Amanda Coffey
PaulAtkinson
MAKING
SENSE OF
QUALlTATlVE
DATA
Complementary
Research
Strategies
SAGE Publications
InternationdlEducationaland ProfessionalPublisher
Thousand Oaks London New Delhi

Concepts and Coding
Linking Concepts and Data
Many analyses of qualitative data begin with the identification of key
themes and patterns. This, in turn, often depends on processes ofcoding
data. The segmenting and coding of data are often taken-for-granted
parts of the qualitative research process. All researchers need to be able
to organize, manage, and retrieve the most meaningful bits of our data.
The usual way of going about this is by assigning tags or labels to the
data, based on our concepts. Essentially, what we are doing in these
instances is condensing the bulk of our data sets into analyzable units by
creating categories with and from our data. This process is usually
referred to as coding, although that can imply a rather mechanistic
process. We prefer to think in terms of generating concepts from and
with our data, using coding as a means of achieving this.
We stress here that although coding may be part of the process of
analysis, it should not be thought of as the analysis in itself. In other
words, coding should not be seen as a substitute for analysis. It would be
as much a mistake to think that coding is an activity that is universally
Concepts and Coding 27
understood across the qualitative (orindeed quantitative)research spec-
trum. Rather, the term coding encompasses a variety of approaches to
and ways of organizing qualitative data. As parts of an analytical process,
however, attaching codes to data and generating concepts have impor-
tant functions in enabling us rigorously to review what our data are
saying.
The analytic procedures that underpin coding procedures establish
links of various sorts. First, codings link different segments or instances
in the data. We bring those fragments of data together to create catego-
ries ofdata that we define as having some common property or element.
We define them as being about or relating to some particular topic or
theme. The coding thus links all those data fragmentsto aparticular idea
or concept. As we will see, such concepts are in turn related to one
another. Codes, data categories, and concepts are thus related closely to
one another.The important analytic work lies in establishing and think-
ing about such linkages, not in the mundane processes of coding. The
importance ofthe work lies in how we use the codings and concepts,not
in whether we use computer software to record them or rely on manual
ways of marking and manipulating the data.
Important analytic work also lies in the identification of relevant con-
cepts. We use the data to think with, in order to generate ideas that are
thoroughly and precisely related to our data. Coding can be thought about
as a way of relating our data to our ideas about those data. Because codes
are thus links between locations in the data and sets of concepts or ideas,
they are in that sense heuristic devices. Coding reflects our analytic ideas,
but one should not confuse coding itselfwith the analyticwork of develop-
ing conceptual schemes. As Seidel and'Kelle (1995, p. 52) note, "codes
represent the decisivelinkbetween the original'raw data,' that is,the textual
material such as interviewtranscripts or fieldnotes,onthe onehand and the
researcher's theoretical concepts on the other."
In practice, coding can be thought of as a range of approaches that
aid the organization, retrieval, and interpretation of data. Miles and
I-Iuberman (1994)suggest that coding constitutes the "stuff of analysis"
(p. 56), allowing one to "differentiate and combine the data you have
retrieved and the refl3ttions you make about this information" (p. 56).
'They argue that coding is'aprocess that enables the researcher to identify
~neaningfilldata and set the stage for interpreting and drawing conclu-
sions. They describe codes as
' r..I.

28 M A K I N G S E N S E O F Q U A L I T A T I V E D A T A
tags or labels for assigning units of meaning to the descriptive or inferential
information compiled during a study. Codes usually are attached to "chunks" of
varying size-words, phrases, sentences or whole paragraphs, connected or
unconnected to a specific setting. They can take the form of a straightforward
category label or a more complex one (e.g. metaphor). (Miles & Huberman,
1994, p. 56)
They go on to say how they see codes being used to retrieve and
organize data:
The organizing part will entail some system for categoriziilg the various
chunks, so the researcher can quickly find, pull out and cluster the segments
relating to a particular research question, hypothesis, construct or theme.
(Miles & Huberman, 1994, p. 57)
Later in this chapter, we provide some examples, drawn from the data
on anthropology students and their mentors, illustrating how codes can
be assigned to chunks of data and how we can then use these codes to
generate concepts and themes. Before we do so, however, it might be
useful to review some of the different ways in which coding can be
approached.
On the one hand, coding can be thought about in terms of data
simplification or reduction. If the codes are kept to a general level and
their number relatively small, then the data are reduced to their bare
bones, stripped down to a simple general form. This coding approach
can be compared directly to simple forms of content analysis (Krippen-
dorf, 1980). The addition of simple, broad analytic categories or codes
can thus be used to reduce the data to manageable proportions. Here the
analyst is concerned primarily with the identification of a simple con-
ceptual schema. The main goal of such coding is to facilitate the retrieval
of data segments categorized under the same codes. Coding in this
context is essentially a process of indexing the data texts, whether they
be fieldnotes, interview transcripts, or other documents. Data are re-
duced to equivalence classes and categories. The qualitative analyst will
thus be able to retrieve chunks or segments of textual data that share 3
common code. Such code-and-retrieve procedures can be used to treat
the data in quasi-quantitative ways by, for example, aggregating in-
stances, mapping their incidence, and measuring the relative incidence
of different codes. Such coding and retrieving can be implemented in a
Concepts and Coding 29
variety of manual styles. Texts can be marked up physically with mar-
ginal keywords or code words, different colors can be used to mark or
highlight the texts, and index cards can be used to cross-reference
instances to numbered pages or paragraphs in the data.
Such a code-and-retrieve procedure alsohas been implemented using
a number of computer software packages. There are now a number of
applications designed specifically for the analysis of qualitative data,
some of which we introduce in more detail in Chapter 7. Many of the
contemporary programs incorporate code-and-retrieve functions. Such
data-handling procedures also can be accomplished to varying degrees
by the general cut-and-paste functions ofword-processing software (see
Stanley 81 Temple, in press).
