See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/289099722

Coding transcripts and diaries

Article · January 2010

CITATIONS READS
78 2,126

1 author:

Meghan Cope
University of Vermont
29 PUBLICATIONS   1,116 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Mapping American Childhoods View project

Qualitative and Mixed Methods in Geography View project

All content following this page was uploaded by Meghan Cope on 03 January 2017.

The user has requested enhancement of the downloaded file.

Part III
“Interpreting and
Communicating”
Qualitative Research
18
Organizing and Analyzing Qualitative Data
Meghan Cope

Chapter Overview

This chapter discusses some ways qualitative data can be organized and analyzed
systematically and rigorously to produce new knowledge. We begin with some
comments on how to make sense of your data, and explore practices of
“memoing,” concept mapping, and coding. Much of the subsequent discussion
revolves around coding as a process of distilling data and identifying themes. The
chapter reviews different types of codes and their uses, as well as several ways to
get started with coding in a qualitative project. Specifically, a distinction is drawn
between descriptive codes, which are category labels, and analytic codes, which
are thematic, theoretical, or in some way emerge from the analysis. The building
of a “codebook” is also discussed, stressing the importance of looking critically at
the codes themselves, identifying ways in which they relate, minimizing overlap
between codes, and strengthening the analytical potential of the coding structure.
Finally, several related issues are covered, such as coding with others, the use of
computer-aided qualitative data analysis software (CAQDAS), and integrating
coding and mapping.

Introduction: Making Sense of Data

As is evidenced by this volume and the notable expansion, over the past decade, of publications
on qualitative methods generally in the field of geography, scholars and students are increasingly
engaged not only in doing qualitative research but also in thinking and writing critically about
methodologies, including the ways that we evaluate, organize, and “make sense” of our data (see
also Clifford, French, and Valentine, 2010; DeLyser et al. 2010; Gomez and Jones, 2010). As the
“qualitative turn” in geography and related fields has matured, the techniques for gathering and
producing data have expanded in scope and depth; this, in combination with critical reflection on
issues of methods, context, and researcher positionality, has produced ever more thoughtful,
reflexive, and creative research projects and, ultimately, a more robust production of knowledge.
It is an exciting time to be doing qualitative geographic research, and, appropriately, the
standards are high, requiring practitioners to be well-informed and attentive to principles of
rigorous scholarship.
Towards that goal of rigorous scholarship, the practices of organizing and analyzing
one’s data deserve special attention. The job of the researcher is, in effect, one of synthesis and
translation; we are tasked with observing and engaging with the world, making sense of the
resulting data, and representing (re-presenting) the facts, stories, ideas, and events that were
shared with us in a coherent manner. The focus for this chapter will be on the middle piece of
this—making sense of data—while recognizing that it cannot be wholly separated from the other
parts of the process.
Increasingly, qualitative data are taking more diverse forms, well beyond the classical
text-based transcribed interviews, oral histories, archives, or diaries (though these are all still
very important sources), into visual data (photos, video, maps, sketches); sources resulting from
participatory engagement with community partners (collaborative maps, exhibits, activism); and
varied creative techniques such as walk-along interviews, artwork, theatrical productions, photo-
voice, self-directed video, vernacular mapping, material culture analysis, and other methods that
foster less hierarchical interaction with participants. In the context of such an expansion of
research techniques, particularly those based less and less on text, it is natural that we should
revisit some ideas about “making sense of data.” I propose, however, that many of the same
principles that have appeared in previous editions of this volume remain valuable: organizing
and reducing data to manageable chunks, identifying themes, and paying attention to rigorous
interpretation. Towards that end, I review some common techniques for the purposes of
description, classification, and connection (Dey 1993) of qualitative data; these techniques are
memoing, concept mapping, and coding.
Making Meaning I: Memos

A memo (short for memorandum) is usually a short note to oneself or research collaborators
jotted to capture a quick insight, to serve as a reminder of a future task, or to draw connections
between multiple referenced items. In ethnography, researchers typically rely on heavy use of
field notes to record the events, dialogues, and observations made while engaging with a
community, yet ethnographers often use memos as a sort of intermediate-level mechanism to
remind themselves of something, to reflect on patterns or connections, to contextualize events,
and to forge new links between emerging themes. For example, Annette Watson and Karen Till
(2010) provide a particularly rich example of the use of memos as a first stab at interpretation
and reflection on their experiences and the material in their field notes:

Both of us understand our writing as a safe space to explore, think through, and represent
knowledges, desires, and fears. In memos, we question our experiences and assumptions,
pay attention to processes, respond to our embodied and emotional presences, consider the
material and visual cultures that constitute what is being studied, scrutinize various
relationships with research fields and partners, and elaborate upon our insights. We also
make connections to other studies of previous work in memos, and raise critical questions
that inform future theoretical readings. (Watson and Till 2010, 128)

The crucial elements of memos, as demonstrated by Watson and Till, can be summarized
as follows. First, memos are quick and informal. They are jottings that by their very nature allow
the researcher space and freedom to explore possible connections and viewpoints, to put an idea
down in its earliest undeveloped form, and to remind oneself to pay attention to this or that in the
future. Second (and despite its informal nature), memoing constitutes a valuable interpretive
practice towards making meaning of the data. Along with margin comments, sticky notes,
sketches, and other annotations researchers commonly make on and about their data, memos are
useful for sorting out ideas, identifying patterns and similarities, recording “Aha!” moments, and
generally beginning the process of organizing and analyzing. Finally, memos are reflexive. That
is, they foster critical review and contemplation, including self-reflection; memos represent a
chance to step back and consider alternate interpretations, to critique one’s own role in the
research process, and identify—however tentatively—linkages between events and discussions
that might not have been seen while in the midst of research mode.
The benefits of memoing are not limited to participant-observation research practices, of
course. Memos can be useful for any research project across the spectrum of qualitative,
quantitative, and mixed-methods scholarship because they serve the important functions of
reminding us of ideas to return to, as well as connecting, processing, and critiquing the process
and early findings of the research endeavour. Reflecting the value of memoing, most CAQDAS
programs include memo functions, including some that allow one to make memos about memos!

