60
Openness to transformation means openness to the local contingencies
that complicate one’s agenda and may even force one to reset or
abandon one’s priorities.The contingencies of field inquiry are not to
be viewed only as obstacles to one’s inquiries but as opportunities to
learn which inquiries are the ones that really matter.These contingencies
should be celebrated, for they are where all real discoveries lie
(Liberman 1999: 50).
‘Can you help me? I am conducting a qualitative study and have
collected all my data on topic X and I’m not sure what to do with
it now. How should I analyse it?’ This is my version of an occasional,
but regular, request that appears on some of the qualitative
email lists I subscribe to.This sort of request typically comes from
a novice qualitative researcher, newly subscribed to the list. The
answer to this question is that if you have been collecting your data
carefully you have already begun to analyse the data. Hopefully the
person has already begun to think about what they are interested in
and what issues those data have raised.This provides the beginning
of data analysis.
Data analysis in most qualitative research begins during data
collection. This practice is consistent with the theory of data
analysis discussed in Chapter 1 that emphasised the dialectical, or
hermeneutic, relationship between theory and data. This chapter
reviews the practicalities of integrating data analysis and data
collection. Many texts on qualitative data analysis begin their
discussion with what to do after data have been collected. If data
3
Data analysis during data
collection
QualitativeAnalysis 9/11/01 1:37 PM Page 60
analysis begins only after the data have been collected, researchers
will have missed many valuable opportunities that can
be taken only at the same time as they are collecting their data.
This is particularly the case if you are using the methodology
of grounded theory. However, it also applies more generally to
most other research methods that are interpretive, inductive and
exploratory.
Waiting until after data have been collected to begin data
analysis can lead to some significant problems during data analysis.
If researchers leave the decision about what sort of data analysis
they want to conduct until after they collect their data, they may
have precluded, or made difficult, certain types of data analysis.
For example, it is difficult to conduct a narrative analysis if
the researcher asks short, directed questions that cut off the
interviewee. Further, during data analysis the researcher will typically
discover and notice unanticipated issues that have arisen early
in the data collection. If data analysis is left until afterwards these
issues will not be noticed during data collection; they will therefore
not be pursued during the data collection and cannot be pursued
in any depth during the data analysis.
This chapter does not attempt to review all possible methods of
data collection. There are numerous books that have described
these processes well, and interested readers are referred to books
such as Kellehear (1993), Mason (1996), Daly et al. (1997),
Denzin (1997) and Rice and Ezzy (1999). Rather, this chapter
focuses on conducting data analysis while the data collection is
being undertaken.
Integrating data collection and data analysis
The integration and interpenetration of data collection and data
analysis is practised by a number of qualitative research traditions,
including ethnography (Rosaldo 1989), participatory action
research (Nelson et al. 1998) and grounded theory (Strauss 1987).
Simultaneous data collection and data analysis builds on the
strengths of qualitative methods as an inductive method for building
theory and interpretations from the perspective of the people being
studied. It allows the analysis to be shaped by the participants in a
more fundamental way than if analysis is left until after the data
collection has been finished. Renato Rosaldo describes the method
of interpretive ethnographers as follows:
Data analysis during data collection
61
QualitativeAnalysis 9/11/01 1:37 PM Page 61
Ethnographers beginning research with a set of questions, revise them
throughout the course of inquiry, and in the end emerge with different
questions than they started with. One’s surprise at the answer to a
question, in other words, requires one to revise the question until lessening
surprises or diminishing returns indicate a stopping point
(Rosaldo 1989: 7).
Theoretical questions, and answers, are shaped and reshaped in
an ongoing dialogue with the experience or subjects being studied.
Rosaldo began his research of Ilongot subsistence farmers in the
Philippines, searching for an explanation for what motivated them to
headhunt. He did not accept their claim that it was an expression of
their rage associated with bereavement and looked for some other,
‘deeper’, reason. However, as his fieldwork progressed, and with his
own experience of bereavement following the death of his wife due
to an accident, Rosaldo came to understand both what the Ilongots
meant by rage in bereavement, and shifted the focus of his questioning
from headhunting to the experience of bereavement. Examining
the implications of his data for his research questions during his fieldwork
led Rosaldo to modify his research questions, which in turn
provided him with a much more sophisticated understanding of the
experience he had set out to study.
One of the central canons of grounded theory is that data
collection and data analysis are interrelated processes (Glaser &
Figure 3.1 Relationships between data analysis and
data collection
(Adapted from Lofland & Lofland 1971: 132.)
Qualitative Analysis
62
QualitativeAnalysis 9/11/01 1:37 PM Page 62
Strauss 1967; Becker 1971; Strauss & Corbin 1990): ‘In grounded
theory, the analysis begins as soon as the first bit of data is
collected’ (Corbin & Strauss 1990: 6). In grounded theory, data
collected early in the research process are used to guide the questions
that are asked as the research progresses. Data gathered early
in a research project guide both the formulation of concepts and
the sampling process (a technique referred to as theoretical
sampling, described below). Grounded theory is conducted this
way because it is assumed that researchers will not know all the
important research questions, sampling dimensions or theoretical
concepts before they begin collecting data (see the section on
grounded theory in Chapter 1). The research questions, the
sampling frame and the theoretical concepts are discovered only
while the data are being collected:
Each investigator enters the field with some questions or areas for
observation, or will soon generate them. Data will be collected on
these matters throughout the research endeavour, unless the questions
prove, during analysis to be irrelevant. In order not to miss anything
that may be salient, however, the investigator must analyse the first
bits of data for cues. All seemingly relevant issues must be incorporated
into the next set of interviews and observations (Strauss &
Corbin 1990: 6).
Examining data right from the beginning of data collection for
‘cues’ is what makes grounded theory ‘grounded’. It is also the foundation
of inductive theory building. Data collection is guided either
by preconceived theories and ideas about what is important, or data
collection is guided by the cues that present themselves during the
data collection process. As was argued in Chapter 1, a sophisticated
understanding of theory building recognises that interpretations are
a product of both previous understandings and the influence of
events in the world. Ethnography, hermeneutics and grounded
theory all emphasise this interweaving of theory and data.
Postmodernist and poststructuralist thought can be seen as at
least partially consistent with this understanding of the interpenetration
of data collection and analysis (Becker & McCall 1990; Denzin
1997). Postmodernists argue, for example, that research reports
should be seen more as contributions to ongoing conversations
about a research issue rather than as final analyses of ‘the truth’
(Lather 1993). It could be argued that the postmodernist point that
Data analysis during data collection
63
QualitativeAnalysis 9/11/01 1:37 PM Page 63
preexisting theory and interpretations influence the data collection
process is simply a more sophisticated way of describing what
grounded theorists have been doing for some time. Game (1991)
argues that mainstream sociology and poststructuralist practice
seem incompatible because sociology emphasises objectivity, and
the independence of the researcher, whereas poststructuralist semiotics
emphasises the interpenetration of meaning and experience.
While this criticism is true of much sociological research, particularly
the demographic and statistical research characteristic of
mainstream American sociology, it is not an accurate analysis
of qualitative sociological research, particularly research within
the interpretive or symbolic interactionist tradition. As Becker &
McCall (1990) observe, symbolic interactionists emphasise that,
in order to understand social life, researchers must examine the
meanings that shape the processes of interaction. The situated,
interpretive and processual emphasis of symbolic interactionists
has many similarities to the practice of poststructuralists. There
are, however, some significant differences between these two
approaches, for example in relation to their understanding of the
role of political and ethical questions in the research process (see the
discussion of rigour in Chapter 2). Nonetheless, they both argue that
research practice should explicitly combine the processes of data
collection and analysis.
The aim of qualitative research is to allow the voice of the
‘other’, of the people being researched, to inform the researcher.
The finite nature of human perception means that researchers
always choose to focus on one or another aspect of a phenomenon.
The voice of the participant, rather than the voice of the researcher,
will be heard best when participants not only provide the data to
be analysed, but when they also contribute to the questions that
frame the research and contribute to the way the data are analysed.
One way of achieving this is by ensuring the interpenetration of
data collection and data analysis.
Techniques for integrating analysis and collection
This section describes a number of practical techniques that can be
utilised alongside qualitative interviewing to begin the data analysis
process. Long interviews are one of the most common methods of
data collection utilised by qualitative researchers, both as a method
on their own and as part of other methods, such as ethnography or
Qualitative Analysis
64
QualitativeAnalysis 9/11/01 1:37 PM Page 64
participatory action research. Here I do not describe the techniques
of interviewing: there is a wide variety of excellent books that
describe in detail the method and process of long interviews
(Holstein & Gubrium 1995; Rubin & Rubin 1995). Rather I
describe a number of techniques that can be utilised during data
collection, using interviewing as an example, that facilitate the
concurrent analysis of data while they are being collected (see text
box below).
Techniques for data analysis concurrent with early data collection
• Team meetings and peer debriefing.
• Checking interpretations with participants.
• Transcribing, reading and coding early data.
• Writing journals and memos.
Team meetings and peer debriefing
In 1997 I led a team of three researchers interviewing people
living with HIV/AIDS about how they understood their future
(Ezzy 2000a). After each researcher had conducted one or two
interviews we had a team meeting to discuss our progress. One
member had interviewed a person with hepatitis C. While discussing
this interview it became apparent that co-infection with
hepatitis C significantly influenced the experience of living with
HIV/AIDS. A question about hepatitis C was therefore added to
our theme list after the first few interviews. In this way, the interview
theme list was updated continually during the data collection
for this project as new topics and new emphases were identified in
the interviews and then discussed in team meetings. Most of the
data analysis for this project was conducted after the interviews
were complete. However, discussing the research while data
collection was being conducted allowed a preliminary analysis of
the data. As a consequence the research was able to adapt and
include previously unanticipated dimensions of the experience of
living with HIV/AIDS.
Anselm Strauss is perhaps one of the most accomplished
collaborative qualitative researchers of the twentieth century
(Maines 1991). As such, his discussion of team meetings as part
of the research process is worth reading carefully if you are
involved in a collective qualitative project (Strauss 1987). Strauss
Data analysis during data collection
65
QualitativeAnalysis 9/11/01 1:37 PM Page 65
suggests that team meetings should be taped, transcribed, and
included in the memo files of the research project. Strauss identifies
four main benefits from team meetings. First, discussing the
data of the research stimulates ideas about its meaning and significance.
Second, some issues that arise during the discussion
are elaborated and developed that provide additional depth of
complexity and quality of analysis to the research. Third, the
issues raised may lead to team members choosing to follow up
issues through new data collection, the addition of questions to the
research schedule, or reviewing data collected earlier for an
analysis of the issues raised. Finally, team discussions may inform
the writing up of the project, particularly if they are transcribed.
Regular team discussions force researchers to confront common
research issues and encourage a focus on similar lines of inquiry.
The development of a shared analytic framework during data
collection makes writing up team research considerably easier.
Using team meetings to work the tensions between individual
interests and the team project can lead to a healthy development
of both: ‘In terms of the forward thrust of the entire project team
discussions not only ensure commonality of perspective, but also
the possibility of individual growth and a measure of autonomy in
the further pursuit of ideas: pursuit—it is important to emphasize—within
the common framework of analysis’ (Strauss 1987:
139). Backett-Milburn and associates’ (1999) reflection on a
collaborative feminist research project similarly points out that
negotiation and compromise is required if the differences between
team members’ interests and positions are to lead to a stimulating
synergy rather than dispiriting arguments.
For the solo researcher, peer debriefings can provide similar
benefits to team meetings. Peer debriefing is ‘the process of
exposing oneself to a disinterested peer in a manner paralleling an
analytic session and for the purpose of exploring aspects of the
inquiry that might otherwise remain only implicit within
the inquirer’s mind’ (Lincoln & Guba 1985: 308). Spall (1998)
suggests that peer debriefing should be conducted at crucial junctures
during the research and that it has three main benefits. First,
it makes the researchers more aware of the influence of their
personal values and theoretical orientations on the collection and
interpretation of the data. This point also includes discussions of
issues that may be ethically or legally problematic. Second, debriefing
sessions provide researchers with an opportunity to explore
Qualitative Analysis
66
QualitativeAnalysis 9/11/01 1:37 PM Page 66
and test their theories and interpretations of the data through
discussion with a colleague familiar with their discipline. Finally,
debriefing allows researchers to discuss problems with, and
planning of, the methodology.
