as in any other genre of science, research using qualitative
interviewing does not just happen by chance, but is designed
by researchers. What sets research interviews apart from everyday
conversations is the much greater extent to which they are
planned and reflected upon in advance—and subsequently analyzed.
In everyday life, we generally just follow the conversational
flows with the people we meet, but qualitative interviews are prepared,
conducted, analyzed, and reported according to some kind
of plan or what is normally referred to as a research design. Some
designs imply a tight structuring of the research process while others
are much more loose and flexible, but even the most stretchy
and negotiable design is still a design (which serves some purposes
better than others). In other words, my claim is that one cannot
not design an interview study, for even the conscious choice of
omitting any preparation and just talking to people around one
about something is also to have chosen a design (which, in most
cases, however, will probably not be conducive to knowledge
production).
In this chapter, I will discuss a number of issues that should be
taken into account when considering research designs in qualitative
interviewing. I outline a well tried-out step-wise approach
to design that is informative of what to do, when to do it, and
research design in
interview studies
2
46 : QUALITATIVE INTERVIEWING
why do this rather than that in the research process. I will introduce
some conceptual distinctions between inductive, deductive,
and abductive designs, and I shall refer to three examples
from paradigmatic interview studies to show more concretely
how different designs enable different research processes and
results. Some projects are designed for discovery and generally
demand a quite disciplined analytic awareness, while others
are designed more for understanding something and come
close to ethnographic research that seeks to take advantage of
whatever conversations emerge in the field that the researcher
is interested in. It is also possible to design a study in order to
construct something new (new practices or new kinds of public
discussions).
Going Through an Interview Study
Some approaches to interviewing focus almost exclusively on the
concrete encounter between an interviewer and an interviewee. In
contrast to this, I will here argue that a well-designed interview
project has a thread that runs through the entire process and connects
the research question with what goes on in the interview and
also with subsequent transcription, analysis, and reporting. Thus,
when designing an interview study, and also when readers evaluate
the appropriateness of different designs, it can be helpful to
think through the process in a step-wise manner. Here I break the
process down into four common steps: Preparation, interviewing,
analysis, and reporting.
These should not be thought of as discrete phases, but are generally
overlapping and cyclic, so that one may, for example, return
to one’s interviewees and conduct additional interviews after having
analyzed their initial statements, or even re-thematize the entire
project upon recognizing that the project has come off to a misguided
start and in fact concerns something other than one had first
imagined. Whether one prefers tight or loose designs, one should
always make the best of the flexible and inductive research logic that
normally guides qualitative research. This makes possible an iterative
design—a form of designing-as-we-go-along—which is normally
a vice in quantitative and experimental research, where the
99th research participant must be treated in exactly the same way
as the 1st participant in order to ensure reliability. But in qualitative
RESEARCH DESIGN : 47
research, it can be a virtue to amend one’s design in the process. It
would be foolish to continue with a bad interview guide that does
not result in valid answers, for example, if it turns out after a couple
of interviews that the guide is problematic. Of course, any amendments
should be carefully noted and reflected upon (and when
relevant, should also be mentioned in the final report), and it is
advisable for interview researchers to use a log book to keep track
of the decisions that are made throughout the process, as these are
often difficult to reconstruct correctly in hindsight.
Preparation
The first thing to consider when preparing an interview study is
to make clear what one wants to study. What is the theme that
one is interested in? People’s life stories, experiences, or actions?
The second thing is to consider whether qualitative interviewing
is suitable for the given research theme. In their book on designing
qualitative research, Marshall and Rossman (2006) argue that
there are three broad areas of study to which qualitative methods
can favorably be applied (p. 55):
Individual lived experience.•
Language and communication.•
Society and culture.•
Qualitative interviews can be used, and have been used, to
study aspects of all three, but they lend themselves most naturally
to the study of individual lived experience. In fact, when one
wants to know how an individual experiences some phenomenon,
interviewing has a certain primacy among the different methods.
Interviews can also be used to study language and communication,
since human beings use the interview situation itself to communicate
through language. But generalizing from communicative
processes in the interview situation to the broader world of human
communication is a thorny issue. If one wants to study “naturally
occurring talk” (e.g. how doctors communicate with patients),
it can be important to obtain naturalistic data rather than just
interviewing people about how they believe they communicate
outside the interview context. Finally, since society and culture
are co-constituted by conversational processes such as interviews,
48 : QUALITATIVE INTERVIEWING
qualitative interviewing remains a relevant method for studying
these aspects of the social world; but, again, it can be quite important
to use other sources of data (e.g. observations, documents,
cultural objects) to get a more complete picture. Qualitative interviewing
is very often a relevant method, but it is not always a sufficient
method vis-à-vis the phenomenon of interest.
It is not uncommon, however, to find that people have fallen
in love with interviewing as a method, and then seek to apply it
to answer questions that are ill-suited for this kind of method.
Judging from my experience as a teacher of qualitative interviewing,
it happens quite often that novice interviewers want to
know, for example, if it makes a difference (in relation to a given
subject matter) whether people are men or women, young or
old, or homosexual or heterosexual. In small projects, they then
recruit, say, two women and two men and ask them about their
opinions about something. There is nothing intrinsically wrong
with this, but the problem arises when the researchers want to
use this limited material to draw conclusions about general differences
between men and women (for example). For, obviously,
it is entirely possible that the selected men have opinions that are
stereotypically associated with women (and vice versa).
The problem is that interviewers very often have inherited a
research logic taken from experimental research and see this as the
scientific method, intent, as it is, to investigate differences between
groups. So they wish to have a “control group,” let us say of men, in
order to verify whether their understandings of women are valid.
This runs counter to a genuine qualitative research logic, and in
fact makes very little sense. If I want to make an interpretation
of the works of William Shakespeare, it would be quite strange to
demand that I use Homer or Dante as controls. Or, if I were to
conduct fieldwork in rural Russia, what sense would there be in
demanding that I use an urban area in the United States as a control?
Naturally, it can be very interesting to do comparative studies
in qualitative research (as the study described in Box 2.2 below
will exemplify), but this must be done in very careful, analytic
ways and is usually quite different from testing a hypothesis about
general differences between groups. Qualitative interviewers need
to be aware that qualitative research functions differently from
experimental research, and that the whole idea of using controls in
this way normally makes little sense.
RESEARCH DESIGN : 49
In qualitative interviewing, we should in general pose research
questions that contain a “how” instead of a “how much.” A
research question such as “How do young people experience being
admitted to hospital?” is generally preferable to the comparative
question “Are women more anxious than men when admitted to
hospital?” The latter question invites us to think in terms of causes,
effects, and control groups, and in order to answer the question in
a statistical sense, one would need a large number of interviewees.
A question like “How do people cope with the loss of a loved one?”
is in general better for qualitative projects than questions that seek
to find causal effects, such as “Does psychotherapy reduce the risk
of depression after a loss?” The latter question is interesting and
relevant, but it is also extremely difficult to answer with qualitative
interviewing. Instead, one would need to enlist a large number of
research participants, administrate standardized tests, and compare
the effects statistically in order to assess whether the findings
are statistically significant, i.e., not just a chance result. In most
cases, it is relevant to conduct qualitative interviews when one
wants to know about how people experience something, reason
about something, or act in relation to something. The result of an
interview study may well be (and often is) that people do so in different
ways (and the researcher can for example construct a typology
of ways of experiencing, reasoning, or acting in order to show
this), but this is different from saying that being a man/woman is
causing one to experience something in a given way.
Design questions that should be answered when preparing an
interview study are generally of five broad kinds: What should be
studied? Why is it relevant to do so? How should the subject matter
be studied? Who should be interviewed—and how many? I will
deal with each question in turn.
• What should be studied? The question of what should always
be addressed before the question of which methods to use. One
should employ the methods that suit the theme rather than skew
the theme to make it fit preconceived ideas about methods. As discussed
above, the strength of qualitative interviewing is its ability
to throw light on the hows of human action and experience: How
is something done (e.g. patienthood), and how is something experienced
(e.g. anxiety) can favorably be studied using qualitative
interviewing. So, in general, it is helpful to formulate one’s research
50 : QUALITATIVE INTERVIEWING
interest in terms of a list of hows. At this stage, it is also relevant to
reflect upon methodology in a more overarching way—for example,
is the researcher aiming to discover something unknown about
how people do X, to construct a better way of doing X, or to understand
how people experience X? (I return to the models of discovery,
construction, and understanding below). It is also relevant to
consider the philosophy of science that one adheres to (implicitly
or explicitly). There can be quite a difference between approaching
the human world as a range of conversational exchanges (a perspective
found in discourse analysis and conversation analysis, for
example), and conceiving of the human world as a reality that is
structured by intentional acts of human consciousness (a perspective
found in parts of phenomenology, for example).