Coding and retrieving is the procedure most often associated with
coding as an analytic strategy. The role of coding in such a conceptuali-
zation is to undertake three kinds of operations, according to Seidel and
Kellc ( 1995,pp. 55-56):(a) noticing relevant phenomena, (b) collecting
examples of those phenomena, and (c) analyzing those phenomena in
order to find commonalities, differences, patterns, and structures. Seidel
'~ndKelle are clear that even when coding is used to reduce data, codes
are heuristic dcvices. In this sense, coding qualitative data differs from
cluantitative analysis, for we are not merely counting. Rather, we are
attaching codes as a way of identifying and reordering data, allowing the
data to be thought about in new and different ways. Coding is the
mechanics of a more subtle process of having ideas and using concepts
about tlic data. It can be viewed as
nothing more than a preparation for this process which is based on a careful
inspection and analysis of raw data (that is segments of text) and on their
con~parisonfor the sake ofidentifyingpatterns and structure. (Seidel &Kelle,
1995, p. 58)
As well 3s data simplification and reduction, coding can be concep-
tualized as data complication. Coding need not be viewed simply as
rcclucing data to some general, common denominators. Rather, it can
hc used to expand, hansform, and reconceptualize data, opening up
rnore diverse analytical possibilities. We go on to say more about this
ldtcr in the chapter, but it is important to recognize at the outset that in
ihc hands of commentators such as Anselm Strauss, coding can refer to

30 M A K I N G S E N S E OF QUALITATIVE DATA
a different kind of orientation toward one's data from that implied by
data reduction. The general analytic approach here is not to simplify the
data but to open them up in order to interrogate them further, to try to
identify and speculate about further features. Such data complication is
not used to retrieve and to aggregate instances to a restricted number of
categories; rather, it is intended to expand the conceptual frameworks
and dimensions for analysis. Coding here is actually about goingbeyond
the data, thinking creatively with the data, asking the data questions, and
generating theories and frameworks.
In practice, coding usually is a mixture of data reduction and data
complication. Codinggenerally is used to break up and segmentthe data
into simpler, general categories and is used to expand and tease out the
data, in order to formulate new questions and levels of interpretation.
One should try to ensure that coding does not lose more than is gained.
It is especially important to avoid the use of coding merely to apply
simple and deterministic labels to the data. Data reduction or simplifi-
cation of that sort is not the main analytic purpose ofqualitative coding.
Coding should be thought of as essentially heuristic, providing ways of
interacting with and thinking about the data. Those processes of reflec-
tion are more important ultimately than the precise procedures and
representations that are employed.
Seidel and Kelle (1995, p. 58) capture this by saying that "codes do
not serve primarily as denominators of certain phenomena but as Izetl-
ristic devices for discovery." This is apparent whichever model or ap-
proach to coding is adopted. Take, for example, the approach of Tesch
a Iza-(1990).Tesch describes qualitative analysis in terms of decontextu 1'
tion and recontextualization. Decontextualizing data involves segment-
ing portions ofdata and slicing up the data set.Tesch defines segmenting
as dividing data into portions that are comprehensible by themselves and
large enough to be meaningful. Decontextualization means separating
data extracts from their original context while retaining meaning. Seg-
mented data are then organized and sorted as part of a process of
recontextualization. Tesch (1990) suggests that the first step of sorting
consists oftagging text segments with information about the category ofthe
organizing system into which it belongs (orseveral categories i f the scginent
is relevant to more than one).Many researchers call this process "coding."
(P17.1)
Concepts and Coding 3 1
Tesch (1990) suggests that once data segments have been coded, they
are still not ready for interpretation. Drawing on the work of Marton
(1986),who argues that each quotation has two contexts-the one from
which itwas taken and the "pool ofmeaning" towhich itbelongs-Tesch
suggests that an organizing system for data is based on developing pools
of meaning. Concepts are identified or constructed from prior material,
theoretical frameworks, research questions, or the data themselves. The
segmented data are coded according to those organizing categories and
then re-sorted, again according to those categories. Data segments are
reassembled or recontextualized. Coding as part of this process aids the
recontextualization of data, giving a new context for data segments. In
this way, Tesch regards coding as a means ofproviding new contexts for
viewing and analyzing data. Decontextualizing and recontextualizing
help to reduce and then expand the data in new forms and with new
organizing principles. To put it another way, segmenting and coding
data enable the researcher to think about and with the data.
Strauss (1987) provides perhaps the best example of using coding to
complicateand expand qualitativedata.We discussthe work of Strauss and
his collaborators in more detaillaterin this chapter. For now, it isimportant
to note that Strauss advocates coding as an essential analytical procedure.
He argues that qualitative researchers must learn to code well and easily.
Strauss also is keen to stress that coding is often misunderstood to be a
simple and unproblematic procedure. The argument here is that coding is
much more than simplygivingcategoriesto data; it is also about conceptu-
alizing the data, raising questions, providing provisional answersabout the
relationships among and within the data, and discoveringthe data. Strauss
argues that codingshould be used to open up the inquiryand move toward
interpretation. Coding is thus about brealung the data apart in analytically
relevant ways in order to lead toward further questions about the data. To
paraphrase Strauss (1987), coding can be viewed as a way toward the
excitement and inevitablepayoff of grounded conceptualization.
We can get ideas for coding from a variety of sources that are not
mutually exclusive. We can start with a simple framework for coding
based on what we as researchers are interested in. Reading through data
extracts, one might discover particular events, key words, processes, or
characters that capture the essenceof the piece. Alternatively, one might
code the data extracts using a code list created prior to reading the data.
Miles and Huberman (1994)suggestthat this method, of csating a "start

32 M A K I N G S E N S E O F Q U A L I T A T I V E DATA
list" of codes prior to reading the data or even prior to the fieldwork, is a
useful way of beginning to code. These codes or categoriescan come from
a variety of sources. For example, we can start from our theoretical or
conceptual frameworks-coding data accordingto key concepts and theo-
retical ideas. We might have hypotheses that could be used to select code
words to identifysegmentsofthedata,in ordertotest or modifythose ideas.
Equally,we could startwith preselectedcodes taken from ourreadingin the
general area, or a comparative area, or previous studies. Key variables and
concepts can be derived from the research literature.
Another way of beginning to code is to start from the foreshadowed
researchquestion thatinspiredtheresearch project.Onemightbeginwith the
data and categorizethem in a more inductive fashion,startingwith the local
categoriesof the actors or informants themselves.We can thus categorizethe
data more in accordance with the indigenous terms and categories of the
culture or the individual informants. We will try to illustrate such thought
processeswhen we explore some of our own data later in this chapter.
It isworth stressinghere that codesare organizing principlesthat arenot
setin stone. They are our own creations, in that we identifyand selectthem
ourselves.They are tools to think with. They can be expanded, changed, or
scrapped altogether asour ideasdevelopthrough repeated interactions with
the data.Startingto createcategoriesis away ofbeginning to read and think
about the data in a systematic and organized way.