Below is an example of how debriefing can inform the research
process. It is a personal communication in response to a paper I
have published on the experience of unemployment (Ezzy 2000b).
Although it is not the result of a debriefing session, it elegantly
illustrates the sort of information that a debriefing can provide to
the researcher. In her response to my paper, Henshaw discusses
the different ways in which respondents portrayed themselves in
interviews. Some told stories describing themselves as in control,
others recounted victim narratives. In the paper I point out that
some respondents reported telling different types of stories to
different audiences. However, I do not really explore the question
of what type of ‘audience’ I, the researcher, might have been for the
participant:
In short, you have not relayed to the reader the exact context of who
you are in relation to your interviewees, nor indicated how they might
have viewed you. For example, if you were perceived to be ‘an authority
from an important university’, your interviewees might have had
a vested interest in presenting themselves in the best possible light. On
the other hand, you might have been a complete stranger and
provided them with a great deal of space to represent themselves in
whichever way they wished . . . Therefore, because you have not
made the relationship (i.e., yourself as the context) between you and
your participants quite clear enough, the reader has some confusion
about where to situate the veracity of your observations (S. Henshaw,
personal communication).
Henshaw’s criticism of my paper is justified, and in response
I could provide more information about how I presented myself,
as an interviewer, to the interviewee. Unfortunately Henshaw’s
analysis came after the paper was published, but it nonetheless
illustrates the value of obtaining peer debriefing not only on draft
papers but also on all aspects of the research process. Research
reports, including journal articles and books, are a contribution to
an ongoing dialogue and debate within academic, political and
participant communities. Understanding does not come only from
individual researchers locking themselves away and reflecting on
their data. The responses of others to our interpretations are a
Data analysis during data collection
67
QualitativeAnalysis 9/11/01 1:37 PM Page 67
central part of the process of developing a trustworthy account.
Team meetings and peer debriefings provide a valuable opportunity
to begin this dialogue with other researchers early in the
research process.
Checking interpretations with participants
Later in this chapter I discuss the suggestion that research participants
should be included in all aspects of the research process.
Feminists and participatory action researchers have pointed out
that this has important emancipatory political implications for the
research process. However, it also provides a mechanism for
developing the dialogue with the research participant that is at the
heart of the qualitative research process. Whether or not participants
are involved from the beginning, in the design of the research
or as members of a steering committee, it is important to consider
how evolving interpretations of the data can be checked with
participants. Lather and Smithies (1997) asked participants to read
drafts of their research reports. In my own research I try to integrate
data collection by checking my evolving interpretations with
participants.
The aim of a good in-depth interview is to obtain the story or
interpretation of the person being interviewed. From this
perspective it is important not to try to suggest to the person how
you, as the interviewer, might expect them to respond.This is not
an argument for being neutral, it is an argument for ensuring that
the interviewer genuinely listens to the voice of the interviewee
(Rice & Ezzy 1999). However, during a long interview I typically
begin to develop my own summary of the interviewee’s experience.
I begin to place the person’s experience in the emerging
theory that I have about the issues being studied.Towards the end
of my interviews I often ask people about this interpretation that
I have developed of their experience. This serves as a check on
whether I have understood what they are saying. For example, in
my study of unemployment and mental health I identified a link
between a number of factors. People who felt confident about the
future, even though unemployed, also typically expressed dissatisfaction
with the job they had left or lost, and were financially
secure at least for the short term. This positive orientation to
unemployment was a product of both the person’s social location
and a product of the type of story he or she told about the
Qualitative Analysis
68
QualitativeAnalysis 9/11/01 1:37 PM Page 68
experience. The following extract is an example from one of my
interviews that illustrates how I examined this link by discussing
it with one of my participants at the end of the interview. Gail was
a single mother in her forties who had been working full-time as
a teacher, but who also had a contract to write a book. She saw
her time of unemployment as an opportunity to finish writing
her book.
Doug: Can I just check with you if my understanding is
correct?
Gail: Sure. [Laughs]
Doug: It seems to me that you were a little bit frustrated with
the work you were doing at the hospital.
Gail: Yeah, frustrated a lot, not a little bit.
Doug: When the position finished and they redeployed you,
you wanted to get out. Financially it is attractive to you
because you have got the redundancy package.
Gail: Reasonably attractive. Don’t let’s get carried away too
much with the value. It is not that much money.
Doug: And that you feel moderately financially secure for a
short while?
Gail: [Laughs]Yes, a very short while.
Doug: But more importantly, you feel confident that you can
get work in the future if you need to support yourself
financially.
Gail: Yeah, but that could be a false confidence, don’t forget.
Doug: I mean, do you feel confident or not?
Gail: Well, if I worry now about not getting a job, I am going
to start applying for jobs now and put all my energy into
looking for another job, which will deflect from my work
on my book. So whether I am using denial so that I don’t
worry about it or whether I am confident about it
probably doesn’t matter very much at all. Let’s just say
that I am confident!
Doug: I understand.
This extract elegantly illustrates the usefulness of this sort of
checking with the participant. It demonstrates that Gail is aware
of several different possible interpretations of her current experience.
She could begin to worry about the future.This would lead
her into searching for another job and, as a consequence, she
would not be able to finish her book. However, she has chosen
instead to interpret her situation in a way that provides her with
Data analysis during data collection
69
QualitativeAnalysis 9/11/01 1:37 PM Page 69
some self-confidence. Checking my interpretation of her experience
with her at the end of the interview brought out the nature
of her experience in a clearer light than would otherwise have
been the case.
Some researchers advocate returning transcripts of interviews
to interviewees for checking (Mason 1996). This may be a useful
strategy for checking details of the interview. A similar strategy is
to send participants summary vignettes that the researcher
has prepared from their interviews (Lather & Smithies 1997).
A summary vignette, through the process of selection, contains
preliminary data analysis. Checking a summary vignette with a
participant allows the participant to engage with the researcher as
they are doing their data analysis. The participant may point to
under- or overemphases and suggest complexities that were not
originally envisaged.
Transcribing, reading and coding early data
There is a temptation—that should be resisted—when conducting
long interviews to leave transcribing the interviews until after data
collection is complete.This is particularly the case if the researcher
has some funds allocated to having the tapes transcribed. It is
easier to leave transcription to be organised all at the one time.
However, there is considerable value to be obtained from
researchers themselves transcribing the first couple of interviews
they conduct before conducting the remaining interviews. First,
this allows interviewers to observe themselves in action, which can
be both painful and enlightening. As I transcribed my first interview,
I remember saying to myself: ‘Did I really say that?’. It was a
painful experience as I noted how I cut the participant off in the
middle of an account, and completely missed a cue on an important
issue. Second, transcribing the interview takes considerable
time, and encourages detailed reflection on the issues of the
research. I also began jotting down notes and ideas about theories
and concepts while transcribing the interview. Transcription
served as a preliminary form of data analysis. I began to make
links between the experience of the participants and concepts and
theory. Irvine (1999), in her exemplary study of Codependents
Anonymous, describes a similar process during her fieldwork utilising
participant observation. Note-taking fed directly into the
process of data analysis.
Qualitative Analysis
70
QualitativeAnalysis 9/11/01 1:37 PM Page 70
In the initial stages of my fieldwork I developed simple codes from my
notes from meetings. At first I coded what appeared to be parts of the
meeting, such as ‘Setting up’, ‘Leading’, ‘Speaking’ and ‘Sharing’.
Before long I began to develop more sophisticated codes within each
of these coded categories. For example, within ‘Speaking’ and
‘Sharing’, I developed codes for ‘Dysfunctional Childhood’, ‘Abuse’,
‘Hitting Bottom’ and ‘First Steps in Recovery’, among others (Irvine
1999: 13).
These codes were then compared against ongoing observation, so
that Irvine’s theoretical coding scheme developed alongside her
data collection. The two-way process of data collection and data
analysis allowed Irvine both to develop a more sophisticated theoretical
model and to collect data that were relevant to her research
questions and evolving theoretical scheme. Coding is discussed in
more detail in Chapter 4, but it is important to underline the value
of beginning the coding process during data collection.
Journals and memos
Write [your fieldnotes] as lushly as you can, as loosely as you can, as
long as you put yourself into it, where you say ‘I felt that’ (though not
to too great a degree). And as loose as that lush adverbalized prose is,
it’s still a richer matrix to start from than stuff that gets reduced into
a few words of ‘sensible’ sentences . . . you’ve go to start by trusting
yourself and writing as fully and lushly as you can (Goffman 1989:
131).
Many researchers advocate developing a sophisticated filing system
from the beginning of data collection as the foundation of the data
analysis process (Lofland & Lofland 1971; Strauss & Corbin
1990). This filing system can include a variety of memos or
journals on the practicalities of conducting fieldwork and emergent
interpretations of the significance of data collected for the project
as a whole. Journals and memos are a systematic attempt to facilitate
the interpretive process that is at the heart of qualitative
research. Understandings, interpretations and theories do not
emerge from data through some mechanical process. They are a
product of researchers thinking and talking about their research.
Keeping a journal and regularly writing memos encourages
Data analysis during data collection
71
QualitativeAnalysis 9/11/01 1:37 PM Page 71
researchers to reflect routinely on their emerging understanding of
the data.
Writing memos is particularly central to grounded theory, as it
forms the foundation of the emergent coding scheme. Strauss
defines a theoretical memo as ‘writing in which the researcher puts
down theoretical question, hypotheses, summary of codes, etc.—
a method of keeping track of coding results and stimulating
further coding, and also a major means for integrating the theory’
(1987: 22). Most grounded theorists write memos regularly, typically
from the beginning of the research project, as part of both
data collection and data analysis, which occur concurrently.
Looking back over my journals and memos from several
research projects, there is a pattern in how they develop that is
similar to the pattern described by Strauss (1987). My journals
begin with questions, suggestions about what I expect to find, and
ideas for reading. They move through notes that remind me of
people I should talk to about my findings, suggestions for sampling,
detailed discussions of particular interviews, and thoughts
about how particular books I was reading might relate to the
interview material. They also contain attempts to develop categories
and concepts, linking these to particular participants or
observations. Towards the end of the journal I begin to focus
more on the structure of the analysis as a whole and how
particular cases might fit into, or suggest modifications to, this
structure.
The memo on Michelle (see text box, p. 73) is an example of how
theory and categories for data analysis emerge during the writing of
memos during data collection. The memo was written quite soon
after the interview, probably after I had transcribed the tape. In the
memo I develop an emerging theory of what factors might influence
how people respond to losing their job.The theory is not fully developed;
it is after all only a memo early in the research. But it contains
an indication of the theory I later developed of the different ways that
people respond to job loss (Ezzy 2000b). In contrast to the interview
with Gail reported earlier, Michelle was distressed, and the memo
suggests that this might be linked to the importance of working
among her friends and the pleasure she found in her last job.
However, the memo also suggests that she was not as distressed as
some other interviewees (one of whom was suicidal) and links this to
her being active, not having friends at her last workplace, and being
financially supported by her parents.
Qualitative Analysis
72
QualitativeAnalysis 9/11/01 1:37 PM Page 72
Michelle: a memo
A very unpleasant job loss—no notice:
— implication that she was worthless
— enjoyed the work
— stopped her obtaining her traineeship certificate.
But:
— no friends at last workplace
— involved in dancing and public speaking and this helped
(cf. journal page 7).
Note that all her important friends are working.This
contributes to the maintenance of her own understanding of
work as central to her future.The social construction of
identity.
Working is also central to her understanding of a satisfying
future, even if she has children at some stage.
Unemployment leads to a loss of direction (interview transcript
page 3).
Depressed, angry and frustrated, but not suicidal.
Kept spirits up by active job search, regular activities such as
dancing and public speaking, which provide a sense of
achievement.