• Why is it relevant to study this? The human world is rich and
varied, and it is possible to raise an endless number of questions
that can be answered using qualitative interviewing. But not all
questions are equally relevant, and I believe that there is an ethical
obligation to use the privileges one has as a researcher to study phenomena
that are relevant, and where there might even be a chance
that the results of the study may improve the world (however little
this may be). This is not to say that all research projects should be
directly relevant to a certain practice, for example, for basic research
also may be very enlightening for human beings. Sometimes it is
the case that what initially appears “useless” may turn out to be the
most useful. But most readers of interview reports will quite legitimately
expect the qualitative researcher to have given thought to
why this piece of research is relevant, and to whom it might be relevant.
As the critical psychologist Ian Parker has put it, as researchers
we are “always participating in the activity of either reproducing
the way the world is or transforming it” (Parker, 2005, p. 13), and
there can be quite a difference between designing for reproduction
and designing for transformation. Should the interviewees, for
example, be expected to be enlightened and lead better lives upon
having participated in the study? And what are the ethical challenges
in this kind of transformative research (cf. the discussion of
transformative interviewing in Chapter 1)?
Furthermore, qualitative inquiry has grown and expanded
and is now an enormous field, and it is very often the case that
other researchers have already studied the phenomenon that one
RESEARCH DESIGN : 51
is interested in. Has the researcher done an adequate literature
review that enables her to assess whether her own questions may
contribute with something new? “Because I can!” is not a good
answer when asked about why you want to study something, and
there is in today’s qualitative research—as in all other forms of
research as well—an enormous amount of repetitions that do not
teach us anything substantially new. The sad truth is that one specific
implicit answer to why it is relevant to study something is
dominant, viz. that it is relevant because the researcher has an aim
of getting tenure or living up to a publication pressure. In such
cases, it can be argued that the researcher could make better use
of her time by studying something that could actually make a difference
to people (and hopefully enable her to get tenure in the
process!).
Someone who has put a lot of emphasis on the relevance
aspects of human and social science is Bent Flyvbjerg (2001).
In a book on how to make the social sciences matter (i.e. how
to make them relevant), he has argued that the social sciences
must become phronetic, which means that they must conceive of
themselves as practical sciences that are involved in the societal
subject matters that they study. Phronetic researchers place themselves
within the context being studied and focus on the values
of the practices of communities by asking three “value-rational”
questions: Where are we going? Is this desirable? What should
be done? (p. 60). The raison d’être for the social sciences,
Flyvbjerg thinks, is developing the value-rationality of society,
i.e., enabling the public to reason better about its values and
social practices. This is just one approach to the why-question
of qualitative studies, which argues that, ideally, qualitative
research is valid when it enables people to improve the practice
that is studied.
As a bridge between “why” and “how,” it is also relevant to consider
the ethical aspects of interview research. Ethical aspects concern
both the why, as we have seen, and obviously also the how,
since research should proceed concretely in an ethically sound
manner. This involves asking questions about one’s own project.
These questions include:
Possible beneficial consequences of the study: Can the study
improve the lives of human beings in any way?
52 : QUALITATIVE INTERVIEWING
How to obtain informed consent. This can be a challenge in qualitative
interviewing, since researchers often develop and change
their focus in the course of the research process, sometimes
making it necessary to inform participants about changes in the
direction of the research process and to ask them for renewed
consent.
How to protect participants’ confidentiality. This is particularly relevant
in qualitative inquiry, where researchers often deal with
intimate aspects of people’s lives.
The consequences of the study for the participants: Is there a risk
that they can be harmed psychologically by taking part in the
study, e.g. by engaging in conversations about painful experiences
from the past?
In many countries, it is demanded that researchers obtain
approval from an ethics committee (often institutionalized at local
universities), but it is important not to reduce ethical issues to a
“check-list approach,” so that one simply goes ahead without further
reflection after having gained formal approval. In qualitative
research, unexpected ethical questions can easily arise, and it is
important to remain open to these questions throughout a project
rather than believing that ethical questions can be dealt with once
and for all before a project is initiated. In any case, most readers
of interview reports would like to know something about how the
researcher has handled the four ethical issues mentioned here.
• How should the subject matter be studied? The question of
how is clearly the biggest one, where both theoretical and methodological
questions must be raised. Should the interviews be
conducted to capture “lived experience” or should they be seen
as a form of situated interaction in their own right, for example?
Also, more concrete and practical questions must be raised, such
as: How can the research questions be translated into an interview
guide that makes sense to the interviewees?
The how questions cannot of course be answered abstractly, but
only concretely, depending on what one wants to find out by conducting
the study. The particular context and practicalities of the
study are here of paramount importance. If the researcher is asked
to deliver a result in two months and is given limited resources
for interviewing, transcribing, and analyzing the interviews, then
this is obviously very different from a large, well-funded research
RESEARCH DESIGN : 53
project that might go on for years and involve hundreds of interviewees.
Too many method books are concerned only (or mainly)
with ideal research situations, and do not take concrete situations
and barriers into account. But some of the most interesting
pieces of research seem in fact to have come from non-planned
situations that led to small-scale studies, when researchers simply
stumbled upon something that emerged as pressing to study
(see Brinkmann, 2012a, for examples and discussions of this kind
of research).
When dealing with the how, the researcher should first and
foremost consider whether interviewing is an appropriate way of
answering the questions that interest her. Surprisingly, this consideration
is often completely by-passed in reported interview
research. Ideally, a research report gives the reader a list of reasons
that explain why this particular method was the most relevant one
to employ. And, if the answer is affirmative, the researcher should
decide whether individual interviews or group interviews are preferable.
These different forms of interviews were covered in the previous
chapter along with the different media that may structure the
conversation (face-to-face, telephone, Internet). Suffice it here to
say that individual, face-to-face interviews still represent the standard
choice in qualitative interviewing, because of the interpersonal
contact, context sensitivity, and flexibility that enables interviewers
to take advantage of the research logic of qualitative inquiry.
At a more general level, the how-question also implies a decision
whether to work inductively, deductively, or abductively
(or whether to combine these modes). The following section is
adapted from Brinkmann, 2012a):
Induction is the process of recording a number of individual
instances (e.g. stories about what it means to learn something
new) in order to say something general about the given class of
instances (e.g. learning). According to traditional formal logic
(which is deductive), inductive inference is not strictly valid, for
even if the first 99 girls we have observed do in fact “throw like a
girl” (cf. Young, 1980), it may be the case that the 100th does not.
Nevertheless, qualitative research is most frequently characterized
as inductive, since researchers will often enter the field without too
many preconceived ideas to test, but will rather let the empirical
world decide which specific questions are worth seeking answers
to. Grounded theory is one well-known approach that seeks to
54 : QUALITATIVE INTERVIEWING
optimize the inductive process in qualitative inquiry (Glaser &
Strauss, 1967). Some methodologists almost identify qualitative
research as such with an inductive approach—e.g. Flick (2002,
p. 2), who talks about “traditional deductive methodologies,” by
which he means quantitative research, being superseded by more
adequate qualitative “inductive strategies.” But as we shall see later
in the chapter, it is indeed possible to work deductively in qualitative
interviewing, although it demands a different approach to
design. Inductive designs are particularly well suited to study new
and emergent phenomena, where it is premature to formulate specific
hypotheses.
Deduction is a phase in the knowledge-producing process of
deducing testable hypotheses from general ideas or theories, and
then seeking to falsify these. In philosophy of science, this theory
was famously developed as a general approach to the scientific
method by Karl Popper and was known as falsificationism. The
idea was that only those theories that result in hypotheses that are
in principle falsifiable deserve to be called scientific. This, according
to Popper, excluded Marx and Freud from the rank of scientists.
The deductive model may serve the natural sciences well, but
is less helpful as a general model in the human and social sciences.
The main problem with the deductive approach (and also with falsificationism)
is that in cases where empirical observations apparently
contradict one’s hypothesis or general theory, scientists often
will not know whether to reject their hypothesis or ignore their
observations (because they are methodologically weak, for example).