What to code, or what categoriesto create, will alwayspartly depend on
the intent of your data analysis. Strauss (1987) makes the distinction
between sociologicallyconstructed codes and in vivo codes.The latter refer
to codes that derive from the terms and the language used by social actors
in thefield,or in the course ofinterviews.The systematicuse ofin vivocodes
can be used to develop a "bottom up" approach to the derivation of
categoriesfrom the content of the data. Initial coding, then, should help us
to identify themes, patterns, events, and actions that are of interest to the
researcher and that provide a means of organizing data sets.Coding can be
more or less complex, depending on the level of analysis.
The Coding Process
The segmenting of data using codes or categories, as we have indi-
cated, can be achieved in a variety of ways, through the application of a
Concepts and Coding 33
variety of analytical strategies. For the purposes of illustration, we have
taken an extract from an interview with an established academic in
anthropology. The interview extract focuses on the question of what
makes a good doctorate. Set out below is the interview extract, to which
we have attached a number of coding categories.
Extract From Interview WithDr. Fitton (Kingford University)
Odette Parry: What do you think -
makes a good PhD?
Dr. Fitton: I think PhDs should show a sub-
stantial contribution to research, but I
don't think that necessarily means inno-
vation for innovation's sake.
I personally would want to favour a PhD
which showed a very sound knowledge
of theoretical positions, an ability to sort
outthose positions and put forwardsome-
thing in a logical, coherent, structured
fashion. I'd favour someone who was
able to do that over someone who has
studiedsomethingthat no onehad thought
ofstudyingbefore, andyou're encourag-
ing something that is peripheral, mar-
ginal, not necessarily of significance. So
I think that what I would look for is a
very sound acquaintance with theoreti-
calwork, an ability to sortit out, and take
it further-have a sufficient substantial
commentary on that work, combined in
the case of anthropology with fieldwork,
and showing that the fieldwork had been
done in a way which shows empathy
with the people you'd studied, and that
the fieldwork and the theoretical part
h ~ dbeen merged together.
Quite a la11order. 1'h look for a "feel" about
the work, I wouldn't have a list of guid-
ing points, because I don't think you can
do that-they are too different. It has
Good
PhD
conti-ihtion
originality
- Ability
theory
logic
I
I
originality
theory
I
fieldwork
I
I
theory
1
indeterminate

34 M A K I N G S E N S E O F Q U A L I T A T I V E D A T A
been said that the strength of anthropol- eclecticism
ogy is its eclecticism, it relies on qualita- qualitative
tive analysis rather than quantitative. -I+ neth hods
Odette Parry: This is really a general -Why
question. Why do you think people PhDs
do anthropology PhDs? I
Dr. Fitton: In some cases it's the obvious rea-
son that doing a PhD will hopefully lead
to the first rung of the academic track.
My own motivation was not that clear. I
was surprised when 1did get a job at the
end of it, but to further an interest I
wanted to take as far as I could. I expect
most people doing a PhD are doing it to
further an interest they have. There seems
to be a trend towards PhDs written to do
with development, so you could say that
a concern for other societies is another
factor. So it's not just a selfish endeav-
our. I can think of one student I've had,
the interest in doing a PhD wasn't there,
there was an external push, she was ex-
pected to. get high qualifications. And
because her heart wasn't in it, she didn't
have the necessary enthusiasm and drive
for it.
motive
motivatio~l
academic
career
interest
H-
intrillsic
interest
qualificat'lons
enthusiasm
As shown here, with a relatively simple approach to the process of
coding, different levels of complexity can be explored. What we have
done here, in fact, is summarize a series of successivedecisions about the
data and its categorization.When they are superimposed in one display,
it easilycan look as if theywere all derived simultaneously,from the same
set of interests and concerns. That is not the case, and here we try to
indicate some of the decisions that have gone into such a coding out-
come.
At the simplest level, the data can be reduced to two possible generic
categories: "a good PhD" and "why people do a PhD." These reflect
directly the questions that Odette Parry asked and reflect two of the
substantive problems that the research team brought to the data-collection
Collceytsand Coding 35
exercise. The research was intended to examine someofthe personal and
institutional factors that influence the PhD process, including views on
the PhD itself as a means of academic socialization. Clearly, coding the
data according to those themes adds nothing initially to our under-
standing of the data. It is essentially a data-reduction task. Segmenting
and coding the data in that particular way would at least allow us to
characterize what each stretch of the interview was about in terms of
general thematic content, in this instance relating directly to the topics
of the interview elicitations and responses. Such wide, generic categories
would facilitate the retrieval of different segments of data that deal with
descriptions of good PhDs and academics' speculations as to doctoral
candidates' motivations.
The application of these and equivalent codes, reflecting substantive
research questions, would be one basic way of organizing the data. Such
procedures have considerable practical value. The nature of qualitative
data means that data relating to one particular topic are not found neatly
bundled together at exactly the same spot in each interview (and field-
notes usually have even less predictable organization). The ability to
locate stretches of data that, at least ostensibly, are "about" the same
thing is a valuable aspect of data management. Such coding, therefore,
can be a useful preliminary to more detailed analysis.We could proceed
a little further in the same vein. Another way of thinlung about this
particular data extract would be to "code" in terms of specific abilities
or competencies identified by the informant. That is, as we have indi-
cated, the first half of the interview extract definitelydescribes a number
of abilities or attributes linked to the production of a good PhD. Coded
as "ability," this data segment could be used to search for and compare
other aspects of the data set where graduate students' abilities are re-
ferred to. For example, PhD supervisors might in other contexts refer to
the abilities required for supervising doctoral work or for being a suc-
cessful academic more generally. Likewise, doctoral students in inter-
views might also describe their experience in terms of the abilities they
bl.ought to their studies, developed over the course of their research,
failed to develop, and so forth. Each category of actor might refer to their
own competence aqd that of other actors. As is clear, these very general
categories promot~thereordering of the data in accordance with pre-
liminary ideas or concepts. They are not necessarily the final ways in
which the data can be examined and explored. This first approach to the

36 M A K I N G S E N S E O F Q U A L I T A T I V E D A T A
data, however, does little or nothing toward complicating the questions
we can ask about them. If we are going to use coding to generate more
interesting and complex ideas about our materials, then we need to do
somethingmore. Most fundamentally,we need to think more about how
we interact with the data.
Coding at such a very general level is a first step toward organizing
the data into meaningful categories. It can be seen as the first level of
coding. As we demonstrate in the data extract above, coding also can be
thought about in a more complex way. Using a good PhD, or ability, or
why a PhD as initial codes or categories, a number of subcategories can
be generated and used to segment the data. In this extract, we have
identified a number of such categories and attached codes accordingly.