Financially dependent on parents.This means she survives
financially, but wants the financial independence
provided by work.
(From my journal for my study of unemployment Ezzy 2000b).
Qualitative data analysis is an interpretive task. Interpretations
are not found—rather they are made, actively constructed through
social processes. Data collection in qualitative research is not something
easily separated off from data analysis. Researchers make
many choices during data collection that are integral to how the
data are analysed and will be analysed—choices, for example, about
what or who to sample, what to ask, what to pursue, and what to
ignore. These choices are a product of the researcher’s developing
interpretation of the phenomena being studied. This interpretive
task is the beginning of, and integral to, qualitative data analysis.
The interpretive process of analysing qualitative data includes:
team meetings and peer debriefing; checking interpretations with
participants; transcribing, reading and coding early data; and
writing journals and memos. These, along with a variety of other
Data analysis during data collection
73
QualitativeAnalysis 9/11/01 1:37 PM Page 73
procedures, are ways of building an interpretation of the phenomena
through a dialogue with the phenomena and with other people’s
interpretations of the phenomena. Qualitative researchers should
aim to make the interpretive process explicit and integral to their
research, right from the beginning of the research.
Sampling and saturation
The most important point about sampling, as it relates to qualitative
data analysis, is that the sample is purposeful. The two most
common, and undesirable, sampling techniques employed in qualitative
research are convenience samples and snowball samples
(Patton 1990). The main disadvantage with these methods is that
the only rationale is ease or convenience. A purposeful sample is
one that provides a clear criterion or rationale for the selection of
participants, or places to observe, or events, that relates to the
research questions. A wide variety of sampling techniques have
been documented in a number of qualitative methodology texts
(Miles & Huberman 1994; Rice & Ezzy 1999). A sample that aims
for maximum variation, for example, would be most useful if the
aim of the research was to document the variations and patterns in
a particular phenomenon. Another sampling technique might focus
on extreme or deviant cases in order to illustrate processes that
would otherwise be difficult to observe. The important point is
that the reasons for the sample are clearly related to the research
questions. Theoretical sampling, used by grounded theorists, illustrates
this link between sampling choices and research questions.
In grounded theory, the units of analysis are sampled on theoretical
grounds (Glaser & Strauss 1967; Strauss 1987; Strauss &
Corbin 1990).This means that the sample is not defined prior to the
research but as the theoretical dimensions emerge during the
research. For example, during my research on unemployment I
developed a theory that the level of distress experienced by an unemployed
person was strongly influenced by the level of financial
distress.This theory was developed out of interviews with people on
relatively low incomes with and without significant debt. Among
these people the level of debt correlated with the level of distress, and
formed a central part of their talk about why they were or were not
distressed.That is to say, while being poor is not pleasant, it is much
more tolerable if you have no debt and have no-one else who is financially
dependent on you. I had also interviewed wealthy unemployed
Qualitative Analysis
74
QualitativeAnalysis 9/11/01 1:37 PM Page 74
people who were not distressed and were living quite comfortably on
their income. To examine this proposition more fully, I sought out
unemployed people who had a high income while unemployed and
also had a sizeable financial debt or considerable financial obligations,
such as a large dependent family.These people were distressed
both financially and more generally, despite their relatively high level
of income. Interviews with them underlined that it was not the
absolute level of income but the level of income relative to financial
obligations that caused distress.The earlier interviews had also indicated
that this was the case, but the theoretical sampling procedure
made the argument even stronger. Sampling conducted on theoretical
grounds resulted in a more sophisticated understanding of the
nature of financial distress experienced by unemployed people.
Theoretical sampling stops when the researcher decides the
study has reached saturation.The idea of theoretical saturation was
first formally described by Glaser and Strauss (1967). This idea
appears to have had its source in Everett Hughes, who advised his
students to keep interviewing until they did not hear anything new
(Hintz & Miller 1995).To be able to do this requires, of course, that
researchers are analysing their data as they are collecting it, otherwise
it would be very difficult to identify when saturation had been
achieved.This advice has implications both for the sample size and
for the types of units sampled. Strauss (1987) observes there will
always be new issues that can be pursued. However, data collection
has to end at some point, and in theoretical sampling this point is
decided on theoretical grounds, as a consequence of concurrent
data analysis and data collection.
Including participants in all aspects of the research
Feminist theory highlights the centrality of relationships to the
research process: ‘Research is an inherently relational process that
involves shared stories, actual bodies, and real voices’ (Way 1997:
704). Relationships necessarily involve power differentials: ‘The
prominence given within feminist methodological literature to the
importance of understanding what methods “do” both to research
participants and to research “findings” has been very important in
reconstituting knowledge-claims and in helping to develop a more
democratic social science’ (Oakley 1998: 725). A more democratic
research practice is typically achieved by formally involving the
researched as participants in the research process.
Data analysis during data collection
75
QualitativeAnalysis 9/11/01 1:37 PM Page 75
Drawing on participatory methodologies, some feminists have
included research subjects as co-researchers in their projects. The
aim of making subjects co-researchers is to avoid exploiting the
‘subjects’ and to empower women to research issues that concern
them. Lather and Smithies (1997), for example, refer to their
‘subjects’ as participants or contributors to their research. Rather
than taking ‘control’ of the data once the interviews were
completed, they included their participants in the writing process
as an ‘editorial board’ (Lather & Smithies 1997: 215).
Participatory research and feminist research share many objectives
and have been utilised together (Maguire 1987). Participatory
research developed in response to a similar desire by researchers in
developing countries to include participants in the research process
(Yeich 1996). Brydon-Miller reports that it was first used in the
early 1970s ‘by Maria Lissa Swantz to describe work then being
conducted inTanzania that drew on the knowledge and expertise of
community members in creating locally controlled development
projects’ (1997: 658).The research attempts to include participants
at every stage of the process: ‘They participate in a process of developing
research questions, designing research instruments, collecting
information, and reflecting on the data in order to transform their
understanding about the nature of the problem under investigation’
(Nelson et al. 1998: 884).
Morrow and Smith report that during their qualitative focus
group study of sexual abuse survivors they invited some participants
to become co-analysts during the data analysis phase of
the research: ‘The 4 coanalysts (termed participant-coresearchers)
continued to meet with Morrow for more than a year.They acted as
the primary source of participant verification, analysing video-tapes
of the group sessions in which they had participated, suggesting
categories, and revising the emerging theory and model’ (Morrow
& Smith 1995: 26).
Involving participants as co-researchers can be challenging, for
feminists and non-feminists alike. Not only does it question the
traditional presumed expertise of the academic researcher, it
requires innovations in data gathering, analysis and writing (Olesen
1994). This can be particularly problematic if, for example, the
researcher and the researched do not share similar political
objectives, such as a commitment to feminist emancipation.
Formally involving participants as co-researchers is not simply
a response to political concerns but also part of a practice that aims
Qualitative Analysis
76
QualitativeAnalysis 9/11/01 1:37 PM Page 76
to hear the voice of the other, to use the hermeneutic turn of
phrase. Feminist methods emphasise the need to ‘hear voices’, and
this draws attention to the human connection: ‘to the relationship
between speaker and listener, to the possibility of different
languages, and thus to the potential for misunderstanding or
mistranslation as well as to the ability of people to see and to speak
about themselves and the world in more than one way’ (Gilligan,
quoted in Way 1997: 705).
The extent to which participants are involved in research varies
considerably. Many qualitative researchers ask participants to read
transcripts of interviews, others include participants on steering
committees or as part of a consultative process in designing the
research. The degree of involvement reflects, in part, the political
orientations of the researcher, and the political objectives of the
research. I argue that there are no universally correct standards for
acceptable levels of participant inclusion in research projects. My
point is that the past practice of reducing the role of participants to
merely providing information in interviews will probably provide
neither the most useful data nor the optimum political outcomes.
The extent of involvement of participants has now become a
question that each researcher must address in developing
a research plan and practice.
Summary reflections
Most qualitative researchers do not presume to know all
their research questions before they start data collection. Additional
research questions can be discovered and researched only by
conducting data analysis, even if this is of a very preliminary kind,
during the process of data collection. Many qualitative researchers
do not know the dimensions along which they will sample for their
data. These dimensions can be discovered only by conducting
preliminary data analysis during data collection, and thus following
the practice of theoretical sampling. Many qualitative researchers
seek to include participants in all aspects of the research process.
Participants can only suggest additional, or different, research
questions; or suggest alternative sources of data, if they are
provided with, and/or included in, preliminary analyses conducted
during the process of data collection.
Qualitative data analysis is a process of interpretation. Data are
not interpreted after they are collected. Although interpretation
Data analysis during data collection
77
QualitativeAnalysis 9/11/01 1:37 PM Page 77
does happen after data collection, data collection itself is an interpretive
process. Choices about what to ask and who, or what to
sample, are products of interpretive understandings. If the
researcher conducts systematic data analysis during data collection,
then the process of data collection will be guided not only by the
researcher’s preexisting interpretations but also by the emerging
interpretations of participants.
Conducting data analysis during data collection results in a
more sophisticated and subtle analysis of the data.The interpretive
process begins when the researcher begins to reflect on his or her
research. From this moment data analysis begins, and should be
systematically conducted. Doing so will make for an easier, richer,
more subtle and more useful analysis.
Further reading
Interviewing
Holloway,W. and Jefferson,T. 1997 ‘Eliciting narrative through
the in-depth interview’ Qualitative Inquiry, vol. 3, no. 1,
pp. 53–71.
Holstein, J. and Gubrium, J. 1995 The Active Interview,Thousand
Oaks, Sage.
Rubin, H. and Rubin, I. 1995 Qualitative Interviewing:The Art
of Hearing Data,Thousand Oaks, Sage.
Journals and memos
Mason, J. 1996 Qualitative Researching, London, Sage.
Strauss, A. 1987 Qualitative Analysis for Social Scientists,
Cambridge, Cambridge University Press.
Strauss, A. and Corbin, J. 1990 Basics of Qualitative Research,
London, Sage.
Empirical studies with good descriptions of their data analysis
methodology
Irvine, L. 1999 Codependent Forevermore: The Invention of
Self in a Twelve Step Group, Chicago, University of
Chicago Press.
Qualitative Analysis
78
QualitativeAnalysis 9/11/01 1:37 PM Page 78
Data analysis during data collection
79
Lather, P. and Smithies, C. 1997 Troubling the Angels: Women
Living with HIV/AIDS, Boulder, HarperCollins.
Morrow, S. and Smith, M. 1995 ‘Constructions of survival and
coping by women who have survived childhood sexual
abuse’ Journal of Counseling Psychology, vol. 42, no. 1,
pp. 24–34.
Rosaldo, R. 1989 Culture and Truth, London, Routledge.
Sampling
Luborsky, M. and Rubinstein, R. 1995 ‘Sampling in qualitative
research’ Research on Aging, vol. 17, no. 1, pp. 89–113.
Patton, M. 1990 Qualitative Evaluation and Research Methods,
2nd edn, Newbury Park, Sage.
QualitativeAnalysis 9/11/01 1:37 PM Page 79
80
Traditional research designs have considerable limitations . . . A sort
of sympathetic magic seems to be involved, the assumption being that
if you go through the motions of science then science will result. But
it hasn’t . . . Understanding of ordinary behavior has not accumulated;
distance has (Erving Goffman, quoted in Manning, P. 1992: 141).
The quantitative, and functionalist, research designs that Goffman
refers to as ‘traditional’ are still present today. Influenced by positivism
and the attempt to be ‘objective’ and ‘scientific’, many
research designs systematically distance the researched from the
researcher. As a consequence the researcher is much less likely to
‘hear’ the ‘voice’ of the participants. It is relatively easy to conduct
a quantitative survey, or a group of one-off qualitative interviews,
with a group of participants. It is easier to publish the results of a
quantitative survey, takes less time to analyse, and will probably
further the career ambitions of the researcher more effectively,
depending on his or her disciplinary background. It is much harder
to spend time with people listening to their voices, understanding
their perspectives and sharing in their problems. Such research is
harder to publish, takes more time and energy to conduct, and is
often devalued when it comes to career progression. However, this
sort of research provides a much more sophisticated understanding
of the issues, will facilitate the formulation of more effective
policy, and is politically and ethically sensitive.