That said, researchers using qualitative interviewing can work
deductively and use single cases as a test bed for general theories,
following a deduction of the form “If this is (not) valid for this
case, then it applies to all (no) cases.” (Flyvbjerg, 2006, p. 230). To
give an example: If it turns out that even the most abstract forms
of human knowledge—mathematics—are situated and acquired
contextually, then we have reason to think that all forms of human
knowledge are thus situated (Lave, 1988). By studying an extreme
case of something, one may become able to deduce general consequences
for the entire class of the given something.
To give a more concrete example of how an interview project
can be based on a deductive design, we may mention Kvale’s classic
(1980) study of the classroom effects of grading. The study arose
in connection with a new Danish policy of restricted admission to
RESEARCH DESIGN : 55
college based on grade point averages from high school (see also
Kvale & Brinkmann, 2008, p. 108). The researchers decided to
interview Danish pupils and teachers about their experiences with
grading, and they formulated several hypotheses in advance such as
(1) Grading influences the process of learning and the social situation
where learning occurs, (2) the prevalence of the grading perspective
would increase with a restricted admission to college based
on grade point averages. These hypotheses were deduced from theories
that were prevalent at the time (particularly Marxist ones) about
extrinsic motivation for learning in a capitalist society (implying
that learning to obtain grades prepares young people to work primarily
to obtain an income). Operating with a deductive approach
like this demands particular care on behalf of the researchers not to
automatically verify the hypotheses (this danger is often referred to
as confirmation bias), and, in this case, the research group employed
independent coders of the material to ensure high reliability in
the interpretations and did many other things to avoid confirmation
bias. In any case, this example shows that it is possible to use
a deductive model when designing qualitative interview projects,
although inductive approaches are much more common.
Both the induction and deduction models normally work best
when researchers already know the phenomena that they are
studying in the research process (although the deductive model
demands much more specificity in this regard). It is tacitly presupposed
that we have some stable entity that we can study repeatedly
in a number of cases to build general knowledge (induction) or that
we already have general ideas from which we can deduce particular
consequences to test (deduction). But when we talk about the volatile
conversational world of human beings, this is often not the
case. Thus, a third kind of reasoning is needed, and fortunately we
have what is known as abduction, which is suitable when we wish
to study things that are emerging and as yet unknown.
Abduction as a form of reasoning is associated with the pragmatist
Charles S. Peirce. Peirce is often credited with being the
original pragmatist because of his formulation of what has since
been known as the “pragmatic maxim”: “Consider what effects,
which might conceivably have practical bearings, we conceive the
objects of our conception to have. Then our conception of these
effects is the whole of our conception of the object.” (quoted from
Bernstein, 2010, p. 3). According to Peirce, things are their effects.
56 : QUALITATIVE INTERVIEWING
Abduction is a form of reasoning that we employ in situations of
uncertainty; when we need an understanding or explanation of
something that happens or some effect. It can be formalized as
follows: (1) We observe X; (2) X is unexpected and breaks with
our normal understanding; (3) but if Y is the case, then X makes
sense; (4) therefore we are allowed to claim Y, at least provisionally.
As an example, let us say (1) that we observe a person who
waves her arms wildly. And let us say (2) that this is unexpected
in the context (the situation is not, for example, an aerobics class).
We can then conjecture (3) that an aggressive wasp is attacking the
person. This would make the person’s behavior understandable,
even expected, and therefore (4) we infer that this is the case (at
least until we arrive at a better interpretation).
As this example testifies, abduction is a very pervasive form of
reasoning in everyday life. And it is likewise widespread, although
more implicitly, in interview studies. In most, if not all, forms of
qualitative inquiry, there is an abductive aspect, especially connected
to (3), which we may refer to as the creative moment in
the analytic process. This is when researchers employ their sociological
imagination (Mills, 1959) and develop conjectures about
how to understand something, which they then test in practice
by looking at evidence for and against (this will be exemplified in
Box 2.1, where the researchers develop five different conjectures to
explain a given phenomenon). From the abductive angle, research
is never finished, as the human world itself is never finished, but
constantly in the making. Designing interview studies abductively
thus means designing for dialoguing with an evolving reality of
persons in conversation rather than attempting to formulate theories
that are universally true.
To sum up, it is often relevant for readers of interview reports
to know whether the study was conceptualized inductively, deductively,
or abductively. An inductive approach demands careful exposition
of the theme being investigated and a close description of the
steps taken from data generation to formulation of general patterns,
types, or ideas in the material. A deductive approach has as its key
issue how to design in a way that minimizes the risk of confirmation
bias, the tendency to have one’s hypotheses confirmed. Finally, with
an abductive approach, it becomes imperative to justify and check
the interpretive conjectures that are voiced by the researcher. Some
studies, of course, successfully combine the different approaches at
RESEARCH DESIGN : 57
different stages in the research process (like the study described in
Box 2.1 below, when the researcher, Janice Morse, began with an
open and inductive approach and, upon discovering a specific phenomenon,
developed different hypotheses abductively that were
tested in a comparative design with deductive elements).
• Who should be interviewed? This is the question of selection
and sampling. According to Roulston (who follows LeCompte and
Preissle), one needs to draw a distinction between selection
and sampling (Roulston, 2010, p. 81). Selection refers to the
general decisions concerning who should be in focus in the study
(e.g. adults suffering from depression) and sampling refers to the
process of finding a subset of the population that has been selected
as relevant (e.g. 20 depressed adults, an equal number of women
and men, recruited from Clinic X in Y-ville, representing “adults
suffering from depression”). In most quantitative studies, the goal
is to obtain a representative sample, which may enable researchers
to generalize from the sample to the general population. This can
also be a goal in qualitative research, but because most qualitative
projects aim for thorough analyses in depth—rather than
larger and broader analyses—they often employ other sampling
strategies.
Sampling becomes a particularly pertinent issue in case-study
research, because researchers study just one single case, and
Flyvbjerg (2006) discusses a number of different ways of selection,
based on different interests.
Random selection can be employed to avoid systematic biases
in the sample (here the size of the sample is decisive for generalization,
but this is often not relevant for qualitative studies). In
general, random selection as a conscious choice is employed only
in quantitative projects.
Information-oriented selection is normally more relevant in
qualitative inquiry. The goal is here to “maximize the utility of
information from small samples and single cases. Cases are selected
on the basis of expectations about their information content”
(Flyvbjerg, 2006, p. 230). This means that the researcher’s knowledge
about the field becomes relevant. With information-oriented
selection, the researcher can choose to look for (1) extreme cases
in order to be able to say something about the phenomenon in
its purest form (e.g. adults suffering from severe depression),
58 : QUALITATIVE INTERVIEWING
(2) maximum variation cases in order to obtain information about
the significance of different and perhaps opposing circumstances
(e.g. adults with mild versus severe depression), (3) critical cases in
order to obtain knowledge that allows for deductions and falsifications,
which were discussed above (e.g. “if X is found among most
people with mild depression, we have reason to believe it will be
found among everyone who suffers from depression”), and (4) paradigmatic
cases that look for the typical in order, as Flyvbjerg says,
to “develop a metaphor or establish a school for the domain that
the case concerns” (p. 230). Sometimes qualitative interviewers do
not have the luxury of choosing a sampling strategy, but must stick
to the respondents that they are able to recruit. Like other forms
of selection and sampling, the consequences of this should also be
reflected upon in the research report.
Regardless of how one ended up with one’s groups of participants,
the process of selecting, sampling, and recruiting the participants
should be described. Readers of interview reports will also
expect the researcher to reflect on possible limitations brought to
the study because of the actual group of participants. What does it
mean if the participants were “self-selected” as volunteers? Does it
matter if the researcher accepted “who she could find” as participants,
without being able to select among them? Does it matter if
the group is skewed in terms of gender, social position, or ethnicity
for example?
• How many interviews need to be conducted? This is arguably
the most typical question raised by interviewers at research
courses, but also readers of interview reports will often ask whether
the number of people interviewed was sufficiently high (normally
they do not ask whether it was sufficiently low, although this question
may also be relevant). People frequently ask this question with
a quantitative logic in mind: The more interviews, the more valid
and reliable the analysis will be. But this is rarely the case. As Kvale
has said, the only logical answer to the question “How many interviews
should I conduct?” is: “Interview as many subjects as necessary
to find out what you need to know.” (Kvale & Brinkmann,
2008, p. 113). If the goal of one’s study is to find out how it is to
be Barack Obama, then it might be sufficient to interview just this
one person. If the researcher has given careful thought to how to
select interviewees, a small number of interviews may be enough
RESEARCH DESIGN : 59
to answer one’s research question. If the point is, for example, to
test whether a supposed general feature exists in some population,
then it might suffice to interview a few critical cases (e.g. those
cases where it is least likely to find the feature, and if it is found
there, it is likely to be found everywhere).
