘manufacture’ their data rather than to ‘find’ it in the ‘field’. Despite their earnest claims to do something quite different from quantitative research (more ‘humanistic’, more ‘experiential’, more ‘in-depth’), such manufacture of data to answer a specified research problem is precisely the method which quantitative research espouses.

Potter (1996, 2002) has roundly criticised researchers who use his own approach (discourse analysis) for depending too much on interview data and has argued for a greater use of naturally occurring data. Closely following my concept of ‘manufactured’ data he shows how interviews, experiments, focus groups and survey questionnaires are all ‘got up by the researcher’. Instead, he proposes what he humorously calls The Dead Social Scientist Test. As he describes it:

‘The test is whether the interaction would have taken place in the form that it did had the researcher not been born or if the researcher had got run over on the way to the university that morning’. (Potter,1996:135)

Potter’s Test suggests that the default data source for qualitative researchers should be those contexts which societal members ordinarily assemble for themselves. Faced with the ubiquity and complexity of such contexts, why would any researcher seek to create a special research setting in order to study how people behave? To those who argue that some members’ practices are difficult to access, we can agree but point out that such unavailability is only apparent and based on commonsense assumptions about where phenomena (e.g. ‘the family’) are to be found.

Yet, despite these cogent arguments, ‘artificial’ research settings, such as interviews and focus groups, have become predominant in qualitative research and even ethnographers usually feel compelled to combine and test their observations by asking questions of informants.

Of course, in all research, choice of data must, in part, depend upon our research problem. Equally, there is no question that all polarities should be investigated – particularly where, as here, they involve an appeal to ‘nature’ (see Speer 2002).

However, my own research experience teaches me that, all things being equal, it is usually a good ploy (and certainly an aid to the sluggish imagination) to begin a research project by looking at naturally occurring data. While iron rules are rarely a good idea in research, this rule has worked for me and many of my students.

Harvey Sacks continually reminded his students that our intuitions rarely give us a good guide to how people actually behave. We cannot rely on our memory of what someone said because such memory will not preserve the fine detail of how people organise their conversation. Nor is this problem soluble by using mechanical equipment to record research interviews. For people’s own perceptions are an inadequate guide to their behaviour.

By contrast, naturally occurring data can serve as a wonderful basis for theorising about things we could never imagine. As Sacks puts it, using what ordinarily happens in the world around us means: ‘we can start with things that are not currently imaginable, by showing that they happened’. (1992,1:420)

Potter has recently extended Sacks’s arguments and I can do no better that present below the five virtues that Potter finds in working with naturally occurring data (Potter,2002:540 adapted):

  • Naturally occurring data does not flood the research setting with the researcher’s own categories (embedded in questions, probes, stimuli, vignettes and so on).
  • It does not put people in the position of disinterested experts on their own and others’ practices and thoughts.
  • It does not leave the researcher to make a range of more or less problematic inferences from the data collection arena to topic as the topic itself is directly studied.
  • It opens up a wide variety of novel issues that are outside the prior expectations embedded in, say, interview questions.
  • It is a rich record of people living their lives, pursuing goals, managing institutional tasks and so on.

None of Potter’s five points, deny that interviews or experiments can ever be useful or revealing:

‘However, they suggest that the justificatory boot might be better placed on the other foot. The question is not why should we study natural materials, but why should we not?’. (Potter,2002:540)

Extracted from David Silverman A Very Short, Fairly Interesting, Reasonably Cheap Book about Qualitative Research, Chapter Two


Potter, J. (1996) Discourse analysis and constructionist approaches: theoretical background. In J.Richardson (ed.) Handbook of Qualitative Research Methods for Psychology and the Social Sciences,
Leicester:BPS Books: 125-140.

Potter, J. (2002) Two kinds of natural. Discourse Studies, 4 (4): 539-42.

Sacks, H. (1992) Lectures on Conversation, edited by Gail Jefferson with an Introduction by Emmanuel Schegloff, Blackwell, Oxford, 1992, 2 volumes.

Speer, S. (2002) ‘Natural’ and ‘contrived’ data:a sustainable distinction?. Discourse Studies, 4 (4): 511-25.

* The original post was featured on methodology.co.uk, SAGE’s forerunner to Methodspace. Read the original post and comments here.

This entry was posted in Uncategorized. Bookmark the permalink.