4
Coding data and interpreting
text: methods of analysis
QualitativeAnalysis 9/11/01 1:37 PM Page 80
The problem of overemphasising ‘scientific’ procedure, and
as a consequence using inappropriate methodology, is well illustrated
in Tanya Luhrmann’s (1989) anthropological ethnography
of magic and Witchcraft in contemporary England. The Witch
magicians studied by Luhrmann engaged in a variety of magical
practices in which they attemped to change physical realities
through magical techniques. Luhrmann says that ‘the point of the
study was to understand how [someone] . . . could come to treat
apparently outrageous claims as sensible topics for discussion’
(1989: 17). Prior to Luhrmann conducting her research, her
participants had been studied by a sociologist, who had handed
out questionnaires with categorical questions, and acted as if he
was an expert who knew more about the subject than his participants.
Questionnaires with fixed response questions are good
‘scientific’ technique. However, in his quest to follow correct
technique this sociologist had completely misunderstood the
nature of Witchcraft, and his research report was derided by the
Witches as equivalent to an observational study of trees that
concluded trees do not grow.
Qualitative researchers study meaning.The quality of research
into meanings and interpretive processes can not be assured simply
through following correct procedures. Interpretations and meanings
are situated. A method applicable to one research situation will
be inappropriate in another. Qualitative research is demonstrably
trustworthy and rigorous when the researcher demonstrates that he
or she has worked to understand the situated nature of participants’
interpretations and meanings. The quality of qualitative data
analysis depends on following well-thought-out procedures, and on
ensuring that these procedures reveal the structures of understanding
of participants.
Luhrmann reports that in order to understand the magical
practices of Witches she decided that she had to try to share
their subjective experience: ‘I decided that I would understand
magic best if I did what people did to become magicians’. She
read, studied and practised magic with a variety of Witches in
England. Luhrmann did not hide the fact that she was conducting
qualitative research, and this did not worry her participants
because she worked hard at fitting in with them. She did not
tape-record her conversations or pass out questionnaires, which
she thought would only damage her acceptance among the
groups in which she participated. She was more interested in
Methods of analysis
81
QualitativeAnalysis 9/11/01 1:37 PM Page 81
understanding the experience of these Witch magicians than in
ensuring that she systematically followed a procedure. Luhrmann
wanted to understand: ‘what it felt like to have a tarot reading,
how magicians argued for their practice, what they meant when
they said that they “saw” the Goddess. I did not see a complete
or representative set of magicians . . . but I did gain considerable
participatory insight into the way some people found magic
compelling’ (1989: 17).
While following correct procedures does not necessarily
produce trustworthy qualitative research, this does not mean that
qualitative research, and qualitative data analysis in particular,
should be completely unsystematic. Phillip Manning identifies
three distinct methodological practices in Erving Goffman’s work:
‘(1) metaphor; (2) unsystematic observation; and (3) systematic
observation’ (1992: 141). The strength of Goffman’s analyses lies
in his ability to move between practices that provide interpretive
understanding, such as metaphor, and practices that provide
convincing evidence, based on systematic observations. This
mixture of practices is an art that results in research that is both
evocative, in the sense that it produces new insight, and convincing,
because it rests on systematic research.
This chapter reviews four analytic strategies: content analysis,
thematic analysis and grounded theory, narrative analysis, and
cultural studies methodology. Clearly these do not exhaust the
possible analytic strategies available to qualitative researchers.
There are a number of other analytic strategies described in the
literature, including phenomenological methods (Moustakas 1994)
and conversation analysis (Potter 1996). I have focused on the
most commonly utilised methods, aiming to provide an overview of
the variety and extent of the available analytic strategies. Discussion
of each analytic strategy focuses on practical examples from
published research.
Content analysis
Content analysis is the most deductive of all forms of data analysis
discussed in this chapter. Deductively derived theory and deductively
driven data analysis work ‘down’ from preexisting theoretical
understandings (Glaser & Strauss 1967).The categories of analysis
are developed through logical deduction from the preexisting
theory. In this way preexisting theory is tested against empirical
Qualitative Analysis
82
QualitativeAnalysis 9/11/01 1:37 PM Page 82
data. Content analysis begins with predefined categories.Thematic
analysis, discussed below, allows categories to emerge from the data.
For example, feminist theory argues that the oppression of
women is partly the product of a culture that systematically
presents stereotyped images of men and women. From this
general theory, categories can be deduced of what constitutes
stereotypical images of men and women. A content analysis to test
this theory could, for example, examine images of men and
women in popular magazines, counting the number of times
images of men and women conform to stereotypical roles. A
content analysis of Australian popular magazines that employed
precisely this method found that popular magazines are ‘still
presenting stereotyped gender roles, lifestyles, and body management’
(Ring 1997: 3).
Content analysis, as with any other form of data analysis, begins
with the identification of the population from which units are
sampled.A sample is then drawn, typically using some form of stratified
sampling (Rice & Ezzy 1999). For example, an American study
of representations of masculinity in school textbooks identified
children’s reading books utilised in schools as their population
(Evans & Davies 2000). Evans and Davies then sampled, or selected,
two series of children’s textbooks published by two different publishers.They
then selected first-grade, third-grade and fifth-grade books
to provide a distribution of ages of intended readers.
Content analysis next defines the units of analysis and the
categories into which these will be placed. Evans and Davies used
characters in the stories as their units of analysis. Before their
analysis they developed an ‘instrument’ for categorising the characters
based on a inventory developed for classifying personality
traits in children’s stories. They identified eight stereotypical
masculine traits and eight stereotypical feminine traits and developed
definitions of these traits before beginning their analysis. For
example, Evans and Davies define the stereotypical masculine
trait of aggression as ‘actions and motives with intent to hurt or
frighten; imparts hostile feelings’ (2000: 261). Similarly, they
define the stereotypical feminine trait of affection as ‘openly
expressing warm feelings; hugging, touching, holding’.
Data analysis involves reviewing each unit of analysis and
categorising it according to the predefined categories. The occurrences
are then counted and comparisons made, often using
statistical or quantitative methods. Evans and Davies reviewed each
Methods of analysis
83
QualitativeAnalysis 9/11/01 1:37 PM Page 83
Qualitative Analysis
84
character portrayed in the stories and identified which of the
sixteen traits they portrayed. They then tabulated their results.
They found, not surprisingly, that males were portrayed with the
stereotypically male characteristics and females with the stereotypically
female characteristics. For example, 24 per cent of the male
characters demonstrated aggression, compared to only 4 per cent
of the female characters. Similarly, 33 per cent of the female characters
demonstrated affection, against only 18 per cent of the male
characters. Chi-square statistical tests were used to demonstrate the
statistical significance of these differences.
The final stage of content analysis is the interpretation of
results. Results are compared with the predictions of the preexisting
theory and conclusions for the theory are drawn. In
Evans and Davies’ case, while very few female characters
demonstrated aggression, a significant number of male characters
demonstrated affection. This might suggest that, while
children’s fiction does present stereotypically gendered characters,
there are examples that counter the stereotypes. However,
this is not the case. Here Evans and Davies resort to a form of
thematic analysis, because they identify themes in their data that
they had not specified prior to conducting their research. Specifically,
they noticed that whenever a male exhibited feminine
traits, he did so as part of some socially unacceptable behaviour.
Boys who exhibited feminine traits were ‘sissy’ and derided for
showing too much interest in domestic chores, or for wanting to
play with girls, or for expressing an interest in quiet play rather
than aggressive play. In short, Evans and Davies (2000) report
that masculine stereotypes in American schoolbooks are portrayed
in the same manner they were twenty years ago. This
analysis of gender stereotypes is consistent with Gergen’s (1992)
content analysis of best-selling American autobiographies, but
contrasts with Jagger’s (1998) content analysis of dating advertisements,
which found that gender stereotypes may be changing
with body image and lifestyle choice may now be more
important than previously central financial and occupational
attributes.
Content analysis is a useful way of confirming or testing a preexisting
theory. When the research question is clearly defined and
the categories of analysis have been well established by preexisting
research, content analysis may be an extremely useful method of data
analysis. It is not, however, a very useful way of building new theory.
QualitativeAnalysis 9/11/01 1:37 PM Page 84
Content analysis
• Identify categories prior to searching for them in the data.
• Select the sample to be categorised and identify units of analysis.
• Count, or systematically log, the number of times the categories
occur.
(Adapted from Kellehear 1993.)
As the example from Evans and Davies suggests, when new theories
or interpretations are required the researcher typically requires a
more inductive methodology such as thematic analysis. To explain
that masculine characters with apparently feminine traits were not
representatives of a new masculinity but derided male ‘sissy’ characters,
Evans and Davies utilised thematic analysis to identify the
category of sissy male inductively from the data.
Content analysis assumes that the researcher knows what the
important categories will be prior to the analysis. It restricts the
extent to which the data are allowed to ‘speak’ to the researcher. Put
another way, it severely limits the extent to which the ‘other’ can
have a voice as part of the research process. For this reason, in
qualitative research content analysis tends to be used in conjunction
with other forms of data analysis that are more inductive and
sensitive to emergent categories and interpretations.
Content analysis can be useful as a stage of data analysis as
it allows the relevance of preexisting theory to be tested, and it
can be used as a way of assessing the applicability of a theory
that emerges during thematic or content analysis. Grounded
theorists sometimes use content analysis in this way (Strauss
1987; Strauss & Corbin 1990). Strauss, for example, reports that
emergent hypotheses are ‘checked out’ or ‘verified’ (1987: 16)
during the analysis both through searching for new data and
through going back over old data and recoding them according
to the new categories. This is, of course, a variation on content
analysis.
The skills acquired through learning to conduct content
analysis form the basis of many of the skills of the more inductive
forms of research, such as thematic analysis and semiotics. The
novice qualitative researcher should first learn, and become thoroughly
familiar with, the principles of content analysis. This will
provide a solid foundation for successfully completing other forms
of qualitative analysis.
Methods of analysis
85
QualitativeAnalysis 9/11/01 1:37 PM Page 85
Coding in thematic analysis and grounded theory
‘In short, coding is the process of defining what the data are all
about’ (Charmaz 1995: 37). Coding in thematic analysis and
grounded theory is the process of identifying themes or concepts
that are in the data. The researcher attempts to build a systematic
account of what has been observed and recorded.Theory emerges
through this coding process. Coding links the data to an emergent
theory. In this section I use my own research with people with
HIV/AIDS to illustrate the various types of coding typically utilised
in thematic analysis and grounded theory.
Coding is an easy process that most people have already
performed. For example, as an undergraduate I used to use a
simple coding method when I was writing undergraduate essays.
I would start preparing to write the essay two or three weeks before
it was due, reading all the relevant chapters and articles. Then,
typically on the night before it was due, I would start writing the
essay by reading through all my notes. I would usually have two or
three pages of notes on each reference, and perhaps a dozen different
references. As I was reading through my notes I would notice
that there was a discussion of the same topic in two different
readings. So I would call this topic number 1. For example, in an
essay about the experience of unemployment, I found several
readings focusing on the different stages that unemployed people
go through after losing their job (Ezzy 1993). As I continued
reading through my notes I would notice other common themes
or topics. A number of studies of unemployment, for example,
focused on the secondary ‘functions’ of employment that are lost
when a person becomes unemployed. I would code this topic
number 2. Eventually I would assign most of my notes with codes
that linked them to one or another of my topics. After my first
reading through I would go back and find the uncoded bits of notes
and try to work out whether they fitted into my existing topics or
required a new topic, or could be left out of the essay. Next I would
write down all my topics on a piece of paper and rearrange them
until I thought I had an argument. I would then write my essay
based on the list of topics and my coded notes.
Perhaps you write essays differently? It doesn’t really matter.