Normally, fewer interviews that are thoroughly analyzed are
preferable to many interviews that are only superficially explored.
It is always relevant to bear Harry Wolcott’s maxim in mind: “Do
less, more thoroughly” (Wolcott, 2009, p. 95). Qualitative interviewing
distinguishes itself by its ability to get close to people’s
lives, not by including a huge number of participants. One cannot
get close to the lives of 50 or 100 people in an interview study. If,
for some reason, such a large number of participants is needed,
a survey would possibly have been better and more economical.
And if the study has included 50 participants, but only the voices
of a handful of people are reported (which is not unusual), then
the reader easily becomes skeptical: What happened to all the
other people who were interviewed? Did their words not matter
to the researcher? As a rule of thumb, it can be said that interview
studies tend to have around 15 participants, which is a number
that makes possible a practical handling of the data (although l5
interviews of 20 transcribed pages equals 300 pages to be analyzed,
which is quite a bit). The aim is not statistical representativeness
(although it can be, e.g. in mixed methods studies), but instead
the chance to look in detail at how selected people experience the
world.
Interviewing
The preparation phase, with its many considerations about theme
and research approach (induction, deduction, abduction), should
also include a review of extant literature and normally ends
with the creation of an interview guide, which is also sometimes
referred to as an interview protocol (Rubin & Rubin, 2012). The
guide translates the research questions (e.g. “How do young people
in late modernity experience transitions?”) into questions that
can be posed to interviewees in a language that makes sense to
them (e.g. “Could you please describe what happened when you
moved away from your parents’ house?). Some interviewers prefer
a simple list of questions in a specific order, whereas others
60 : QUALITATIVE INTERVIEWING
prefer a page with two columns, one with research themes on one
side and another with interview questions that reflect the different
themes on the other side. This makes it possible for the interviewer
to get an overview of where she is in the conversational process
and likely ensures that all relevant themes are covered. It is preferable
to memorize the guide as much as possible in order to be able
to maintain eye contact with the interviewee. This also facilitates
a flexible approach to the order of the questions, and may allow
the interviewee to cover something that the interviewer had only
expected to touch upon later in the conversation.
Interviewers should think about whether a receptive style or a
more assertive style is preferable. For sensitive and personal topics,
a supportive, receptive, or responsive approach is often helpful
(Rubin & Rubin, 2012). In their introduction to responsive
interviewing, Rubin and Rubin emphasize flexibility of design
and highlight the interviewer’s acceptance of what interviewees
say, along with a need for adjusting “to the personalities of both
conversational partners.” (p. 7). On the other hand, if the goal is
to study how people justify their beliefs, deliberate about difficult
matters, or give accounts of their opinions, a more confrontational
style may be required, which demands particular ethical
sensitivity in order to ensure that the conversation is conducted
respectfully.
Different styles of interviewing were covered extensively in the
previous chapter, and I shall not repeat myself here, but merely
emphasize once more that it is preferable to create some sort of
alignment between one’s research interest, interview style, and
the kind of analysis that one expects to carry out. For example,
if one’s research interest is to capture illness narratives, then it is
important to create a corpus of stories from the participants that
lend themselves to narrative analysis. It then becomes important
to ask interviewees to produce narratives, which can be done quite
simply by asking “Could you tell me the story of what happened
when you received the diagnosis?” Small linguistic guides (e.g. to
“tell the story” instead of “describe the situation” or “reflect upon
the meaning of...”) often prove to be immensely important when
the material is to be analyzed in the next step. Likewise—even if it
may sound trivial—if the goal is to analyze the phenomenological
essences of particular experiences, then it is pertinent to ask for
concrete descriptions; or if the research interest concerns people’s
RESEARCH DESIGN : 61
account-giving practices, the interviewer should not forget to ask
people to give accounts, such as by justifying opinions or answering
other why-questions.
Analysis
When the interviews have been conducted, a more focused analytic
phase begins. Like the other phases, analysis is not reserved
to a post hoc interpretation of transcripts, because the analytic
task already begins during the interviews, e.g. when interviewers
attempt to understand and interpret what the interviewees are
trying to say. It is very common that interviewers summarize a
narrative or description and ask the participant for verification or
further reflections. This was the fourth element in the conversational
flow illustrated in Box 1.1 in the previous chapter. Doing
this is, in a rather simple way, already beginning to analyze the
statements by trying to achieve a form of interviewee validation
in situ.
Also, the process of transcribing the recorded conversations
should be thought of as part of the analysis. Transcribing necessarily
means translating from one medium (the spoken word) to
another (the written word), and researchers should think about
how they are going to transcribe early on in the process. Many
different approaches to transcription exist. These include very
detailed conversation analytic approaches such as Gail Jefferson’s,
which demands the marking of overlap between speakers, emphasis,
volume, delay, and so on, and which is very time-consuming;
verbatim transcriptions that may include laughter, hmms, and
breaks; and reconstructive transcriptions that “polish” and provide
order to the often messy utterances of the speakers. There is
no golden standard of transcription. Everything depends on the
purpose of one’s investigation and on what is possible in practice
(what resources in terms of time or salary for assistants are
available). But it is obvious that if one’s analysis concerns the fine
machinery of turn-taking, or how the form of speech shapes the
meaning of what is said, then there is a need to transcribe the finer
details of talk, whereas a more rough transcription might be in
order if the purpose is to study the life stories of the participants.
In any case, to transcribe is always to analyze in the original sense
of analysis (literally “to break down into units”).
62 : QUALITATIVE INTERVIEWING
Not all researchers transcribe the entire empirical corpus. Some
prefer to work directly with the sound recording, which can be
coded in most contemporary software programs for qualitative
analysis, and some transcribe only selected portions of the corpus.
After transcription, a more focused analysis of the material
can be carried out, and here the options are legion and depend on
the philosophical and theoretical position of the researcher and
obviously also on the purpose of the study. Again, we may use the
distinction between inductive, deductive, and abductive strategies
to describe three broad approaches to analysis:
Induction in its different varieties is the most widespread
approach to analysis. Some qualitative researchers talk about analysis
as “analytic induction,” which, in the broadest sense, refers to
“the systematic examination of similarities within and across cases
to develop concepts, ideas, or theories.” (Pascale, 2011, p. 53).
Analysts using this strategy will inductively code data to identify
patterns and formulate potential explanations of these patterns.
So, a key component of analytic induction is coding. Coding can
be either concept-driven or data-driven. Concept-driven coding
uses codes that have been developed in advance by the researcher,
either by looking at selected portions of the material or by consulting
the existing literature. Data-driven coding implies that the
researcher starts out without codes, and develops them upon reading
the material. In principle, anything can be coded depending on
the research interest. Gibbs (2007) suggests the following examples:
particular acts, events, activities, strategies, states, meanings,
norms, symbols, level of participation, relationships, conditions or
constraints, consequences, settings. Also reflexive codings can be
used that record the researcher’s role in the process (pp. 47–48).
Coding also plays a significant role in the inductive methodology
known as grounded theory, originally developed by Glaser and
Strauss (1967). Grounded theory is an inductive strategy for theory
development without a prior theoretical framework. Many grounded
theorists work with open coding in a process of “breaking down,
examining, comparing, conceptualizing and categorizing data”
(Strauss & Corbin, 1990, p. 61). Grounded theories are developed
through the use of conceptualization to bind facts together, rather
than through inferences and deductive hypothesis testing. Since the
creation of grounded theory in the 1960s, it has branched in many
RESEARCH DESIGN : 63
different directions, including the more constructivist position represented
by Charmaz (2011) and the postmodern variant known as
situational analysis, developed by Clarke (2005). Charmaz makes
clear that grounded theory is inductive at its core and will proceed
with analysis by comparing data with data (developing codes),
comparing data and codes (developing tentative categories), and
developing categories into overarching concepts that are compared
with (other) theoretical concepts. These are, roughly, the analytic
stages recommended in grounded theory. But Charmaz notes that
grounded theory also has an abductive component since it highlights
the importance of being surprised in the development of codes, categories,
and theoretical concepts. So, as always in qualitative inquiry,
it can be fruitful to mix the different analytic strategies.