Some of those more detailed codes come more or less directly from the
informant's words, such as enthusiasm. Others are our summary glosses
ofwhat the informant seems to be referring to or describing at a particu-
lar point in the text. For example, we have glossed one of the motivating
factors as altruism, not the actual word used by the informant but used
by us to capture descriptions of motivation based on or ascribed to the
desire to help others or a commitment to another culture. Other codes
reflect more directly our conceptual interests. For example, we have
categorized one segment as referring to indeterminate kizowledye. Dr.
Fitton is talking about evaluating a PhD with reference to its "feel" and
talks about a lack of specificguidelines. Our identification ofthe segment
in this way, and our decision to code as we have, reflects our own interest
in pursuing how anthropologists talk about their own knowledge and
the knowledge of other anthropologists, past and present. Here, there-
fore, we note an appeal to an apparently indeterminate criterion in
evaluating doctoral students' work. In coding it aswe have,we can collate
all the instances where similar appeals to indeterminate knowledge have
been made, talked about, or denied. We can inspect those instances
further to examine the varieties of indeterminate knowledge and their
sources, how indeterminate knowledge is evaluated, how it is learned,
and so on.
These more detailed subcategories can be represented by a single code
attached to a discrete segment ofthe data. Subcategories also can overlap
one another. Codes and their segments can be nested or embedded
within one another, can overlap, and can intersect. The same subcate-
gory can be applied several times in a single unit of data (such as an
Concepts and Coding 37
interview),and the same segment can have more than one code attached
to it. In the data extract above, the dense nesting and overlapping of the
more detailed codings can be seen. These dense patternings are quite
characteristic of code maps of qualitative data. After all, ordinary social
action, including conversational talk, does not present itself to us in
neatly bounded packages. When we segment the data by attaching codes,
we often reflect how topics run into one another and how there may be
multiple issues to concern ourselves with simuItaneously.
Coding is never a mechanistic activity. Because of our selection of a
data extract, we hope that the outcomes are fairly transparent, but they
are by no means automatic. We need to decide, for example, not only
what aspects ofthe datato tagwith codesbut alsowhat levelsof generality
or detail to go into. As we have indicated in the extract above, we have
identified and defined three levels of generality. The most general cate-
gories are two: what makes a good PhD and the motivation to undertake
a PhD. In this instance-and by no means is this always true-the
categories correspond with the thrust of questioning in the interviews
and were part of the agenda followed in the semistructured interviews
themselves. We have here identified an intermediate category, having to
do with the abilities or competencies required for the successful comple-
tion of a PhD. The third and most specific level of categorybreaks down
thosc more general themes into more specificand detailed codes. As can
be seen horn how we have marked those codes on the extract, they also
late to stretchesofthe interviewofdifferent lengths.The more detailed,
specific codes are embedded within the longer, more general ones. This
is a fcature of dense, detailed coding that becomes especially important
in the con~puter-assistedhandling of qualitative data, which we discuss
in Chapter 7. The identification of codes and decisions about levels of
detail are far from straightforward. As Weaver and Atkinson (1994, p. 32)
put it:
[Ilf we decide to delineate a number a number of general, inclusive catego-
ries, much of the text will be coded with a single code (or co~ljunctioilof
codes). The advantage of this strategy is that it should maximize the usef~~l-
[lessof the codes; they are likely to be applied to enough segments to justify
thc purpose of reco~ltextualization.However, it may also have several disad-
vantages. First, since so much text will be coded with the same category, there
ii~ightbe difficulty in locating particular episodes significant to analysis; a
likely scenario is that the researcher will have to siphon through rearns of

38 M A K I N G S E N S E O F Q U A L I T A T I V E D A T A
irrelevant data, despite recontextualization. Second, coding may be too
crude, and this might make the analysisseem rather vague,lackingdetail,or
the exploratory avenues of analysisbeing superficiallyrestricted.
On the other hand, codingschemes that aretoo detailed canbe equally
problematic:
[I]fwe decide to define a large number of categories, with fairly exclusive
meaning, the problems are reversed. Coding will be more detailed and
intricate,and there will be a greater differentiationof segmentsaccordingly.
However,if thesegmentationoftextistoo intricate,in thatspecificcategories
are attached to very small segments of text, important contextualinforma-
tion may be lost, and thereby some of the segments' meanings. (Weaver&
Atkinson, 1994, p. 32)
As a consequence,Weaver and Atkinson made an explicit decision to
include codes of different degrees of generality so that data retrieval
could be undertaken at different levels.
We can develop and illustrate decision-making processes, as well as
the significance of different levels of coding, with another extract from
the interview data. As will be seen, it deals with a set of issues similar to
those identified from our first extract.
Extract From Interview WithDr. Throstle (Southersham University)
Odette Parry: What sorts of skills do you think it im-
parts, the actual process of doing a PhD?
Dr. Throstle: It's a very big question. Again I think in
anthropology you learn a whole lot of things that
you don't normally learn in aPhD, which ispartly
to do with fieldwork. That trains you to carry on
on your own both academically and personally,
it's social skills training of a very exacting kind.
.. .
I think one of the peculiarities of anthropology is
that unlike most other disciplines, certainly in the
social sciences, you're dealing-unlike history
for example-you don't start from one body of
documentation and convert it into another kind
of body of documentation, you start with people's
Concepts and Coding
lives and conversations with them and you have
to turn that into an academic text.
Which is why it takes such a long time and why it's
so difficult, because these two things are miles
and miles apart. And it's very common, I think,
for graduate studentswhen they come back from
the field to react against what they're doing, to
feel that what they're writing is somehow a be-
trayal or it falls far short of the relationships they
had when they were in the field. Writing a rather
dull piece of academic work somehow feelslike a
betrayal.
Nevertheless most students learn to do that, and in
the process learn an immense amount not only
about the people they study, but also writing
skills and how to produce a high-level academic
text.
I think one of the problems of fieldwork, if you're
away for a long time-and most anthropologists
are-you lose touch to some extent with your
academic and your home culture, and then you
have to get back into it, and it's often a slow
process when you come back.
It would be easy to treat this extract in much the same way as we did
the first extract, and to deal with it initially in terms of Odette Parry's
elicitation. We could thus relate it primarily to the kinds of skillsthat the
question asked about. Below, therefore, we have categorized the data
extract accordingly. It will be seen that we allowed ourselves to stick
closely to the overarching theme implied in the question about skills.