What I want to demonstrate is that coding and categorising is
something most people have already performed in tasks as simple
as writing notes in the margins of books and articles. Coding in
Qualitative Analysis
86
QualitativeAnalysis 9/11/01 1:37 PM Page 86
qualitative data analysis is more complex than this, partly because
most qualitative researchers work with much larger sets of data.
However, the process is similar. The initial identification of topics,
often referred to as open coding, is exploratory, looking in the data
for codes. As the coding scheme becomes more developed new
forms of coding, referred to as axial and selective coding, are used
that enable the development of an argument, or central story,
around which the research report is organised.
Thematic analysis is part of the early procedures of data
analysis in grounded theory, but grounded theory goes beyond
thematic analysis.The term ‘grounded theory’ should be used only
to refer to studies in which data collection and data analysis are
conducted concurrently alongside theoretical sampling and other
techniques distinctive of grounded theory, such as the constant
comparative method (Strauss & Corbin 1990). Green observes
that many published qualitative research papers routinely parrot
the phrase that ‘the data were analysed during grounded theory’
(1998: 1064). This suggests that a sophisticated procedure has
been followed. However, Green points out that results presented
in the paper often suggest that the data analysis has utilised only
thematic analysis and not the sophisticated methodology of
grounded theory: ‘Unfortunately, what follows may be merely an
account of some key themes in the data, with brief textual quotes
in illustration, and sceptical readers remain unconvinced that qualitative
analysis is anything other than journalistic reportage’ (Green
1998: 1064). The more straightforward procedures of thematic
analysis may be appropriate for some studies, but it is important
that the researcher clearly identify which data analysis methods
have been utilised.
Both thematic analysis and grounded theory employ similar
techniques for analysing data. One difference between the two is
that grounded theory utilises theoretical sampling in which
emerging analysis guides the collection of further data (see the
discussion of sampling in Chapter 2), and this is not done in
thematic analysis. Grounded theorists have also developed a
sophisticated methodology for the development of codes, particularly
in relation to the development of core codes during selective
coding, and again this is not necessarily done in thematic analysis.
Thematic analysis can be employed either as part of a grounded
theory analysis or for the analysis of data that have already been
entirely collected. In this section I discuss the practice of coding
Methods of analysis
87
QualitativeAnalysis 9/11/01 1:37 PM Page 87
qualitative data, initially reviewing grounded theory and thematic
analysis together where they share similar methodologies, then
moving to a discussion that focuses on the more sophisticated techniques
unique to grounded theory.
Thematic analysis aims to identify themes within the data.
Thematic analysis is more inductive than content analysis because
the categories into which themes will be sorted are not decided
prior to coding the data. These categories are ‘induced’ from the
data. While the general issues that are of interest are determined
prior to the analysis, the specific nature of the categories and
themes to be explored are not predetermined.This means that this
form of research may take the researcher into issues and problems
he or she had not anticipated.
A clear example of thematic analysis is provided by Crisp
(2000), who designed a qualitative study of persons with disabilities
focusing on their interaction with health and rehabilitation
professionals. He was particularly interested in examining the
different ways people with disabilities respond to, and perceive,
health and rehabilitation professionals. Crisp recruited 35 disabled
people whom he interviewed using a semi-structured format; he
then transcribed these interviews. Within the parameters of his
general research question Crisp inspected the data, using thematic
analysis techniques, to develop a typology of responses to rehabilitation
professionals. The categories of analysis were not defined
prior to the analysis, but emerged during the analysis. Crisp
reports that at the beginning of his data analysis the ‘data was
inspected to elicit the conditions that underlie life events, interactions
with others, strategies and tactics that are adopted by
respondents, and consequences. It was initially coded openly
by scrutinising interview transcripts line by line or word by
word; by looking for in-vivo codes, terms used by respondents; and
by making comparisons for similarities and differences between
events and incidents’ (2000: 358).
The first stage of coding during thematic analysis and of
grounded theory is often described as open coding, as suggested in
the quote from Crisp in the previous paragraph. Glaser describes
open coding as a way to ‘generate an emergent set of categories and
their properties’ (1978: 56). More specifically, Strauss and Corbin
describe open coding as ‘the part of analysis that pertains specifically
to the naming and categorizing of phenomena through close
examination of data’ (1990: 62). Orona puts it more graphically
Qualitative Analysis
88
QualitativeAnalysis 9/11/01 1:37 PM Page 88
when she suggests that ‘coding each line is the guts’ of grounded
theory (1990: 1249).
For example, during my research with people living with
HIV/AIDS (Ezzy 2000a), as I read through my data the first time,
I made notes in the margins beside the following lines from my
interviews:
‘I have only got a couple of years to go . . .’
‘I was determined to live for ever . . .’
‘The future was I was dying.’
‘The future is still unknown to me . . .’
‘The life expectancy was anything from 5 to 10 years . . .’
‘That was living for the moment.’
‘I was just waiting around to die . . .’
The notes in the margins of the interview transcripts highlighted
the importance of the future and planning, and how a changed
understanding of the length of their life had affected participants.
As I read through these marginal notations I noticed a theme
emerging about how people thought about time. I then relabelled
all these lines with the code ‘temporality’ and wrote a memo to
myself noting that the code ‘temporality’ could be further broken
down depending on differences in the way people were oriented
towards the future. That is to say, I noted an emergent theme of
time. All the interviews contained a similar theme linked to their
concern about how people living with HIV/AIDS understood the
temporal nature of their lives. I named this theme with the code
‘temporality’. Some people were confident, expecting that they
would live out a normal lifetime. Others expected to die soon, and
were angry or depressed as a consequence. A third group of
participants expected their life to be shorter, but had accepted this
and celebrated the life that they had (Ezzy 2000a).
Open coding often involves considerable experimentation. At
first I experimented with the code ‘future orientation’. However, I
realised that the issue was not just how people thought about the
future but included how people felt about the present, given what
they expected to happen in the future. I experimented with a variety
of conceptual labels, or categories, or codes (all these words mean the
same sort of thing), until I found codes that seemed to fit the data.
This process sounds very straightforward. However, it is
anything but straightforward. It requires considerable effort and
Methods of analysis
89
QualitativeAnalysis 9/11/01 1:37 PM Page 89
reflection. Orona reports that ‘In the beginning, I literally sat for
days on end with the transcribed interviews spread out before me,
absorbing them into my consciousness and letting them “float”
about’ (1990: 1249). She wrote memos, talked about them with
friends, and explored any ideas that came to her.
One of the strengths of this form of grounded theory is that it
relies on hunches and intuition, or creativity, nuance and detail
(Orona 1990).The process is not linear or clear. Rather, it is often
confusing, frustrating and somewhat chaotic. This is both its
weakness and its strength. It leads to new ways of understanding as
new ideas are put together or participants’ interpretations are seen
in new light. However, it is also difficult, time-consuming and
demanding of energy.
The process of ‘constant comparison’ is one of the central
methods utilised by grounded theorists in developing and identifying
codes. Strauss and Corbin describe the process of constant
comparison as integral to the coding process: ‘As an incident is
noted, it should be compared against other incidents for similarities
and differences’ (1990: 9). Comparisons allow data to be grouped
and differentiated, as categories are identified and various pieces of
data are grouped together. Through the comparative process,
events that at first seemed entirely unrelated may be grouped
together as different types of the same category, or events that
seemed similar may be categorised differently. For example, one
participant’s discussion of ‘living for the moment’ at first seemed
quite unrelated to another participant’s discussion of his plans to
buy a house. However, as temporality was explored as a possible
category, these events were compared and identified as different
aspects of the same category.
Codes have properties, and these properties have dimensions
(Strauss & Corbin 1990). For example, the code ‘temporality’ has
the properties of how people feel about the future and how they
feel about the present. How people feel about the future varies
along a dimension. Some despaired because of the loss of their
future, others were more philosophical about the uncertain nature
of the future, and others were confident of a long future. Exploring
the properties and dimensions of a code can lead to the code being
broken into two separate codes, or it might lead to its being amalgamated
with a similar code.
Varying the units of analysis can be an important strategy during
open coding. Strauss and Corbin (1990) suggest experimenting
Qualitative Analysis
90
QualitativeAnalysis 9/11/01 1:37 PM Page 90
Methods of analysis
91
with coding lines, sentences, paragraphs and whole documents. In
my own research I moved between coding lines, paragraphs, interactional
events, narratives about episodes, and the structure of the
interview as a whole. Printing out interview transcripts with very
wide margins, preferably in landscape orientation, makes this sort of
coding much easier.
The next step in coding is described as axial coding by Strauss
and Corbin. Axial coding involves ‘specifying a category (phenomenon)
in terms of the conditions that give rise to it; the
context (its specific set of properties) in which it is embedded;
the action/interactional strategies by which it is handled,
managed, carried out; and the consequences of those strategies’
(1990: 97). The aim of axial coding is to integrate codes around
the axes of central categories. Orona describes the transition to
axial coding, saying that there came a point where she ‘felt’ that
she had to stop reading the transcripts of interviews with participants
and instead read all her own notes and memos carefully:
‘For several days, I sat wading through the notes and placing
them into what I felt were the major categories, which by then,
had been abstracted to a higher level. Thus “silent partner”,
“helper”, and “neighbours” had been abstracted to the level of
social relations’ (Orona 1990: 1249). She finally identified four
major themes: social relations, reciprocity, moral obligation and
temporality.
There is some debate among grounded theorists about the
nature and value of axial coding. Strauss and Corbin (1990)
argue for the value of axial coding, whereas Glaser (1978)
argues that it is a process that restricts the inductive, or
grounded, nature of theory building. According to the formalised
method of axial coding developed by Strauss and Corbin
(1990), among others, axial coding should focus on the four
dimensions of context, strategy, processes and consequences.
This focus on dimensions suggests a particular way of constructing
data analysis that focuses the coding process on the
relationship of codes to the analytic ‘whole’ (Schatzman 1991).
The danger, of course, is that in constructing ‘dimensions’
the analyst may decide to focus on issues related to his or her
interests rather than issues that concern the participants. In
her review of this debate, Kendall (1999) describes how her data
analysis became misdirected in precisely this way and developed
serious problems as a consequence. Kendall’s problems appear
QualitativeAnalysis 9/11/01 1:37 PM Page 91
to have arisen as a consequence of beginning axial coding too
early in her analysis rather than as a product of the specific
methods of axial coding itself. Whichever approach is taken,
Kendall’s review of the issues demonstrates that the most important
advice for the qualitative data analyst is to ‘not become
wedded too early to what looks obvious’ as central categories or
themes of the research (1999: 753).
In my study of people living with HIV/AIDS, the central theme
of temporality was coded axially as I identified other codes associated
with the various types of temporal orientations.That is to say,
people who were confident about the future tended to be healthy,
have good networks of friends, believed that medicine would solve
the problems of HIV, and were making decisions in their lives based
on this confidence in the future. On the other hand, people who
despaired about the future tended to have experienced illness, were
socially isolated, felt that medicine did not have the answers to
HIV/AIDS, and found it difficult to plan their lives very far into the
future.
Finally, selective coding (Strauss & Corbin 1990) or theoretical
coding (Glaser 1978) involves the identification of the core category
or story around which the analysis focuses. Crisp reports that in his
analysis ‘More selective coding occurred later when major themes
emerged . . . [and after they were identified] core categories were
repeatedly verified or revised after re-checking the transcribed
interview data, and after asking the respondents whether they
accepted (in everyday language) these accounts of themselves’
(2000: 358).
In my own analysis of stories of living with HIV/AIDS, the core
category was ‘temporality’ (Ezzy 2000a). I identified three different
temporal orientations: some people confidently expected a normal
lifetime; other people expected a short life, and were angry or
depressed about this; and a third group were uncertain how long
they would live, but decided to enjoy what time they had left.These
temporal orientations were linked to a variety of other codes, including
whether people felt in control of their life, whether they were
religious, and whether their values were self-centred or communally
oriented. In other words, the code ‘temporality’ provided the central
code around which all the other codes were fitted.‘Temporality’ also
provided the central ‘story’ of my research report and of the theory
that I developed to account for the different ways in which people
with HIV/AIDS respond to their diagnosis.