A final example of an inductive approach to analysis is
empirical phenomenology, which may serve as an example of
experience-focused analysis, because of its ambition to study the
essential structures of conscious experience. Phenomenology
sometimes applies inductive analysis as a kind of meaning condensation
(Kvale & Brinkmann, 2008, p. 205). This refers to an
abridgement of the meanings articulated by the research participants
into briefer formulations. Longer utterances are condensed
into shorter statements in which the main sense of what is said is
rephrased in a few words. This technique rests on the idea in phenomenology
that there is a certain essential structure to the way
we experience things in the life world (see the previous chapter),
which is what constitutes an experience as an experience of a given
something (shame, anxiety, love, learning something new etc.).
An even more specific approach to phenomenological analysis
has been developed in a psychological context by Amedeo Giorgi
(e.g. Giorgi & Giorgi, 2003). Giorgi breaks the analytic process
down into four steps: (1) Obtain a concrete description of a phenomenon
(through an interview) as lived through by someone.
Read the description carefully and become familiar with it to get
a sense of the whole. (2) Establish meaning units in the description.
(3) Transform each meaning unit into expressions that communicate
the psychological sense of the data. (4) Based upon
the transformed meaning units, articulate the general structure of
the experience of the phenomenon (p. 170).
While grounded theory is an analytic technique that is relatively
independent of specific theoretical perspectives on human
64 : QUALITATIVE INTERVIEWING
life and experience, phenomenology is in a way a complete package
of theory, philosophy of science, and methodology (see e.g.
Langdridge, 2007). The advantage of using such a “package” is
that peers will know what to expect from an analysis since there
is much agreement concerning which questions to pose in an
analysis as a researcher (and to the analysis, if one is the reader
of it). The disadvantage is the rather constricted and standardized
format, which may limit the creative development of qualitative
analyses. Today, many qualitative researchers prefer to move freely
between different methods of analysis. This, however, usually
demands a more careful description of their analytic procedures
so that readers have a chance to evaluate the validity of the possible
idiosyncratic work.
Deduction in the analytic phase can involve the use of hypotheses
derived from theory in an interpretive process. In Box 2.1, we
shall soon see an example of how hypotheses can also be derived
from the empirical material itself and tested in a comparative analysis.
Herbert Blumer once referred to theoretical concepts as sensitizing
instruments that researchers use as tools to be able to look
in fruitful directions and helpful ways (see Clarke, 2005, p. 28).
Some qualitative researchers, e.g. those working on the basis of
philosophical hermeneutics (Gadamer, 1960), believe that we cannot
understand anything without prejudices in the literal sense of
pre-judgments. There is no such thing as understanding something
from nowhere, without presuppositions, for we always need
some interpretive framework in order to distinguish significant
from insignificant aspects of the material. Some of these frameworks
can be formulated explicitly as theories. Psychoanalysis is
one such famous theory that enabled its practitioners and theorists
to see and understand something that was not visible without
the sensitizing concepts of psychoanalysis (e.g. repression, defense
mechanisms, Oedipus complex, etc.). On a less paradigmatic level,
many researchers today approach their empirical material analytically
with theoretical concepts drawn from narrative theory (e.g.
story line, plot, protagonist, antagonist, etc.). In both cases, it is
possible to deduce hypotheses from general theories that can assist
in the analytic process of reading and interpreting the data.
This kind of deductive analytic strategy is very often criticized
for its confirmation bias, which I also discussed above. Critics argue
RESEARCH DESIGN : 65
that analysts will find whatever the theory posits. This, however, is
hardly an issue that is unique to qualitative research, but can be said
to be a universal human tendency. Fortunately, a number of strategies
exist to counter this tendency, e.g. to play the devil’s advocate
against one’s own interpretations. If this is done sincerely, rather than
just as window-dressing, it can lead to new and exciting perspectives
on the materials. Flyvbjerg (2006) cites a number of social scientists
who have argued that qualitative case studies may often lead to a
refining of preexisting theory or even to discarding general theories
that turn out not to hold when confronted with empirical realities.
Flyvbjerg refers to this as case studies functioning like “black
swans,” borrowing the well-known example from logic that general
statements (such as “all swans are white”) can be falsified just by
finding a single instance that contradicts them (e.g. by finding a species
of black swans in Australia). If the analyst meticulously shows
the reader that care has been taken to avoid confirmation bias, then
the ensuing text may become very persuasive, and it often results in
a highly readable product if the researcher constructs the text like
a series of challenges to her own interpretations that are discussed
in turn. A further sign of quality is seen when researchers present
several different interpretations of the phenomenon under scrutiny
rather than just sticking to a single one. This can be achieved by
working with more than one theoretical framework, leading to different
sets of “sensitizing concepts” that may bring forth different
aspects of the material in the analytic process.
Abduction in the analytic phase works from breakdowns in the
understanding of the analyst. The researcher will look for breaks and
contradictions and other matters that somehow “disturb” the common
understanding or convention. Some interview researchers, who
work abductively in a broad sense, look in particular at the social
practice of interviewing itself as the key to open up for analysis.
Roulston (2011) has argued that there is much to learn from “failed”
interviews, i.e., from interviews where things “go wrong” according
to textbooks on interviewing and conventional wisdom. In Table 1.1
from the previous chapter, which drew a distinction between two
conceptions of interviewing—as a research instrument and as a
social practice—we can say that Roulston encourages interviewers
to pay close attention to the nature of the conversational interaction
itself and to look for misunderstandings or other breaks in
66 : QUALITATIVE INTERVIEWING
the conversational flow. Aspects that stand out as strange may often
prove to be valuable to understanding how talking about the subject
matter in a specific way constructs what we may know about it.
As one example of this, we may mention Tanggaard (2007), who
did a research project on learning in a vocational school and conducted
many interviews with students. The researcher conceived of
learningasembeddedineverydayactivities,whereasitwasmadeclear
from interviews with vocational students that learning—according to
the learners’ perspective—was something that took place in a school.
According to Tanggaard, this led to “discourses crossing swords” in
the interviews, implying a struggle about how to define learning. The
interviews did not appear as smooth and responsive, but rather as full
of breaks, misunderstandings, and even antagonisms. This was made
clear to Tanggaard only upon reading the transcripts and employing
what I here have called abductive reasoning. It was her readings of
Foucault’s theory in particular that enabled the articulation of ideas
about how to make sense of the struggle of the conversationalists.
In her study, the opposition between the speakers—interviewer and
interviewee—made it clear that “learning” is not a simple thing, but
is a multi-perspectival phenomenon. The “what” of the conversation
(the subject matter) cannot here be separated from the “how” of the
situated interaction of interviewing.
Thus, unlike analytic induction, thematic analysis, grounded
theory, and phenomenology—all of which aim to capture the
lived experience as reported in interviews—analyses that look at
interviews from an abductive angle in this sense will more often
be discourse-focused and will treat interview data as topics (rather
than resources) and analyze them as accounts occasioned by the
situation (rather than reports about past experience). Needless
to say, it is often helpful for readers of interview reports to learn
how the analysts have treated the data in the process of analysis: as
accounts, as reports, or both.
Reporting
The final step of an interview project is the reporting of the
results. Like with the other steps, reporting cannot be treated
as a discrete stage, but is in its own way important throughout
the process. Ideally, interviewers should proceed with preparations,
interviews, and analyses with the final end product—the
RESEARCH DESIGN : 67
report—in mind. For qualitative research, it is the case that analysis
and reporting in particular often melt together. Writing is
“a method of inquiry” (Richardson & St.Pierre, 2005) throughout
an interview study. Writing is a central way for qualitative
researchers not just to report some findings, in the final instance,
but also to experiment with analyses, compare different perspectives
on the empirical material, and try out a number of
alternative ways of presenting readings of the material. Writing
should therefore be treated as an intrinsic part of the methodology
of interview research and not as a final “postscript” added
on at the end.
Since reporting is treated in great detail in the following chapters
on writing up the methods section and writing up the research
findings, I shall leave it here and encourage the reader to continue
reading! The most important thing to bear in mind in relation to
design questions is that the well designed project reflects on how
to report at the outset. If the goal is to write a single short article,
care should be taken not to obtain too much material to analyze,
but if the goal is to write a long and thorough book, matters
are very different. One should never ignore such practicalities,
since the form of a good report supports the content of what the
researcher is trying to say.
Designing for Discovering, Constructing or Understanding?
The four stages of preparation, interviewing, analysis, and reporting
are characteristic of most interview projects across subject
matters and theoretical paradigms. The underlying goals and
research interests of different projects may, however, be quite different.
I shall here reduce the many possible research interests to
just three and provide an example of each. I will invoke a distinction,
which initially looks rather abstract and philosophical, but
which may have quite concrete implications for the research process.