Odette Parry: What sorts of skills do you think it im- skills from
parts, the actual process of doing a PhD? PhD
Dr. Throstle: It's a very big question. Again I think in academic
anthropology you learn a whole lot of things that
you don't normally learn in a PhD, which is partly personal
to dowith fieldwork.That trains you to carryon on
your own boih academically and personally, it's social
social skills training of a very exacting kind. . . .

r a t i f r r u i I / i J : 3 1
40, MAKING SENSE O F QUALITATIVE DATA
I think one of the peculiarities of anthropology is writing
that unlike most other disciplines, certainlyin the
social sciences, you're dealing-unlike history
for example-you don't start from one body of
documentation and convert it into another kind
of body of documentation, you start with peo-
ple's lives and conversations with them and you
have to turn that into an academic text.
Which is why it takes such a long time and why it's writing
so difficult, because these two things are miles
and miles apart. And it's very common, I think,
for graduate studentswhen they come back from
the field to react against what they're doing, to
feel that what they're writing is somehow a be-
trayal or it falls far short of the relationships they
had when they were in the field. Writing a rather
dull piece of academic work somehow feels like a
betrayal.
Nevertheless most students learn to do that, and in writing
the process learn an immense amount not only
about the people they study,but alsowriting skills
and how to produce a high-level academic text.
I think one of the problems of fieldwork, if you're
away for a long time-and most anthropologists
are-you lose touch to some extent with your
academic and your home culture, and then you
have to get back into it and it's often a slow
process when you come back.
We can see here that so far we have produced a very thin and flat set
of categories. They reproduce only the bare bones of skills and do not
appear to do justice to the dense descriptive language of this particular
academic. Another way of approaching the task, therefore, is to pay
much closer attention to the categories of expression that the informant
actually uses. Rather than using the interviewextract as an extended reply
to our one question, therefore, we pay much closer attention to the
content of the talk. Here, therefore, we approach the data once more,
and now try to identify themes that reflect the informant's views more
closely.It is not necessary for us to use precisely the same words to index,
Concepts and Coding 4 1
or code, those themes; we are interested in exploring them and linking
them with other data segments, not only in labeling them.
OdetteParry:What sorts of skillsdo you think skillsfrom PhD
it imparts, the actual process of doing a
PhD?
i
I Dr. Throstle: It's a very big question. Again I fieldwork1
! think in anthropology you learn a whole normal P ~ DI
lot of things that you don't normally academic independence
learn in a PhD, which is partly to do with personal independence
fieldwork. That trains you to carryon on social skills
your own both academically and per- exacting
sonally,it's social skillstraining of a very
exacting kind. ...
I think one of the peculiarities of anthro- peculiarity
pology is that unlike most other disci- disciplines
plines, certainly in the social sciences, social science
you're dealing-unlike history for ex- history
ample-you don't start from one body people
of documentation and convert it into lives
c~notllcrkind of body of documentation, conversation
you start with people's lives and conver- academictext
sations with them and you have to turn
that into an academic text.
Which is why it takes such a long time and time
why it's so difficult, because these two difficult
things are miles and miles apart. And it's difference
very common, I think, for graduate stu- return from field
dents when they come back from the field writing
to react againstwhat they're doing, to feel relationshipsin the field
t h ~ twhat they're writing is somehow a dull
bctl-ayalor it fallsfar short of the relation- academicwork
shipsthey had when theywere in the field. betrayal
Writing a rather dull piece of academic
work somehow feels like a betrayal.
Nevertheless most students learn to dothat, people
and in the process learn an immense writing
amount not only about the people they high-level
study, but also writing skills and how to academic text
produce a high-level academic text.

M A K I N G S E N S E O F Q U A L I T A T I V E D A T A
I think one of the problems of fieldwork, if fieldwork
you're away for a long time-and most time
anthropologists are-you lose touch to absence
some extent with your academic andyour academic culture
home culture, and then you have to get home culture
back into it, and it's often a slow process return
when you come back.
In indexing the data in this kind of way, we can start to develop a much
denser set of themes and categories.We can start to glimpse some of the
recurrent preoccupations of this particular anthropologist, and we can use
such categorizationstobuild systematiccomparisonsand contrasts with the
viewsexpressedby other facultymembers. In looking atthe data in this way,
moreover, we can start to identify some further themes and issues.
Havingbegun by stayingcloseto the informant's own categories,we can
start to see how they might be categorized further, possibly in relation to
linking categories of our own devising. Take, for example, the first para-
graph of the informant's reply. It will be seen that here she is prefacing a
response about "shlls" by saylng that there is something distinctive about
anthropology as a discipline that makes it different from others, certainly
amongthe socialsciences.Thishelpsusto identifyasuperordinatecategoly,
which we might identifyas the distinctiveness of anthropology. If we identify
that as a category and code the data accordingly, then we can use that to
search for other data extracts in which the anthropologists express what is
specialand distinctive about their subject.
There is also a potentially intriguing theme to be constructed from
this anthropologist's words. Writing is identified as a major academic
shll to be acquired, but that thought is expressed in a particular, and
striking, way. Dr. Throstle talks of the difference between the academic
text that the student must prepare and the people who were studied.
People's lives and conversations have to be turned into text, and students
find the essential difference between the social and the textual to be a
problem. Not only is this a very interesting comment about anthropol-
ogy, but it also connects with our previous category in two ways. First,
it is offered as an example of how anthropology differs from other
disciplines (the requirement of constructing a body of text out of lives
and conversations). Second, it picks up on the significance of difference
in a new way.
Concepts and Coding 43
Let us look ahead in the data once more before commenting further.
The informant talks about the importance of fieldwork. Fieldwork is
introduced as one of the distinctive aspects of the anthropology PhD at
the beginning of the extract. The topic reappears later. Toward the end,
for example, fieldwork is contrasted with coming back, and there is a
contrast between fieldwork, on one hand, and academic culture and
home culture on the other. Again we see Dr. Throstle describing things
in terms of difference, in this case differences between cultures. We can
see now that there is a potential superordinate category relating to
difference that relates to all these aspects of the interview extract. The
differences that are described are different in content, but they all seem
to relate to a coherent set of underlying issues: the distinctiveness of
anthropology, the significance of fieldwork, and the separation of field-
work from other aspects ofthe anthropologist's life.We should note that
this theme, which relates closelyto the content of the data, is clearly one
that we have constructed. It is also one that takes us toward concepts of
a more analytic, even theoretical, relevance. We have thus moved our
coding process from identifying categories that remain close to the
original data to those that imply much broader analytic issues. We may
therefore return to the data once more and apply a further set of codes.
Odette Parry: What sorts of skillsdo you think
it imparts, the actual process of doing a
PhD?