Qualitative Analysis
92
QualitativeAnalysis 9/11/01 1:37 PM Page 92
Coding finishes when the researcher is satisfied that the theory
is saturated. Morrow and Smith explain that in their analysis
‘Codes and categories were sorted, compared, and contrasted
until saturated—that is, until analysis produced no new codes or
categories and when all of the data were accounted for in the core
categories of the grounded theory paradigm model’ (1995: 26).
Note that saturation refers to the relationship between the codes
and the emerging theory. It will always be possible to discover new
information in the data, but saturation is achieved when the coding
that has already been completed adequately supports and fills out
the emerging theory.
Coding in grounded theory and thematic analysis
• Open coding:
— Explore the data.
— Identify the units of analysis.
— Code for meanings, feelings, actions.
— Make metaphors for data.
— Experiment with codes.
— Compare and contrast events, actions and feelings.
— Break codes into subcategories.
— Intergrate codes into more inclusive codes.
— Identify the properties of codes.
• Axial coding:
— Explore the codes.
— Examine the relationships between codes.
— Specify the conditions associated with a code.
— Review data to confirm associations and new codes.
— Compare codes with preexisting theory.
• Selective coding:
— Identify the core code or central story in the analysis.
— Examine the relationship between the core code and other
codes.
— Compare coding scheme with preexisting theory.
A sophisticated approach to coding during grounded theory
mixes both inductive and deductive methods. Codes do not
emerge from the data uninfluenced by preexisting theory. As
argued in Chapter 1, the process of theory building involves an
ongoing dialogue between data and theory. It is, however, very
difficult to balance the emergent nature of codes in grounded
theory against the influence of preexisting theory. Glaser
Methods of analysis
93
QualitativeAnalysis 9/11/01 1:37 PM Page 93
observes that in grounded theory ‘we do not have to discover all
new categories nor ignore all categories in the literature that
might apply in order to generate a grounded theory. The task is,
rather, to develop an emergent fit between the data and a preexistent
category that might work’ (1978: 148). The process of
developing an ‘emergent fit’ involves negotiating between categories
that emerge through data analysis and knowledge of
categorical schemes utilised in relevant literature and theory.The
aim is to avoid the knowledge of existing theory’s forcing the
analysis of the data into these preexisting categories. Through a
process of comparison of emergent categories with preexisting
categories, new and more sophisticated understanding of the
experience can be developed. Emergent codes may be named so
as to be consistent with preexisting theory. However, as existing
theory is integrated in the constant comparative process utilised
by grounded theory, as Wuest observes, this process ‘more often
results in modifying and building the emerging theory such that
it fits both the new data and the relevant concepts from the
existing theory’ (2000: 55).
In my study of people living with HIV/AIDS, the emergent
category of temporality resonated with a number of existing studies
that examined the temporal nature of chronic illness (Frank 1995;
Davies 1997). In my final report (Ezzy 2000a), these preexisting
conceptual frameworks were explicitly integrated with my own
emergent theory of the role of temporality in shaping people’s understanding
of living with HIV/AIDS. My data were coded inductively,
and as I coded I included preexisting theory as part of the constant
comparative process.The research report, therefore, becomes part of
an ongoing intentional dialogue about how to understand the different
ways that people experience and live with HIV/AIDS.
Coding is the process of disassembling and reassembling the
data. Data are disassembled when they are broken apart into lines,
paragraphs or sections. These fragments are then rearranged,
through coding, to produce a new understanding that explores
similarities, and differences, across a number of different cases.The
early part of the coding process should be confusing, with a mass
of apparently unrelated material. However, as coding progresses
and themes emerge, the analysis becomes more organised and
structured. Careful coding allows the researcher to move beyond
preexisting theory to ‘hear’ new interpretations and understandings
present in the data.
Qualitative Analysis
94
QualitativeAnalysis 9/11/01 1:37 PM Page 94
Narrative analysis
In contrast to the qualitative sociology, mainstream academic
psychology has rarely examined the person as a whole. Statistics
disaggregate the individual into measurable attributes. Similarly,
the traditions of grounded theory and thematic analysis, through
the use of cross-case comparisons, tend to disaggregate individuals,
focusing on codes and categories rather than people as the
units of analysis. In contrast, narrative analysis refers to the whole
of a person’s account. The parts of the story become significant
only as they are placed within the context of the whole narrative.
The emphasis on whole people and whole narratives represents
a radical change of focus. First, it emphasises that the nature of an
event or belief is not to be found in the event or belief itself, but
in the relationship of the event or belief to a broader interpretive
framework or narrative. This places ‘purpose’ at the forefront of
interpretation (Freeman 1984). If a researcher wants to understand
the meaning of something, he or she must locate the event or belief
in a broader narrative that defines its purpose, and therefore its
significance. Narrative analysis identifies the broader interpretive
framework that people utilise to turn meaningless events into
meaningful episodes that are part of a story leading out of the past
and into the future.
Plot is one of the central characteristics of a story. Plot is a
literary term for the structure of a narrative, derived from
Aristotle’s Poetics (Martin 1986). Narrative theory applies this
literary analysis of plots to the study of action, arguing that lives are
narrated in the same way as literary texts (Ricoeur 1984).
A succession of apparently unrelated events are configured into a
whole, a story with meaning, by the plot of a story. Plots explain
the point, or purpose, of the events discussed (Ricoeur 1985).
‘A fundamental way we create sense is by shaping the “one thing
after another” character of on-going action into a coherent narrative
structure with a beginning, middle and end’ (Mattingly 1994: 812).
Second, the emphasis on narrative embraces a situated relativity
and points to the ‘in-process’ nature of interpretations. Plots are not
fixed by the events they describe, but are situated constructions,
or acts of reading (Ricoeur 1984). Between the poles of objectivism
and relativism, narrative analysis embraces the situated and continually
transforming nature of interpretations and self-understanding:
‘Ricoeur . . . understands the construction of narrative identity as a
Methods of analysis
95
QualitativeAnalysis 9/11/01 1:37 PM Page 95
process that is constantly open to review. As such, it is the poetic
resolution to the hermeneutic circle’ (Joy 1993: 296). We all tell
stories about our lives, but these stories are always open-ended.
History and identity are not fixed constructs, but neither are they
completely flexible and malleable. Rather, they are somewhat stable,
but continually reinterpreted as we have new experiences and tell
new stories about ourselves, our past, and the world around us.
Stevens and Doerr (1997) provide an excellent example of the
use of narrative theory to examine the response of women to being
informed of a diagnosis with HIV. Drawing on long interviews with
38 HIV-positive women, they identified sections of the transcribed
interviews that reported the women’s responses to their diagnosis.
Their methodology is summarised in the text box below.
One narrative analysis method: Stevens and Doerr’s study of
HIV narratives
1. Identify the story to be examined—in this case, the participant’s
account of being told she had HIV.
2. Analyse the context and content of each story, particularly
focusing on understandings and feelings.
3. Examine how the women described the consequences and aftermath
of the diagnosis.
4. Compare and contrast the stories. Search for similarities and
differences in the structure of the story plots.
5. Examine the effects of background variables such as gender, age,
health status, and time since diagnosis, to see whether these are
related to earlier identified patterns.
6. Examine the transcripts for sections that illustrate the types of
stories identified.
(Summarised from Stevens & Doerr 1997.)
Stevens and Doerr (1997) identify three types of narratives.
Epiphany narratives described the HIV diagnosis as a revelatory
event through which the women found a new meaning to their life,
producing major changes to the way they lived. Confirmation narratives
described the HIV diagnosis as a discovery of something the
women already suspected, and as such it did not change their lives
significantly.These narratives tended to be emotionally muted, with
a tone of resignation. Finally, calamitous narratives described the
event of diagnosis as a shock in which the women felt they had
been given a death sentence, responding with fear, anguish and
Qualitative Analysis
96
QualitativeAnalysis 9/11/01 1:37 PM Page 96
intense emotions. The trauma of the diagnosis violently disrupted
their previous self-narratives and took these women by surprise.
These three narrative types are different ways of plotting an HIV
diagnosis. They link the event of an HIV diagnosis into a series of
episodes, past and future, that give it meaning and significance.
The responses are not, however, fixed. It is possible, for example,
for a calamitous narrative to be replaced with an epiphany narrative
due to some other experience.
The calamitous narrative type is well illustrated by an interview
conducted with a woman with HIV as part of a study of people
living with HIV/AIDS in Australia conducted at the National Centre
in HIV Social Research at La Trobe University (McDonald et al.
1998; Ezzy 2000a).This selection illustrates the nature of the narratives
Stevens and Doerr describe. ‘Sarah’ was 36 years old when she
was interviewed. She was married to ‘Matt’ and worked part-time.
She was infected with HIV about ten years ago, when she was 26
years old, prior to her relationship with Matt. However, she did not
find out she had HIV until about a year before the interview:
I was tired and run down and da da da da, and went off to the doctor
and said look I think I might be pregnant. He said the pregnancy test
is negative but we’ll do a blood test to make sure. I said fine. I said look
while you’re at it, a guy I was seeing died last year, could you run an
HIV test. Pregnancy test came back negative and the HIV test came
back positive. So, I had been telephoned at work, told by the doctor
over the phone telling me the result had come back positive but it could
be a mistake, you’d better ring your husband because he’ll probably
have it if you do have it, and you’d better get in here today. So, I had to
ring Matt. And he’s a plumber, he’s at work, and I’m saying you’ve got
to come with me to the doctor’s today.What is it—I can’t come today.
And I had to say, you know, the AIDS test has come back positive. It
couldn’t have. So, we both went down to the doctor’s. I’m crying,
Matt’s just in shock. We’re sitting at the doctor’s. The doctor said oh
well this is my first case, I don’t know what to say to you.There’s really
no treatment. It will be a very short life span. There’s no way known
you can have children.You’re looking pretty good at the moment but
you’ve had it off this guy for as long as you think you had, you won’t
be . . . I was the doctor’s first patient. He couldn’t understand that
people, in a wealthy middle-class suburb get HIV. People don’t get it
from this area. Particularly heterosexual women. So, he just couldn’t
cope.They rushed through Matt’s blood test. And his blood test came
back negative.That often it is the case that the husband will be negative
da, de, da. So that was sort of the initial sort of shock.
Methods of analysis
97
QualitativeAnalysis 9/11/01 1:37 PM Page 97
Qualitative Analysis
98
This narrative links a past (a previous relationship) with a social
location (white heterosexual woman) and the response of others
(the doctor’s terrifying prognosis) to explain the trauma and shock
of this woman’s diagnosis with HIV. There are a variety of treatments
for HIV/AIDS currently available, although they do not
guarantee a healthy life. The doctor’s incorrect information fed
into a plot of the events that described the movement from
a good life, through the trauma of diagnosis, to an anticipated
difficult life. The event of diagnosis takes on meaning as a
consequence of its placement in a story about this woman’s life.
The plot of her story is of HIV as a calamity.The events of her life
are made sense of—are interpreted—within this broader narrative
frame.
Narrative analysis contains a very broad range of methodologies
for identifying narrative structure, and Stevens and Doerr’s
approach is only one among many. For example, Bamberg’s (1997)
edited collection contains a wide selection of narrative analysis
methodologies, including a statistical cognitivist approach to narrative,
an ethnomethodological conversation analysis, a functionalist
linguistic study, a cross-cultural interpretive approach and a life span
study. Each of these studies utilises different methodologies: some
statistical, some metaphoric, some interpretive, they also draw on
different theoretical paradigms, from psychological cognitivism,
through linguistic theory, to hermeneutics. Riessman (1993) also
provides a useful overview of several narrative analysis methods. In
this short introduction I have attempted only to describe two
approaches that are widely utilised in qualitative social research.