For although there is no direct coupling between kinds of
design on the one hand and research interests on the other, it is
often the case that tight, preset designs are meant to maximize the
chances for the researcher discovering hitherto unknown features
of reality, whereas more flexible designs are meant to facilitate the
researcher’s better understanding of something. In addition, it is
also possible to design in a way that involves constructing some-
68 : QUALITATIVE INTERVIEWING
thing new. In short, qualitative interview researchers should know
(and communicate to their readers) whether they are trying to:
• Discover something that they do not know.
• Construct something that they (or their requestors) would
like to see happen.
• Understand something that they do not understand.
Concerning the relation between research design and research
interest, we may refer to the great hermeneutic philosopher Gadamer
(1960), who argued that the process of understanding something
cannot be codified into methodological rules to be specified prior to
the research process. In order to make sense of Gadamer’s argument,
you may ask yourself the following: What methods do you use when
reading, and hopefully understanding, this book? What methods do
you use when you try to understand the people you meet and talk
to? In general, according to hermeneutic scholars, we do not employ
methods in such cases, and it is misguided when some qualitative
methodologists pretend that certain methodological procedures
will guarantee good and insightful research by themselves. What
we do in order to understand people (a primary aim for qualitative
researchers) is spend time with them and talk to them. So, from a
hermeneutic angle, it will sometimes be the case that a strict design,
based on formal and standardized methodological rules, will not be
helpful when understanding is the aim. This, however, should not
lead us to discard design questions. Hermeneutically oriented interview
research is also designed (albeit not necessarily in a methodologically
strict manner), and hermeneutic interviewers, inspired by
Gadamer, should nonetheless think through the steps of an interview
project as outlined above.
In a recent book, Martin Hammersley (2011) discusses the
three different models of research: The discovery model, the construction
model, and the hermeneutic or understanding model.
They correspond to three different research interests. The discovery
model is probably closest to people’s intuitive ideas about scientific
research: Its rationale is to discover something new; just as
physiologists have discovered the functions of organs and astronomers
have discovered new planets and stars. In 20th century philosophy
of science, however, the discovery model gradually fell out
of fashion, because arguments were voiced (e.g. from the aforementioned
Gadamer) to the effect that human and social science
RESEARCH DESIGN : 69
phenomena are not independent of researchers in a straightforward
manner. What we “see,” in qualitative research, are historical
and cultural phenomena (e.g. patterns of feeling, thinking, talking,
and acting) that are constituted by human activities (discourses,
symbolic interactions, etc.), including the very activity of gaining
knowledge about them. So, the argument went, it is misleading to
say that we “discover” them, for we always already have some kind
of implicit knowledge about them.
Even in light of such arguments, I believe that it is premature to
discard the discovery model. Even if there are valid philosophical
reasons to claim that the phenomena studied by qualitative researchers
are not independent of human activity (a claim I agree with),
I still believe that it makes sense to say—in a more everyday sense of
the term—that researchers can discover aspects of these phenomena.
They can, as Noblit and Hare (1988) once expressed it, “make
the obvious obvious” for us, and thereby discover something that
might have been there all along in our lives, but remained unnoticed,
perhaps due to its very pervasiveness. I have argued elsewhere
(Brinkmann, 2012a) that phenomenological approaches are particularly
adept at making the obvious obvious, and Iris Marion Young’s
paper “Throwing like a girl” is one of my favorite examples of what
wonderful and enlightening descriptions of the mundane may result
from using this strategy (Young, 1980). In the paper, Young shows
how boys and girls learn to move their bodies in quite different
ways, which cannot be accounted for in terms of anatomical differences,
but concerns socialization. When reading her analysis, one is
likely to react with a feeling of recognition; she describes something
we knew all along, but which we did not know that we knew!
For interview projects that aim for discovery, it is very important
to design the study in a way that allows the phenomena to
appear in a way that is not controlled by, or a simple artifact of,
the researcher’s actions. To paraphrase the words of sociologist of
science, Bruno Latour (2000), one must make sure to “allow the
objects to object” to what researchers say about them and do to
them. Only in that way can we attain a level of objectivity that
allows for genuine discoveries. In other words, one must discipline
the researcher’s activities, and this is done most directly by methodology
that may make the research procedures more transparent
to readers. Below, in Box 2.1, I shall present an example of an
interview study that discovered something new about the human
world, which was facilitated by a careful methodological design.
70 : QUALITATIVE INTERVIEWING
Box 2.1 DiscoveryThrough Qualitative Interviewing
When preparing and researching for this book, I sent e-mails to
leading scholars in the field of qualitative interviewing, asking for
exemplars of excellent interview projects and publications. In
her kind reply, Janice Morse directed my attention to a study of
burn patients who had experienced agonizing injuries (Morse &
Mitcham, 1998). Morse is the founder of the International Institute
for Qualitative Methodology at the University of Alberta, which is
a leading qualitative research institution in North America, and
she has also founded the journals Qualitative Health Research and
the International Journal of Qualitative Methods. She has thus been a
very central figure in qualitative research for decades, in particular
qualitative health research, and is extremely experienced
as reviewer, editor, and author of papers based on qualitative
interviewing.
The study was done by Morse herself together with a colleague,
and in her e-mail she explains why she believes it is important. She
describes an interesting process of analyzing the data that eventually
led to the published paper. In the process, she almost stumbled
upon the key to unlock the structure of the data and thereby
of the phenomenon: “While analyzing I walked away from my
computer, and when I came back I glanced at the screen and had
a wow moment suddenly noticing how disjointed the text was.
Everyday language took on a new meaning.” (E-mail, August 30,
2011). “The text,” referred to by Morse, is the textual material from
the participants’ statements. She continues: “I worked from this
single observation to other data sets, ‘testing hypotheses’ with
the already collected data. The moral of this story is that when
you hear the interview in the interview setting, that is one level of
analysis. Another may unfold when you work with the text.”
The study illustrates that qualitative interviewing can lead to
genuine discoveries. In this case, the discovery was about the language
use of patients with catastrophic burn injuries. The primary
research question concerned how such patients “get through” the
experience and cope with “resulting disabilities, losses of body
integrity, alterations in their former sense of self, and often the
death of other family members involved in the same accident.”
(Morse & Mitcham, 1998, p. 667).
RESEARCH DESIGN : 71
One aspect of coping with the agonizing physical pain identified
in the literature is disembodiment, a distinct distancing from one’s own
body. And, as Morse recounts, when analyzing the interview statements
of the patients, she discovered that they regularly referred
to parts of themselves as objects, using “it,” “the,” and “this” (e.g.
talking about “the left hand” rather than “my left hand”). In order
to interpret this peculiar use of language, which she had not been
aware of in the course of interviewing the participants, but only
noticed as in a flash on the computer screen, the researchers compared
the language use of burn patients with that of patients with
experiences of pain that stem from other kinds of injuries (e.g.,
spinal cord injuries).They found that other groups of patients (which
would be called “control groups” in quantitative research designs)
use disembodying language to a much lesser extent, if at all.
Methodologically, the interviewers used narrative analysis of
relatively unstructured, but focused, interviews to identify the
phenomenon. Initially, Morse took a phenomenological attitude
and displayed an interest in just letting the patients “tell their
stories” (Morse & Mitcham, 1998, p. 668). But, in a subsequent
analytic stage, and upon discovering the particular use of disembodying
language, she went beyond phenomenological description
to develop hypotheses (or “conjectures,” as they are called
in the paper), based on the patient narratives, to account for their
language use. This represents a deductive phase in the research
process, leading to five different hypotheses (ranging from disembodiment
being caused by loss of sensation, by loss of physical
ability, as something learned from physicians, as a means to
protect the self, and finally as a means of controlling overwhelming
pain). Out of the five different hypotheses, the researchers
found evidence only for the fifth and final one: Disembodiment
is a strategy used to remove the body part in order to remove
the pain, when it is overwhelming (p. 671). Furthermore, the
researchers found that later in the patients’ rehabilitation period,
patients again go back to using possessive pronouns when referring
to the self.
Unlike many other interview studies that claim to have identified
some novel phenomenon, the researchers in this case distinguish
themselves by also looking at negative cases, i.e. examples
of interviews when patients do not use disembodying language.