Dr. Throstle: It's a very big question. Again I
think in anthropologyyou learn a whole
lot of things that you don't normally
learn in a PhD, which is partlyto dowith
fieldwork. That trains you to carryon on
your own both academically and per-
sonally, it's social skillstraining of a very
exacting kind. . . .
I hink one of the peculiarities of anthro-
pology is that unlike most other disci-
plines, ccrtainly in the social sciences,
you're dealing-unlike history for ex-
ample-you don't start from one body
of documentation and convert it into
another kind ofbody of documentation,
L
distinctiveness contrastive
of anthropology rhetoric
significance of
fieldwork
contrast anthropology
with other disciplines
difference between the
field and academic texts

44 M A K I N G S E N S E O F Q U A L I T A T I V E D A T A
you start with people's lives and conver- .
sations with them and you have to turn
that into an academic text.
Which is why it takes such a long time and
why it's so difficult, because these two
things are miles and miles apart. And it's
very common, I think, for graduate stu-
dents when they come back from the
field to react against what they're doing,
to feel that what they're writing is some-
how a betrayal or it falls far short of the
relationships they had when theywere in
the field. Writing a rather dull piece of
academic work somehow feels like a be-
trayal.
Nevertheless most students learn to dothat,
and in the process learn an immense
amount not only about the people they
study, but also writing skills and how to
produce a high-level academic text.
I think one of the problems of fieldwork, if
you're away for a long time-and most
anthropologists are-you lose touch to
some extent with your academic and your
home culture, and then you have to get
back into it, and it's often a slow process
when you come back.
differencebetween field
and academictexts
metaphor:
betrayal
differencebetween the
field and academic
culture
differencebetween the
field and home culture
In attaching codes in this way we have indicated some key generic
issues. That work has been done as more than a mechanistic exercise in
segmentingthe data. It reflects a series of readings and re-readings of the
data, in which the details of the interview and our own emergent con-
cerns interact. We may also note while dealing with the interview text at
this levelthat the repeated references to differences and distances suggest
Concepts and Coding 45
texts. In Chapter 4, we will consider in more detail how we can examine
such formal properties, and we will discussthis particular property again
when we consider contrastive rhetoric (Hargreaves, 1984) as one device
that actors use in producing their accounts of the social world.
In considering form and content here, we have noted one further
aspect of this academic's talk. We have identified the graphic way in
which the distance between field and text is expressed. We are told that
student anthropologists may feel that their work is a betrayal. Again, we
will return to discuss the exploration ofsuch figures ofspeech in Chapter
4, but we note that it may prove useful to identify figures of speech such
as metaphors during theprocess ofcoding.We have therefore added two
codes to the extract above, one identifying the use of contrastive rhetoric,
the other identifying the location of this particular metaphor.
We have been through this one extract from our data and applied to
it different, complementary sets of codes. For the sake of clarity, we have
displayed the data and the codes separately. In the course of repeated
examinations of the data for the purposes of a comprehensive analysis,
we would not necessarilyhave recourse to such separation.We can think
of the successive passes at the data as resulting in overlays of different
codes, ieflecting different levels of specificity or generality as well as
reflectingdifferent sets of analytic themes. These different approaches to
the data could result in different physical disaggregations of the data:
~hysicallycutting up different copies or cutting and pasting segments
into different files with the word processor. If we were using computer
software to perform these tasks, we would be able to retrieve the coded
segments by using different codes or combinations of codes to identify
them.
Our illustrative example is not an exhaustive treatment of the data.
Any other analyst could conceptualize them in different ways. The point
is not to search for the "right" set of codes but to recognize them for what
they arc: links between particular segments of data and the categories we
want to use in order to conceptualize those segments.
two further issues. In identifying these categories, we have also started
to move from a consideration of the content of the anthropologist's talk
in order to yay some attention to its form. We can note that his or her
Beyond Coding and Toward Interpretation
A ,
descriptions draw on a recurrent pattern, a series of contrasts: anthro- Coding clualitative data enables the researcher to recognize and re-
pology/other disciplines, fieldworklhome, and people's livesiacademic conlextudlize data, allowing a fresh view ofwhat is there. +cause coding

46 M A K I N G SENSE OF QUALITATIVE D A T A
inevitably involves the reading and re-reading of data and making selec-
tions from the data, it involves interpreting the data set. However, a key
issue is what to do with data once they have been selected, cut up,
fragmented, coded, and categorized. The move from coding to interpre-
tation is a crucial one, as Wolcott (1994) suggests. Interpretation in-
volves the transcendence of "factual" data and cautious analysis of what
is to be made of them.
Once coding is achieved, the data have to be interrogated (Delamont,
1992)and systematicallyexploredto generate meaning. There is a casethat
coding,whilereorganizing data, alsoinvolves a certain amount of informa-
tion loss. To some extent, that depends on how thorough and detailed the
coding has been. The data loss is much greater if one does not move from
the process of coding to an exploration of how codes and categories relate
to the original data, to other data, to theoretical ideas, and so forth. Ian Dey
(1993)arguesthat categorizingenablesone to think about the data in a new
way. This is only the case if we movebeyond the codes,categories,and data
bits back to what the "whole" picture is or may be.
The move from coding to interpretation has a number of discrete levels.
First, the coded data need to be retrieved. This essentially means that
recontextualized data need to be displayed in such a way that they can be
read easily.The data bits that relateto aparticular code or categoryneed to
be presented together in order for the researcher to explorethe composition
of each coded set. Huberman and Miles (1994) argue that data display is a
key element ofthe analyticalprocess. This can be achieved by organizingall
the data under a particular code physicallyin the same place;by producing
diagrams, matrices, and maps of the code; or by using a retrieval function
on a microcomputing program. Whichever way it is done, the idea is that
the codes or categoriesand the data need to be in such a form that they are
accessibleboth for reading and for exploring.
Second, the move from codingto interpretation involves playingwith
and exploring the codes and categories that were created. Dey (1993)
provides many ideas about howyou can go about doing this. He suggests
that once data are displayed in a coded form, the categories can be
retrieved, split into subcategories, spliced, and linked together. Essen-
tially, the codes and categories you have selected should be used to make
pathways through the data. It isworth remembering herethat such codes
are not cast in stone. As you chose and selected them, so you can
abandon, change, re-sort, rename, and so on. Similarly, once you are in
Concepts and Coding 47
a position to look at all the data across the codes, you should not be
tempted to ignore incidents, events, individuals, or chunks of data that
do not "fit" into the codes. The exceptions, misfits, and "negative"
findingsshould be seen as having as much importance to the process of
coding as do the easily coded data.