For qualitative social researchers, one of the strengths of narrative
analysis is that it provides a constructive way of doing social
research that engages with, rather than denies, the epistemological,
ontological and methodological issues raised by contemporary
social theory. Narrative analysis, as it is typically utilised in social
research, draws on the hermeneutic theory of philosophers and
moral theorists such as Ricoeur (1984, 1985, 1988),Taylor (1989),
MacIntyre (1995) and others (see the discussion of hermeneutics
in Chapter 1). The implications for qualitative methods have been
explored by a number of authors, such as Polkinghorne (1988),
Bruner (1990) and Riessman (1993). Some feminist researchers
have similarly drawn on narrative analysis as part of a feminist
response to the crisis of legitimation and representation (Personal
Narratives Group 1989; Chase 1996; Richardson 1997). Bell’s
QualitativeAnalysis 9/11/01 1:37 PM Page 98
study of women with cancer provides an excellent illustration of
this link between a sophisticated theory and the practice of narrative
analysis:
Sociological theorists are questioning the possibility of producing
accurate knowledge about social life as well as how to account for the
ways knowers produce knowledge:What constitutes knowledge?What
constitutes a subject? What constitutes action, agency, power, or
resistance? Narrative analysis is a particularly strong way of addressing
these questions (Bell, S. 1999: 347, emphasis added).
Susan Bell (1999) utilises narrative analysis in her study of
women with cancer as a consequence of their mothers’ exposure to
DES, a drug prescribed to prevent miscarriage. She reports that at
first she began her study by trying to distance herself from the
women she was interviewing. She strove for ‘objectivity’. Bell did
not tell her participants about her own life or commitments, and
strove to select women who would not know her. However, as she
become more involved in her study she become uncomfortable
about the attempt to distance herself from her participants, objectifying
them as objects of scientific inquiry. Influenced by feminist
theory and narrative analysis, she began to be more open about her
own commitments and experiences with her subjects. Her interviews
became more conversational and reciprocal.
Bell’s methodology differs from Stevens and Doerr’s in that
Bell is less concerned with identifying the overall plot of the narrative
through an examination of what is said. Rather, Bell focuses on
how the story is told. She examines narrative techniques such as the
use of repetition, metaphors, phrasing, and the imagery of the
story. Rather than providing cross-case comparisons, Bell examines
two specific narratives in detail. Bell uses her analysis to show how
the narratives of the two women she describes are integrated with
broader cultural narratives (see text box, p. 100).
Narrative analysis is attractive to Susan Bell for two reasons.
First, narrative analysis explicitly addresses the role of the interviewer
in the construction of interview responses: ‘In my
interpretation of Molly’s and Deborah’s experiences I explore how
my social position helped to construct the interview context, the
production of narratives about their experiences, and my interpretation
of the contexts and narratives’ (Bell, S. 1999: 354).
Contemporary research on long interviews has highlighted that
interviews are not places where an interviewer goes and collects
Methods of analysis
99
QualitativeAnalysis 9/11/01 1:37 PM Page 99
accounts that were preexisting in the participant’s head. Rather,
interviews are places where meanings, interpretations and narratives
are co-constructed (Holstein & Gubrium 1995). To try to be
objective, to try to avoid influencing the interview, is fruitless.
Rather, the researcher should be explicit about his or her role in the
interview process. Narrative analysis facilitates precisely this
explicit analysis of the role of the interviewer in the construction of
the interview narrative.
One narrative analysis method: Bell’s study of cancer narratives
1. Identify narrative segments in the interview transcripts.
2. Examine word choice, phrasing, imagery and structure of
clauses.
3. Focus on the telling of the story: how do people explain what
they did, or what happened?
4. Examine how the stories relate to each other.
5. Look for connections between the personal accounts and
broader cultural and political processes.
6. Locate yourself, as the researcher in the analysis and the
construction of the stories.
(Summarised from Bell, S. 1999.)
Narrative analysis does not attempt to identify the one true
interpretation of participants’ stories. Rather, the goal is to identify
the cultural and social context that facilitates the everyday practice
of telling stories about oneself and one’s world: ‘Understanding the
meaning and significance of a story requires understanding how it
is communicated within or against specific cultural discourses and
through specific narrative strategies and linguistic practices’ (Chase
1996: 55). Narrative theory explicitly engages with the complexity
of the world and the finite nature of human understanding. Human
action is too complex to ever discover a final set of laws to describe
it. Humans are situated, and can never know everything. As
Josselson puts it: ‘narratives are not records of facts, of how things
actually were, but of a meaning-making system that makes sense
out of the chaotic mass of perceptions and experiences of a life’
(1995: 33).
Second, narrative analysis allows Bell to connect the narratives
of individual women to the more general political context: ‘Their
narratives display the ways these women connect their individual
life experiences to changing social and structural conditions in the
Qualitative Analysis
100
QualitativeAnalysis 9/11/01 1:37 PM Page 100
context of the women’s health movement’ (Bell, S. 1999: 353).
Narrative analysis allows the researcher to be explicit about the
political and cultural location of both the narratives of participants
and the researcher. This is precisely the point of Mills’ (1959)
argument, that social researchers need to grasp the link between
sociopolitical processes and individual biography. Narratives exist
at a variety of ‘levels’. Personal narratives told in fleeting encounters
such as interviews contain, represent and misrepresent
narratives that an individual may hold for a considerable time,
and narratives that broader sections of a community may share and
may be embedded in more general cultural processes. Bell
summarises her research, pointing to how her analysis links
precisely these dimensions:
My interpretations of the interview narratives show how two DES
cancer daughters’ perceptions and interpretations are mediated
through the cultures surrounding them, how they live within and in
tension with systems of domination, how their individual biographies
are connected to the structural conditions in which they originate, and
how their narratives are jointly produced by researcher and subject.
These interpretations show how narrative analysis can demonstrate
and explain the production of knowledge (Bell, S. 1999: 385).
Narrative analysis refers to a wide range of analytic methodologies.
I have illustrated the practice of two of these in this section.
On the one hand, Stevens and Doerr’s methodology focuses more
on the structure of the story, on what is said.The methodology they
utilise is similar to, and expands on, the analytic strategies of
grounded theory and thematic analysis. A side range of narrative
studies have utilised similar methodologies (Gergen 1988; Bruner
1990; Mattingly 1994; Frank 1995). On the other hand, narrative
analysis, as conducted by Bell, shifts the focus of the research from
what participants say to how they tell their stories. This involves
‘attending to the cultural, linguistic, and interactional contexts and
processes of storytelling’ (Chase 1995: x). This concern with
broader cultural and political context is shared by cultural studies
and is discussed in the next section.
Cultural studies and semiotics
Postmoderns subvert the authority of modernist metatheory with a
rhetorical conception of science . . . They do so by focusing on
Methods of analysis
101
QualitativeAnalysis 9/11/01 1:37 PM Page 101
the how rather than the what of knowledge, its poetic and political
enablements rather than its logical and empirical entailments
(Brown 1991: 190).
In 1997, McGuigan reported that it ‘remains difficult to see quite
what cultural studies amounts to methodologically’ (1997: 1).
This is because cultural studies is both eclectic, drawing on a
wide range of method from other disciplines, and diverse, with a
wide variation in methodologies that makes it difficult to identify
that which is common to cultural studies. In this section I focus
on cultural studies research as it is relevant to the practice of
qualitative data analysis. McGuigan’s (1997) book provides an
excellent overview of the methodologies of cultural studies more
generally.
Cultural studies has recently taken a ‘turn’ towards qualitative
methodology. Early cultural studies focused almost exclusively
on the ‘text’ of a television program, film or writing.This textually
determinist model provides no room for an examination of how
audiences dialogue with the ‘text’, nor does it conceive of interpretations
as constructed intersubjectively. However, more recent
cultural studies practitioners have recognised that ‘textual
meanings do not reside in the texts themselves; a certain text can
come to mean different things depending on the interdiscursive
context in which viewers interpret it’ (Ang 1996: 38).
Interpretation in cultural studies
• Data, or the text, are interpreted in the light of broader cultural
and social systems.
• What is missing from the text is as important as its manifest
content.
• Preexisting theory is used to interrogate and interpret data.
• Analysis is conducted to reveal the operation of power in
cultural life.
• Some researchers rely on rhetoric and aesthetics to persuade
readers of the authenticity of their work, largely ignoring issues
of systematic observation and analysis.
• The results of data analysis are not framed as scientifically
validated truth, but as historically located, subjective and
relative.
Qualitative Analysis
102
QualitativeAnalysis 9/11/01 1:37 PM Page 102
Cultural studies locates the interpretation of data within an
analysis of broader social and cultural processes. Influenced by
structuralism, it examines not only the manifest content but also
the ‘deep structure’ of a text, or data: ‘One must pay attention to
both, looking not only at classification systems, but also at what is
implied, what is not spoken, what is “really meant”’ (Lamont &
Wuthnow 1990: 290). Another way of saying this is that a semiotic
or cultural analysis of magazines, television programs and
transcribed discussions about these cultural artifacts by audiences
takes the researcher outside the data. It examines the relationship
between the ‘data’ and broader social and cultural frameworks.
Cultural studies tends not to call this process ‘data analysis’,
or ‘coding’, but ‘interpretation’, or ‘reading’. This interpretive
emphasis contrasts with the focus on detail and nuance characteristic
of thematic analysis and grounded theory. Data are not
dissected so much as contextualised. This contextualisation is
not found in the data, but is drawn from a more general analysis of
social and cultural processes, often informed by critical theory
or feminism:
The aim of cultural studies is not a matter of dissecting ‘audience
activity’ in ever more refined variables and categories so that we can
ultimately have a complete and generalizable formal ‘map’ of all
dimensions of ‘audience activity’ . . . Rather, the aim, as I see it, is to
arrive at a more historicized and contextualized insight into the ways
in which ‘audience activity’ is articulated within and by a complex set
of social, political, economic and cultural forces (Ang 1996: 42).
This analytic emphasis is often linked to a strategy of data analysis,
or reading, that examines both what is not present in the data as
well as the manifest content. To put this another way: qualitative
data analysis facilitates the identification of differences in interpretations
and experience among people, events and interactions.
Thematic analysis and grounded theory use codes to attune the
researcher to the structure of these differences in the lives of
participants. Cultural studies is also interested in these differences,
but from the perspective of locating their meanings interpretively
within broader social, cultural and political contexts.
For example, McKinley (1997) utilises cultural studies
methodology to examine how people talk about the prime-time
soap Beverly Hills 90210. Reflecting the influence of cultural
Methods of analysis
103
QualitativeAnalysis 9/11/01 1:37 PM Page 103
studies’ move towards ethnography, McKinley’s data are not the
‘text’ of the program itself but the ‘text’ of talk about the program
among 36 young women, obtained during interviews and transcriptions
of talk while they were watching the program. McKinley
describes how young women watch and talk about 90210 as part of
a process of gender enculturation. She argues that the women
actively work at constructing an understanding of gendered
identity as part of their talk about this television soap. Her method
of data analysis, or interpretation, explicitly ‘works’ the relationship
between the empirical data, or text, and more general theories
of cultural practice and social structure. While she describes
her research as an ‘empirical testing of poststructuralist theory’
(McKinley 1997: 5), her analytic strategy is more complex than
simply analysing data using thematic or content analysis and
comparing the results with preexisting theory. Rather, the preexisting
theory enters into the very way that McKinley conducts
her interpretation of her data:
The [interpretive] questions became, what female identities were seen
as appropriate and/or encouraged as viewers talked about 90210?
What options were hidden? And what identities were made so natural
that they were accepted as real and immutable? . . . poststructuralist
feminist [theory] led me to analyze the talk about females with
an eye towards ways it did—or did not—perpetuate the values of
patriarch and capitalism, and to ask what the role of the television
text was in generating and guiding—or not guiding—this talk
(McKinley 1997: 8–9).
Unlike the methods of grounded theory, cultural studies explicitly
integrates theoretical questions as part of the data analysis
process.This is for two reasons. First, cultural theory locates interpretations
within more general cultural and social processes. Talk
about 90210 among McKinley’s small group of women is informed
by, and representative of, processes of ‘hegemonic patriarchy’:
‘Time and time again, I heard viewer talk working to explore
identities that challenge the patriarchal definition of womanhood,
then retreating to close down these alternate possibilities and reestablish
a conservative status quo’ (McKinley 1997: 9).