This definitely adds to the credibility of the findings, since one
72 : QUALITATIVE INTERVIEWING
The next models of research discussed by Hammersley are the
construction and understanding models. In a practical down-toearth
sense, the construction model builds on the idea that the
goal of qualitative research often is to create something new, e.g.
a new or improved social practice, or even new kinds of people
who have changed and developed as a consequence of taking
part in a qualitative study. In Chapter 1, this was presented
as the transformative model of interviewing. Qualitative interviewers
are increasingly becoming aware that interviewing, as
Briggs (2003, p. 497) has argued, is “a ‘technology’ that invents
both notions of individual subjectivities and collective social and
political patterns.” Different conversational practices, including
research interviews, produce and activate different forms
of subjectivity, and, utilizing some of the more activist forms of
interviewing discussed in the previous chapter, one goal of interviewing
can be to construct subjects in ways that allow for new
kinds of action.
gets a sense of researcher trustworthiness as a reader, and, as
the researchers explain, there is no reason to think that a negative
case invalidates a more general observation (Morse & Mitcham,
1998, p. 670).
All in all, we can say that the study conducted and reported by
Morse and Mitcham was designed in a way that enabled a genuine
discovery. There is generally no formula for how to discover something
new, and, as Morse recounts in her e-mail, it simply struck
her when looking at the computer screen that there was something
peculiar about the interviewee’s statements. Unlike other
designs that aim to construct a new kind of process, or understand
something that one has already identified, this study shows that
qualitative interviewing can lead to discoveries, even if the number
of participants included is relatively small (initially, six patients
with burn injuries were interviewed). The study further illustrates
that it is possible, and in this case desirable, to work with a form
of deduction in the analytic stage, i.e. qualitative “hypothesis testing.”
Conjectures about how to explain the phenomenon were
deduced from the material and tested in a process that involved
comparisons with the experiences and language use of other relevant
patient groups.
RESEARCH DESIGN : 73
In a more abstract epistemological sense, the construction
model is in many ways a direct answer to the perceived problems
of the discovery model. Scholars subscribing to the construction
model (e.g. Gergen, 2001, and other so-called social constructionists)
argue that knowledge is always constructed rather than discovered,
and that we can never know what or how things are in
separation from human activities. There are no “things in themselves,”
but only “things constructed by us.” In its strong form, as
Hammersley critically points out, social constructionism seems
to undermine the very possibility of knowledge, because the only
conception of knowledge available (to strong constructionists) is
one that presents knowledge as what most people believe. This
conflates knowledge and belief, and ignores the everyday necessity
of distinguishing between what is true and what it taken to
be true (Hammersley, 2011, p. 131). It also leads to the unhappy
consequence that a powerful group of people can create truths
by persuading or forcing others to believe certain things that are
in the interest of the powerful group. Furthermore, in its strong
form, constructionism leads to the consequence that most questions
about research design can be ignored, because all knowledge
is seen as a unique function of the steps taken to obtain it. In its
weaker forms, the construction model fades into the understanding
model, albeit with a greater emphasis on the fact that qualitative
inquiry is meant to change aspects of the social world. An
example of this is given below in Box 2.2.
Box 2.2 ConstructingThrough Qualitative Interviewing
One of the interview studies that I keep returning to is the classic
reported in Habits of the Heart: Individualism and Commitment in American
Life by sociologists Bellah, Madsen, Sullivan, Swidler, and Tipton
(1985). The empirical material for this study of North American
character and values consisted of interviews with more than 200
participants, some of whom were interviewed more than once. In
an insightful appendix to the book, the authors present their philosophy
of science as “social science as public philosophy.” They
reject the common view of the social science as “a disembodied
cognitive enterprise” (p. 301), and advocate instead a dialogical
role of the social sciences in which research functions to raise
74 : QUALITATIVE INTERVIEWING
important questions about values for society. In this way, the book
aimed not simply to represent aspects of US culture to its readers,
but to construct a discussion about where the United States are
going as a society.
In order to achieve this, the researchers conducted a special
kind of interview, which they refer to as “active interviews” (Bellah,
Madsen, Sullivan, Swidler & Tipton, 1985). Active interviews correspond
quite closely to the Socratic interviews that I introduced
earlier in this book (see the previous chapter). In contrast to the
interviewer as a friend or therapist, probing deep in the private
psyche of the interviewee, the active interviews were intended to
generate public conversation about societal values and goals. Such
active interviews did not necessarily aim for agreement between
interviewer and interviewee, and the interviewer was allowed to
question and challenge what the interviewee said. In one of the
examples cited, the interviewer, Ann Swidler, was trying to get
the respondent to clarify the basis of his moral judgments crystallized
in his statement that “lying is one of the things I want to
regulate”—and Swidler asked him why:
A: Well, it’s a kind of thing that is a habit you get into. Kind of
self-perpetuating. It’s like digging a hole. You just keep digging
and digging.
Q: So why is it wrong?
A: Why is integrity important and lying bad? I don’t know. It just is.
It’s just so basic. I don’t want to be bothered with challenging
that. It’s part of me. I don’t know where it came from, but it’s
very important.
Q: When you think about what’s right and what’s wrong, are things
bad because they are bad for people, or are they right and
wrong in themselves, and if so how do you know?
A: Well some things are bad because...I guess I feel like everybody
on this planet is entitled to have a little bit of space, and things
that detract from other people’s space are kind of bad...(Bellah,
Madsen, Sullivan, Swidler &Tipton, 1985, pp. 304–305)
Swidler challenges the respondent to examine why lying is wrong,
which is quite a hard philosophical question, and the final question
cited—concerning why wrong things are wrong—seems very
complex, and in standard textbooks on interviewing, the question
RESEARCH DESIGN : 75
could appear as an example of how not to pose an interview question.
The question is (extremely) abstract and invites high conceptual
reflection rather than concrete description. It very much
resembles Socrates’ questions in Plato’s dialogues.
The methodological appendix can be read as a very honest and
straightforward account of how a qualitative research group has
worked. There are no details about the specific analytical steps,
but a rich description of the discussions that the researchers had,
how they interviewed, and the kinds of philosophical literature that
inspired them. This is quite typical of studies that are designed
to construct a discussion, in this case of the question that is the
opening sentence of the book, “How ought we to live?” (Bellah,
Madsen, Sullivan, Swidler & Tipton, 1985, p. vii).
The study by Bellah and co-workers does not represent a strong
form of social constructionism, but rather a view of social inquiry
as already a part of the society it studies. There is no place outside
society from which to obtain an objective view of social processes,
so qualitative social science must instead seek to construct a better
social world by initiating discussions about society and its problems.
Bellah’s study also has great affinities with what Hammersley
calls the understanding model (which is actually the one favored by
Hammersley himself), and this is what I will focus on now.
The understanding model agrees with the critics of the discovery
model that there is no knowledge whose validity is simply
given (Hammersley, 2011, p. 132). But it simultaneously goes
against the radical constructionist conclusion that this implies
that knowledge can be constructed freely by human beings. We
are constrained, as Gadamer would say, by the historical horizon
of our interpretations, by the inescapable framework provided by
culture and history that constitutes our world. We understand reality
from where we stand, but there are still more and less accurate,
fruitful, and, valid ways of understanding from where we are. So,
in that sense, the understanding model assumes that knowledge
can be both perspectival and objective at the same time, and it also
implies that we cannot freely choose where to stand. We are situated
somewhere in the conversational world (cf. Chapter 1), and
that should not be seen as a hindrance to objective knowledge, but
as a precondition of it.
76 : QUALITATIVE INTERVIEWING
When working from the understanding model, qualitative
interviewers must design their research projects in a way that
enables what Gadamer called a “fusion of horizons” that leads
neither to a forgetting of one’s own perspective, nor to totalizing
the other that one seeks to understand. As the moral philosopher
and phenomenologist Lévinas would say, the goal, which
is at once ethical and epistemological, is to avoid the main ill of
Western philosophy, viz. the reduction of the other to the same
(Levinas, 1969). One should respect, and perhaps even celebrate,
difference, and yet try to understand it. Interviewers, who use
conversations to understand the lives of others, must therefore
reflect upon how their own background (standpoints, methodologies
etc.) affects their understanding, but they should not say
(like some constructionists) that this background determines
what they see.
Before summing up and concluding on design issues in qualitative
interviewing, I shall present an interview study that is quite
different from the ones presented in Box 2.1 and Box 2.2, and
which rested on a different kind of design. Although not articulated
in exactly these terms, the example presented in Box 2.3 is
of an interview study which, I believe, nicely illustrates a kind of
fusion of horizons, which was enabled by a very flexible design.