This leads to a further level of the process of moving from coding to
interpretation, that is, the transformation of the coded data into mean-
ingful data. Here, the emphasis is on what to look for in the codes and
categories. Delamont (1992) suggests that one should be looking for
patterns, themes, and regularities as well as contrasts, paradoxes, and
irregularities. One then can move toward generalizing and theorizing
from the data. The emphasis on the "negative" exceptions as well as the
"positive" patterns remains crucial. Huberman and Miles (1994) work
from a similar continuum. They suggest no less than 13 "tactics" for
generating meaning or data transformation. These move from descrip-
tive through explanatory tactics. At the one end of the continuum are
things such JS noting patterns of themes, the "counting" of phenomena
occurring from the data, and comparing and contrasting the data sets.
At the opposite end are the moves toward generalizing, noting and
questioning the relations between variables, and finding conceptual and
theoretical coherence in the data.
We already have illustrated how codes can represent categories of
different sorts.Some ofthose already implyinterpretiveframeworksand
link data segments to emergent concepts. Hence, we do not need always
to think of coding first and theorizing afterward. Our decision making
implies andlytic ideas at evely stage of the coding process. Furthermore,
wc alrcady have seen that our codings can imply systematicrelationships
among categories and concepts. These relationships can form one basis
for the development of interpretations. For example, if we return to the
second of our data extracts, we can see that our codings start to suggest
possible relationships. The different levels of coding suggest that some
c,ltego~iesmay subsume a number of others. This is so for several
ciifkrent codings. For example, Dr. Fitton's responses in the first extract
suggest the following kinds of relationships:
S t ~ ~ d e n tmotivatiol~ Academic career
Intrinsic interest
Altruistic colnrnitlne~lt ;

48 MAKING SENSE O F QUALITATIVE DATA
Good PhDs Originality
Theoretical knowledge
Logical coherence
Quality of fieldwork
In a different way, we have seen how Dr. Throstle's coded comments
can be linked through a common use of difference and distance, which
could linked in the coding scheme through a code of contrast. In other
words, our codes not only establish linkages between data and concepts;
they also can start to map out dimensions within conceptual categories
and to establish superordinate links amongconcepts. We can summarize
this set of more abstract ideas as follows:
Difference Home culturelField
FieldIAcademic culture
People's livesIAcademic texts
AnthropologyIOther disciplines
We will suggest in Chapter 6 that although there is more to it than
linking codes, the establishment of ordered relationships between codes
and concepts is a significant starting point for reflection and for theory
building from qualitative data.
Strauss (1987) develops the use of coding as part of the process of
interpretation and analysis. That is, Strauss links the initial process of
coding (which he refers to as open coding and which is essentially what
we did with the interview extract with which we exemplified our earlier
discussion) to a more refined process of using categories to generate
broader conceptual frameworks. Strauss identifies a set of procedures
that allows initial categories to be elaborated and developed.
We are not going to follow Strauss's own model closely. We do not
recapitulate all the steps and analytic strategies that he identifies or has
codified in his work with Corbin (Strauss & Corbin, 1990).Indeed, it is
worth noting in this context that Glaser (1992) accuses Strauss of taking
the general inspiration and strategicapproach of "grounded theory" and
transforming them into unduly prescriptive recipes for analysis. Never-
theless, the general approach to coding that is to be found in Strauss's
methodological writing is valuable in encouraging the researcher to
move beyond local codings to generate ideas and broader conceptual
frameworks.
Collcepts and Coding 49
In essence, Strauss's approach encourages us to go beyond the essen-
tially "summary" approach to coding in which the data are simply
reduced to a limited set of categories. Strauss does not encourage us
simply to index the data, as it were, or to use the code words merely to
mark and retrieve segments of data. On the contrary, his general ap-
proach exhorts us to expand on rather than to reduce the data, to take
categories and exhaust their full analytic potential. One point is to use
our codings and categories to think with and not to remain anchored in
the data (notes,transcripts, etc.) alone.
From this general perspective, the process of coding is about asking
oneself questions about the data. Those questions help to develop lines
of speculation and hypothesis formation. In accordance with more
general principles of grounded theory, they may also direct further data
collection strategies. In the course of open and axial coding, then, one
takes as u topic a "phenomenon" (in Strauss's terms) and attempts to
identify its dimensions, its consequences, and its relationships with other
phenomena.
For example, if we return to our previous data extract, we start to
think more creatively about some of the themes that we have begun to
identify in our coding. Consider once more the same interview extract
(see py. 38-39). When we first coded that, we were able to identify
stretches of the transcript that seemed to refer to theoretical sophistica-
tion, analytic value, originality, fieldwork, indeterminate knowledge,
contribution to the discipline, and empathy with the people studied.
These were all in an informant's response to the question ofwhat makes
a good PhD, dnd all those separate code words, and the fragments they
refer to, all seem to bear on one major phenomenon, that is, the essence
ofwhat it takes to produce a successful PhD in social anthropology. We
can, therefore, start to think about the various possible "dimensions" of
such dbilities or competencies.
In this one account, we find the established anthropologist construct-
ing his own characterization of how the competencies are (or should be)
combined in one ideal-typical doctoral candidate. In thinlung more
about this phenomenon, we need to ask ourselves, and to ask ofthe data,
what sorts ofabilities or competencies arerecognized and described. Our
first informant has given us the abilities to theorize, to organize and
structure coherent arguments, to empathize with the people studied, to
make an original contribution to the discipline, to do fieldtvork, and to

L-_a 1 1 L - a I 1 1 L E I G 1 8 1 2
50 M A K I N G S E N S E OF Q U A L I T A T I V E D A T A
5
integrate fieldwork and theory. Further, we note that these abilities are
generally characterized as indeterminate qualities rather than being
rendered explicit. In looking at other interviews, we note that other
categories and codings bear on the same general phenomenon. They are
expressed in different ways, for example, the abilities to turn other
people's lives and conversations into academic text, to do scholarship,
to be original, to be critical, to add to human understanding, to put a
new slant onwhat's been written before, and to contribute to the ongoing
debates. Successful anthropologists should also have a distinctive ap-
proach and exhibit openness, humility, and the ability to reflect upon
themselves. The open coding of a series of interviews would therefore
give us a wide range of dimensions of "ability" or "competence" or
"quality" (the general phenomenon is very general indeed).