Another way of describing this analytic strategy is suggested by
Barthes’ (1967, 1972) distinction between three orders of signification.
Drawing on Saussure, the process of signification draws an
Qualitative Analysis
104
QualitativeAnalysis 9/11/01 1:37 PM Page 104
analytic distinction that differentiates signs into two component
parts of signifiers: the physical object such as a printed word, image
or person and the signified, which is the mental concept or word
that refers to the signifier (Fiske & Hartley 1978). The words
‘female identity’ are the signified that refers to the signifier of actual
women. The first order of signification includes self-contained
references, or the manifest content. Talk about female identity in
90210 would be analysed as simply indicative of how the women
understood the relevance of the program for their self-identity.Talk
about female identity is taken to indicate the actual lives of women.
This is the level at which much qualitative interpretation tends to
operate, or at least begin. The second order of signification refers
to meanings that derive from the way society values and incorporates
signs into a range of cultural practices and meanings. Female
identity, both as signifier and signified, in contemporary society is
typically associated with passivity, heterosexuality and subservience
to men. The third order of signification links these
general cultural references into a ‘comprehensive, cultural picture
of the world, a coherent and organized view of the reality
with which we are faced’ (Fiske & Hartley 1978: 41). The talk
about female identity is now understood as part of a more general
patriarchal and capitalist society. Cultural studies is most interested
in this third order of signification.
Second, cultural studies is integrally political, asking questions
about power and power relations that require a more general theoretical
frame. Grossberg puts this forcefully, arguing that cultural
studies is ‘not about interpreting or judging texts or people, but
about describing how people’s everyday lives are articulated by
and with culture, how they are empowered and disempowered
by the particular structures and forces that organize their lives’
(Grossberg 1998: 67). For example, British cultural studies has
been deeply influenced by Stuart Hall’s (1980) argument that
cultural studies is worthwhile only for its contribution to a political
radicalism. From this perspective, culture is analysed in terms of its
relationship to existing social inequalities linked to class, race and
gender. In British cultural studies analysis focused on how the
messages of culture, such as television programs or popular novels,
either supported or confronted these forms of inequality. The
analytic concern was not with the ‘message’ of the text but with
the significance of these messages for broader political issues. A
television program that did not contain any mention of gender
Methods of analysis
105
QualitativeAnalysis 9/11/01 1:37 PM Page 105
inequalities could therefore be analysed as contributing to the ongoing
nature of the oppression of women. There is some debate
among cultural theorists about the value of Hall’s position, although
even the more pragmatic approaches retain a commitment to articulating
the political dimensions of their research: ‘The ambition of
cultural studies is to develop ways of theorizing relations of culture
and power that will prove capable of being utilized by relevant social
agents to bring about changes within the operation of those relations
of culture and power’ (Bennet 1997: 52).
In America, cultural studies has mingled critical theory with
pragmatism and symbolic interactionism, in a tradition of which
C. Wright Mills is emblematic (Denzin 1992). Mills’ (1959)
argument that sociology has a responsibility to conceptualise other
possible ways of organising society, while often ignored within
sociology, has been taken up within cultural studies. As Calhoun
puts it: ‘Cultural studies seeks . . . to explore the ways in which our
categories of thought reduce our freedom by occluding recognition
of what could be’ (1995: xiv).
While cultural theorists recognise participants as active constructors
of their lives, in a significant way participants are determined, or
overdetermined, by broader social processes. Denzin considers that
‘[the participant] does not understand the historical forces that shape
everyday biographical life. Only the analyst understands these forces’
(1997: 236). In this sense, it is impossible to analyse people’s selfunderstandings
on their own terms. In order to make sense of a text
or action, these must be translated into theoretical terms. This is
precisely what McKinley does in her analysis of women’s talk about
a television soap: ‘Gathering and analyzing empirical evidence of
the ways the microprocesses of hegemony play out in talk about the
show have led me to conclude that such talk is implicated—for better
or worse—in the reproduction of dominant notions of female
identity’ (McKinley 1997: 235).
Interpretation in cultural studies is often more reflective of
Goffman’s first two analytic strategies of metaphor and unsystematic
observation than of his third method of systematic observation.
Paul Smith observes that British cultural studies, while committed
to empirical research, has tended to eschew an emphasis on
systematic methods, criticising these as overly ‘positivist’, and
instead using methods that have more in common with the aesthetics
of literary and art criticism: ‘Traditional issues of measurement
and methodology have become displaced in favor of innovative,
Qualitative Analysis
106
QualitativeAnalysis 9/11/01 1:37 PM Page 106
virtuoso interpretations of media texts, youth sub-cultures, popular
music lyrics, etc.’ (Smith, P. 1998: 10). In contrast to the more
systematic ethnography of Willis (1977), cultural theorists such
as Hebdige (1979) drew on a more creative and semiological
approach to cultural analysis, as exemplified in his book Subculture.
Hebdige’s work draws on a variety of eclectic sources for data and
aims to produce a sensitivity to the complexity of youth subculture
through the use of metaphor and models. As McGuigan puts it, his
work reflects a ‘poststructuralist fascination with the play of signifiers
and [moves] away from “experience”’ (1992: 101).
The analytic method of cultural studies contains an unresolved
tension between a theoretical emphasis on analytical deduction
and an empiricist inductivism. Lamont andWuthnow (1990) point
out that American cultural studies has emphasised the role of
observation and empirical induction, and this is linked to a rejection
by some to the radical politics characteristic of British cultural
studies. European cultural studies has emphasised theoretical
deduction as primary in generating new theory and interpretation,
and this is often linked to a more explicit commitment to engaging
with the political implications of their analysis:
Cultural studies has in general been more willing than sociology,
with its strong universalising bent, to grant that knowledge may be
inherently perspectival—or to put it differently, may be both limited
and enabled by the knower’s historical, cultural, and social access to
the world, including the world of intellectual traditions—and more
eager to explore the links between knowledge and social domination
(Long, 1997: 15).
Cultural studies challenges the naive inductivist empiricism of
some grounded theorists. If all interpretation is from a perspective,
then it is pointless to pretend that preexisting theory, or the value
commitments of the researcher, have not shaped the research
process.This does not make empirical research irrelevant, as some
critics argue. Drawing on the theory developed in Chapter 1, I
argue that it is still important to engage in systematic empirical
research, and that the methods of grounded theory have considerable
value in assisting this process. However, cultural theorists
problematise the politics of the interpretive process, asking from
whose perspective, and for whose benefit, the interpretation has
been conducted. This is a question that cannot be avoided if it is
Methods of analysis
107
QualitativeAnalysis 9/11/01 1:37 PM Page 107
accepted that all interpretation is unavoidably political. These
points were taken up in detail in Chapter 2. Ang summarises the
implications of taking this interpretive approach in her description
of the aim of research: ‘It is not the search for (objective, scientific)
Truth in which the research is engaged, but the construction of
interpretations, of certain ways of understanding the world, always
historically located, subjective and relative’ (Ang 1996: 46). The
emphasis on the historical and subjective nature of the products of
research in cultural studies also leads to an emphasis on the role of
the author, and researcher, in producing the research. This is
discussed more fully in Chapter 6.
Compared to the sophistication of the semiotic techniques
utilised to study cultural texts, cultural studies has applied surprisingly
unsophisticated methods to analyse letters, transcripts of
interviews and ethnographic data (Nightingale 1993). This does
not just refer to the sometimes ‘thin’ nature of interpretations
based on barely described, and often significantly attenuated,
evidence and analytic methods (McEachern 1998). Early attempts
at ethnography in cultural studies often also ignored the role of the
researcher in the research process.
‘What occurs, then, in the absence of rigorous ethnographic
observation and description, when the techniques of ethnography
are divorced from ethnographic process, is a co-opting of the interviewee’s
experience of the text by the researcher, and its use as
authority for the researcher’s point of view’ (Nightingale 1993: 153).
This point is almost exactly the same as Glaser’s (1978) point about
the danger of qualitative researchers forcing their data into the categories
of preexisting theories. Radaway admits to the probability that
her own research was shaped in this way when she suggests that ‘my
initial preoccupation with the empiricist claims of social science
prevented me from recognizing fully that even what I took to be
simple descriptions of my interviewees’ self-understandings were
mediated if not produced by my own conceptual constructs and
ways of seeing the world’ (1991: 5). As Radaway’s quote suggests, to
make this point does not require a return to a naive empiricism. To
argue for the importance of rigorous method does not require the
abandonment of the more general political and theoretical orientations
of cultural studies. It does, however, require that the researcher
take equally seriously the voices of participants and the researcher’s
own voice.This can be achieved through a combination of rigorous
method and explicit engagement with more general political and
Qualitative Analysis
108
QualitativeAnalysis 9/11/01 1:37 PM Page 108
theoretical issues. Radaway suggests that if she were to revisit her
research, ‘I would attend more closely to the nature of the relationship
that evolved between the Smithton women and me by
describing the interviews themselves in greater detail and including
representative transcripts from them’ (1991: 5).
Cultural studies reminds us that qualitative research is an interpretive
process. Interpretations are always situated, historical,
subjective and political. Researchers that ignore these aspects of the
interpretive process will produce an analysis that falls into
predictable traps. Ignoring the situated nature of research leads to
claims about the generalisability of results that are less than
convincing. Ignoring political dimensions of the research leads, by
default, to a politics of conservatism. Cultural studies also reminds
us that interpretation is as much an art as a systematic process.
It is arguable that some cultural studies researchers have ignored
systematic analytic procedures to their loss. However, it is equally
arguable that adherence to systematic method may give some
qualitative researchers a confidence that is unfounded. Qualitative
researchers should aim for a balance between systematic observation,
unsystematic observation, and metaphor. As with Goffman’s
research, the correct mixture of these methods leads to research
that is both evocative, in the sense that it produces new insights,
and convincing, because it rests on systematic research.
Summary reflections
Description demands model-building and models always distort; there
is no clear window on a different culture. However, the attempt to build
an account sensitive to interpretive limitations may provide a powerful
understanding of the phenomenon (Luhrmann 1989: 14–15).
Each of the analytic strategies described in this chapter is a way of
summarising and interpreting ‘data’. The aim is not to discover,
finally and objectively, what is ‘out there’. Rather the aim is to
engage with the data as ‘other’, as a participant in a conversation in
which the researcher also participates.This does not mean, however,
that anything goes methodologically. Systematic and rigorous data
analysis strategies are both better at hearing the voice of the ‘other’,
and provide a stronger position from which to contribute to the
ongoing politically imbued conversation in which we live.
Methods of analysis
109
QualitativeAnalysis 9/11/01 1:37 PM Page 109
Qualitative Analysis
110
Further reading
Content analysis
Kellehear, A. 1993 The Unobtrusive Researcher, Sydney, Allen &
Unwin.
Grounded theory and thematic analysis
Charmaz, K. 1995 ‘Grounded theory’ in J. Smith, R. Harre and
L.Van Langenhove (eds) Rethinking Methods in
Psychology, London, Sage, pp. 27–49.
Orona, C. 1990 ‘Temporality and identity loss due to Alzheimer’s
disease’ Social Science and Medicine, vol. 30, no. 11,
pp. 1247–56.
Strauss, A. and Corbin, J. 1990 Basics of Qualitative Research,
London, Sage.
Narrative analysis
Franzosi, R. 1998 ‘Narrative analysis’ Annual Review of Sociology,
vol. 24, no. 1, pp. 517–55.
Josselson, R. and Leiblich, E. (eds) 1993 The Narrative Study of
Lives, vol. 1, London, Sage.
Personal Narratives Group 1989 InterpretingWomen’s Lives,
Bloomington, Indiana University Press.
Riessman, C. 1993 Narrative Analysis, Newbury Park, Sage.
Cultural studies
Alasuutari, P. 1995 Researching Culture: Qualitative Method and
Cultural Studies, London, Sage.
Fiske, J. and Hartley, J. 1978 Reading Television, London,
Routledge.
McGuigan, J. (ed.) 1997 Cultural Methodologies, London, Sage.
QualitativeAnalysis 9/11/01 1:37 PM Page 110