Box 2.3 UnderstandingThrough Qualitative Interviewing
My attention was directed to the study described here by Martin
Packer, author of The Science of Qualitative Research (Packer, 2011) and
editor of the journal Qualitative Research in Psychology, when I asked
him about examples of excellent interview studies. Packer gave
the following kind answer in an e-mail:
“I have not been looking out especially for good examples
of research interviewing, but one example that comes to mind
is Loïc Wacquant’s article “The pugilistic point of view: How
boxers think and feel about their trade” (Wacquant, 1995).This
is part of a larger study, and Wacquant has published almost
a dozen articles and a book. But this one focuses on his
interviews with boxers. I think there are some contradictions
in it (for example, I don’t believe that the boxers forgot
RESEARCH DESIGN : 77
he was a teacher), but he manages to convince me that
he is reconstructing an embedded and embodied form of
understanding.” (E-mail, August 18, 2011).
Upon reading Wacquant’s analysis, I must agree with Packer’s
verdict. The article indeed communicates an embedded and
embodied form of understanding. It does not, perhaps, report
on any new discoveries, nor was the study conducted to construct
changes in specific social practices, but it nicely illustrates how
qualitative interviewing can be designed for understanding and why
this may be a valuable thing to aim for.
Wacquant is a sociologist, born in France, who is working at the
University of Califonia, Berkeley. He collaborated closely with the
famous sociologist Pierre Bourdieu, and is an editor and co-founder
of the journal Ethnography. In his article here, he describes and reconstructs
the boxer’s point of view (which he calls the pugilistic point
of view, referring to boxing as pugilism). From August 1988 until
October 1991, Wacquant conducted participant observation at a
boxing gym in Chicago (on its South Side). He was here educated
in the art of boxing, and his career as a boxer culminated with a
fight at Chicago’s Golden Gloves tournament (which, alas, he lost).
Like Bourdieu, Wacquant wishes to develop a “carnal sociology”
that is not just a sociology of talking or observing selves, but
one that involves embodied people, situated in social practices.
Understanding the lives of people, from this “carnal” point of view
implies living a life that approximates that of the people one wants
to understand. In relation to interviewing boxers, it involves “taking
seriously what ordinary boxers have to say about their occupation:
how they think and feel about this harsh trade to which they are
willing to give so much, what virtues it holds for them, and how it
affects their life and self.” (Wacquant, 1995, p. 490). Wacquant thus
acts as a qualitative interviewer as an inherent aspect of learning
the craft of boxing and talking to his fellow apprentices. He wants to
obtain a view at prizefighting from the “inside looking out” (p. 490).
In the paper, Wacquant draws upon in-depth semi-structured
interviews with 50 fighters, comprising, he says, almost all professional
fighters in Illinois in 1991, and, in the course of the
research process, he ended up with 2,000 pages of transcriptions
(Wacquant, 1995, p. 493). Because he had been in the community
for almost three years before the formal interviews were
78 : QUALITATIVE INTERVIEWING
conducted, he was able to “phrase [his] questions in a manner
congruent with their occupational concerns and thus elicit candid
and meaningful answers” (p. 490). Unlike the majority of
today’s interview studies, which involve participants with whom
the interviewer just meets for an hour or so, Wacquant had prepared
extremely well by spending numerous hours, weeks, and
even years with the participants. So the interviews “were not the
product of a fleeting and superficial encounter but one link in an
extended chain of routine interpersonal exchanges.” (p. 494).
Wacquant’s article is full of detailed analyses of boxers, their
lives and practice, and is informed by an understanding of the
social situation of the boxers, most of whom “reside in segregated
and degraded neighborhoods where violent crime is a basic fact
of everyday life and where physical insecurity infests all spheres
of existence.” (Wacquant, 1995, p. 497). There is thus a combination
of a macrosociological angle and minute analyses of the
skilled bodily trade of boxing—the kinetic technique as he calls it
(p. 504)—and there is first and foremost an acknowledgement of
and respect for the fact that boxing can infuse the lives of boxers
“with a sense of value, excitement, and accomplishment.”
(p. 501). As one of the boxers, Henri (a black light-heavyweight)
said: “It’s a thinkin’ man’s game, but the outside doesn’t see that. The
on’y thin’ they see is jus’ two guys throwin’ punches, you know.
Well, uh, you gotta think about what you gonna do, when you gonna
do it, and how you gonna do it. See, this is what you gotta think
about.” (p. 503).
Toward the end of the article, Wacquant expresses his writing
ambition: to outline “a picture of the pugilistic planet as its main
inhabitants see it, or like to imagine it.” (Wacquant, 1995, p. 519).
He admits that the account is rather one-sided in its emphasis on
the virtues of prizefighting, but a certain ambivalence also creeps
in, e.g. when the interviewee Danny talks about his attitude to his
son’s boxing: “No, no, no fighter wants their son [to box], I mean you
could hear it, you hear it even in [Jack] Dempsey’s age: you never
want your son to fight—that’s the reason why you fight, so he won’t be able
to fight... It’s too hard, it’s jus’ too damn hard.” (p. 523).
In concluding on Wacquant’s study, we can say that his interviews
with the boxers reported in this article and elsewhere were
parts of a larger and longer-lasting ethnographic project of understanding
the world of boxing from the inside. The project was
RESEARCH DESIGN : 79
The philosophical remarks that I introduced above, about different
models of qualitative research designs, should not be taken to
imply that qualitative interviewers ought to become philosophers
(although a dose of philosophical reflection would rarely hurt).
However, it would be helpful if they lived up to a minimal philosophical
requirement, viz. to think through and make explicit what kind
of model (or combination of models) their design rests on. Only if
researchers do so can others evaluate whether the study is theoretically
coherent and designed in a way that makes its goal realistic. It
is certainly possible to work eclectically and combine research tools
from different models, but many interview studies simply proceed
with what looks like a phenomenological design of the interviewing
process (searching for essential structures of human experience),
combined with an oath to the philosophy of social constructionism
that denies reality to all essential structures of experience. This
is unreflective and illegitimate (and contradictory) eclecticism that
makes interpretations of the findings very difficult.
Conclusions
In this chapter, I have introduced a number of theoretical distinctions
that enable us to conceptualize different kinds of design. I
introduced a generic step-wise model that is good to keep in mind
when considering the design of qualitative interview studies,
thus designed for understanding something that many people do
not understand, and even look upon with contempt. Regardless
of one’s pre-understanding of boxing, I doubt that anyone can
maintain a simplistic view of boxing after having read Wacquant’s
descriptions. He has used his conversations with the boxers, aided
by his own life as an apprentice of the craft, to understand their
world—and his paper communicates this understanding in an artful
way to the readers.
With such an interview study, we are at the most flexible and
iterative end of the design continuum. The point is not that this is
a sloppy design, without prior specification of the number of interviewees
and preparation of an interview guide, but rather that it is
the most adequate kind of design when one is aiming at this form
of understanding.
80 : QUALITATIVE INTERVIEWING
involving the steps of preparation, interviewing, analysis, and
reporting, and I outlined three broad aims for qualitative interviewing:
Discovery, construction, and understanding. There are
many questions about design that are common regardless of the
aim of the research project—which I conceptualized as the what,
the why, the how, the who, and the how many—but there are also
some issues that relate more specifically to each of the three broad
aims.
Researchers aiming for discovery need to design their project in
such a way that they minimize the risk that what they discover is
simply an artifact of the study itself. This can be achieved by thinking
carefully about the way that the interviewing is conducted,
including an awareness of possible effects of leading questions, and
also by striving for high validity and reliability when transcribing,
coding, and analyzing the materials.
Researchers aiming for construction need a particular ethical
sensitivity about the way that they expect the world to change in
response to their research endeavors. Not all interviewees have an
interest in changing when taking part in a study, and not all communities
have asked for social change. Furthermore, researchers
need to account for why the research is of value when it is not primarily
an epistemic affair, seeking truth or understanding.
Finally, researchers aiming for understanding, who often need
to work with a more flexible design and to constantly accommodate
to the vagaries of the field they are studying, need to think
about how they can document the different twists and turns that
their research project undergoes. If Gadamer is correct that understanding
is non-methodical, then they cannot rely on methodological
specifications but must find other, more descriptive ways,
of capturing and justifying the knowledge-producing process.
I also introduced three models of reasoning in qualitative
research, viz. induction, deduction, and abduction. Roughly, these
models correspond to designs that are data-driven, theory-driven,
and breakdown-driven (this will be explained more fully in the
next chapter). I tried to show how these models may play a role
when designing a qualitative interview project, and also that they
have particular significance in the analytic phase of research. Three
examples were brought in: Janice Morse’s study of burn patients,
Robert Bellah and colleagues’ study of North American values,
and Loïc Wacquant’s study of boxers.