We can also go on to think about some further features. For example,
under "axial" coding, Strauss recommends thinking about such features
as consequences. Following that line of reasoning, we can go beyond the
data immediately at hand and ask ourselves such further questions as
"What happens if. . .?" For example, we can ask "What happens if there
isn't any fieldwork?" That question can inform further inspection of the
data, further data collection asthe research develops, or both. Inspection
of our anthropology data suggeststhat there may be different answers to
that question, dependingon the antecedents or causes of fieldwork being
absent (another of Strauss's features of axial coding). For example,
fieldwork may not take place because the graduate student has decided
to pursue "library research" (a term used by our informants). Further
exploration of consequences and antecedents shows that library researcl~
is often discouraged, and the absence of fieldwork, though deliberate,
may be stigmatizing. On the other hand, fieldwork may be incomplete
because of conditions beyond the student's control (such as civil disor-
der). This is not stigmatizing, though it may prove to be a major
handicap and is unfortunate. Such observations again help us to open
up questions about the significance of fieldwork in the construction of
a professional career and an academic identity.
In the same vein, we can ask ourselves what the consequences are for
the writing and organizing of the thesis. What are the consequellces of
not having the ability to integrate theory and fieldwork, or of not making
a "contribution" to the area of specialization? Such a reflection perhaps
would take us to other aspects of the data. We might look for acaden~ic
Concepts and Coding 51
supervisors.' accounts of how PhD theses get examined and what aspects
of organization and style are rewarded or penalized. We also need to
explore the students' accounts of their writing, looking for such things
as their strategies for writing, their experiences of writer's block, and
their strategies for writing. We need to examine our data (or, ideally,
collect further data) on how our anthropologists recognize and reward
"a contribution" to the discipline or to a particular debate within it. In
procedural terms, this means cross-reference to the "open" codings
identified elsewhere in the data. In intellectual terms, it means using the
various elements in the data to pursue lines of interrogation and specu-
lation-moving between data and codings to explore and expand on key
analytic themes.
We might also follow others of Strauss's analytic strategies in cod-
ing-thinking about conditions and antecedents, perhaps. We can ask
ourselves, for example, about the antecedents of fieldwork. This ap-
proach would probably lead us to think about and to search out data
concerning the necessary preconditions (organizational, personal, aca-
demic) to embarking on fieldwork. We would then find ourselves ask-
ing-and interrogating the data-about when fieldwork is possible and
who gets to decide about it (where,under what circumstances,with what
intellectual prcparation). We would also start to explore some of the
more mundane antecedents and conditions, the practical tasks and
relationships that go into the practical work of fieldwork. We might thus
start to generate ideas and themes that lead us to think seriously and
systemntically about academic and practical dimensions of anthropo-
logic,~lfieldwork. One would hope to be generating themes that facili-
tated comparative thinking and exploration, for example, contrasting
with other travelers to exotic parts, such as war correspondents, travel
writers, workers in international transport, or others who live on a
long-term bas~saway from their home base and creature comforts.
Conclusion
In this chapter, we have explored some of the rationales for coding
clualitdtive data and introduced some of the different approaches to be
found in the methodological literature and in practice. This discussion
has not becn intended as a comprehensive cookbook as to how to

52 M A K I N G S E N S E O F Q U A L I T A T I V E D A T A
perform coding. We have preferred to discuss contrasting and comple-
mentary strategies rather than prescriptively recommending a single
orthodoxy. We alsohave rriedto suggestfromtime to time that "coding,"
however it may be conceptualized and carried out, is not the final word
on qualitative data analysis.
The segmentation of field data and retrieval of marked data segments
is a valuable resource in the management of qualitative data. It is an
established approach that in recent years has been reinforced by the
development of microcomputing strategies (Weaver & Atkinson, 1994),
many of which essentially recapitulate the same logic of data handling.
They substitute rapid and comprehensive searching supported by soft-
ware for the uncertain and slow process of manual searching and filing.
In and of themselves, however, such procedures by no means exhaust
either the possibilities of the data or, therefore, the possibilities of data
exploration. In particular, it should be apparent that the fragmentation
of data, dependent on code-and-retrieve approaches, or what Tesch
(1990) called the decontextualization of data, does little to preserve
formal features of those data.
Our interview informants may tell us long and complicated accounts
and reminiscences. When we chop them up into separate coded seg-
ments, we are in danger of losing the sense that they are accounts. We
lose sight, if we are not careful, of the fact that they are often couched in
terms of stories-as narratives-or that they have other formal proper-
ties in terms oftheir discourse structure. Segmenting and coding may be
an important, even an indispensable, part of the research process, but it
is not the whole story. Consequently, we turn to a consideration of
narrative analysis in the chapter that follows. We will see how one can
look at the same sort of data from a fresh perspective, paying due
attention to its more formal properties and being sensitive to the storied
quality of many qualitative data.
Suggestions for Further Reading
Dey, I. ( 1993) Qualitative data ana1ysis:A userfiiendlyguidefor sociul scientists. London:
Routledge and Kegan Paul.
An unusually detailed description of theprocesses ofcatcgorizi~rgarrdcodi~~gq~~~~lit~~ti~~~
data. Illustrated with data extracts frorn Woody Allen ai~dVictorin Wood scripts.
Concepts and Coding
53
Addressed to the management of coded data on computer. Dey is the developer of the
software application called Hyperresearch.
Miles, M. B., & Huberman, A. M. (1994). Qualitative data analysis:An expanded source-
book (2nd ed.).Thousand Oaks, CA: Sage.
An exceptioizally systematic and formal approach to the classification and repre-
sciltntioil ofqualitative data. A uiziquely rich sourcefor analytic ideas, with a strong
visual ci?~phasis.
Strauss, A. L. ( I 987). Qualitativeanalysisfor social scientists. Cambridge, UK:Cambridge
University Press.
Ail urllistrally detailed account of Strauss's own approach to data analysis. Illustrated
fhrolcghotlt with extractsfrom Strauss's postgraduate seminars, giving a particularly
privileged insight into his preferred approach.
Strauss, A. L., tL Corbin,J. (1990).Basics ofqualitative research:Grounded theory,procedures,
and tcchitiques. Newbury Park, CA: Sage.
An ncccssiblcii~troductionto codiizgprocedures,followii~gStrauss's approach togroilizded
theorizing. Gocs beyond coding itselftodiscuss the development of ideas and theoretical
fiaii~cworks.
Tesch, R. ( 1990). Qlrnlitative research:Analysis types and software tools. London: Falmer.
Coiwbiiles a gerleral discussion ofanalytic strategies with a survey of computer sofware
packagcs for qunlitative data matlagemellt and analysis,