Making Meaning II: Concept Mapping

Despite the word “mapping” in the name, there is little formal discussion of concept mapping in
the qualitative geography literature, yet it is an important organizational and analytic strategy
that most researchers probably do without really thinking too much about, perhaps because there
is a certain intuitiveness to the process. At its most basic level, concept mapping involves
visualizing data and their relationships. This could be performed through the process of
categorizing, that is, sorting data into piles (literal or digital) that have some internal cohesion.
For example, in studying teen mobility my research partner and I used open-ended survey
questions to query parents’ perceptions of their teens having driver’s licences; we immediately
saw that “safety” and “convenience” were major themes and, in one analytic exercise, began to
sort survey responses along those lines, with subordinate factors contributing to the main
concepts. By sorting and re-grouping responses on a large whiteboard, using statements from
parents who mentioned the convenience of having teens drive, we were able to quickly identify
some clusters: “working parent(s)” signified that both parents were employed full time (or that a
single parent was employed full time); “rural/remote” indicated that these families tended to live
in more rural areas and at farther distances away from school, home, and workplaces; and
“younger children” indicated the households in which teens had younger siblings who also
needed chauffeuring around to activities and events. Discerning these kinds of connections
through the relatively quick process of concept mapping—really a visual brainstorming
exercise—allowed us to form some initial insights that were then followed up with more
systematic examinations of the correlation between “convenience” and the household situations
of those parents and teen drivers. While sorting and piling are common components of concept
mapping, other methods are also frequently employed, such as word clouds, flow charts,
relationship trees, and non-hierarchical graphical plottings that begin to drill down into key
associations within data.
Although the example mentioned here was intended as a relatively quick, exploratory,
and coarse-level analysis, concept mapping has evolved into a tool that is much more formalized
and the results can in themselves be part of the final representation of the research; the latter has
gained traction in such fields as planning, organizational research, and evaluation studies, but
could lend itself to diverse disciplinary fields. Kane and Trochim (2006) acknowledge both the
informal notion of concept mapping (as just discussed) and the much more systematized
methodology in the very first sentences of their book on the topic:

Concept mapping is a generic term that describes any process for representing ideas in
pictures or maps. [Here] however, we use the term only to refer to one specific form of
concept mapping, an integrated approach whose steps include brainstorming, statement
analysis and synthesis, unstructured sorting of statements, multidimensional scaling and
cluster analysis, and the generation of numerous interpretable maps and data displays.
(Kane and Trochim 2006, 1. Emphasis in original)

In this method, which combines qualitative and quantitative analysis, researchers begin their
analysis of, say, open-ended survey questions by isolating unique “statements” (single-topic
fragments of longer responses), then sorting those statements into groups based on their
substantive relationships. This “sort/pile” part of the method could be done by hand using slips
of paper and actual piles or using CAQDAS functions digitally (for example, using the “network”
functions in Atlas.ti, or the “code map” function in HyperRESEARCH). This in and of itself may be
very helpful to researchers and, indeed, my research partner and I stopped at this point, having
identified some valuable conceptual–empirical linkages that warranted further exploration. Kane
and Trochim, however, go on to perform analyses of the clusters that result from the sort/pile
method, including multidimensional scaling (which assesses clusters depending on the strength
of their relationship or similarity) and cluster analysis (to assess connectivity, conceptual
distance, and so forth). For those who are interested in these latter, more statistical, functions, the
work of Kane and Trochim (2006) and their various research partners is recommended.
Ultimately, concept mapping can be as informal or formal as is suitable for the goals of
making sense of data. For many geographers and other “visual” thinkers, it can be a helpful way
of seeing data and relationships in different dimensions, it works well as a group analysis
exercise with multiple researchers and research assistants, and, most importantly, concept
mapping leads to new insights and connections being made.

Making Meaning III: Coding

Coding social data (text, images, talk, interactions) is sometimes derided as tedious, but if you
think of it as a kind of detective work, it can be intriguing, exciting, and very valuable to the
research process. The purposes of coding are partly data reduction (to help the researcher get a
handle on large amounts of data by distilling key themes), partly organization (to act as a
“finding aid” for researchers sorting through data), and partly a substantive process of data
exploration, analysis, and theory-building. Further, different researchers use coding for different
reasons depending on their goals and epistemologies; sometimes coding is used in an
exploratory, inductive way such as in grounded theory in which the purpose is to generate
theories from empirical data, while other times coding is used to support a theory or hypothesis
in a more deductive manner. Several approaches are discussed here, with pointers on how to
organize and begin the coding aspect of a research project.
A short caveat is necessary here: it is important to recognize that coding is not always the
best or only way to rigorously understand qualitative data. Even in his coding manual, Saldaña
points out, “No one, including myself, can claim final authority on coding’s utility or the ‘best’
way to analyze qualitative data… there are times when coding the data is absolutely necessary,
and times when it is most inappropriate for the study at hand.” (2012, 2. Emphasis in original).
Narrative analysis and discourse analysis are two methods that are more appropriate for many
projects and are covered in Chapter 13 of this volume as well as other texts (see Doel 2010;
Dixon 2010).
Types of Codes and Coding
One common type of coding is content analysis, which is essentially a quantitative technique
and by no means represents the full extent of coding for qualitative research. Content analysis
can be done by “hand” or by computer (see discussion of CAQDAS programs below), but either
way it is a system of identifying terms, phrases, or actions that appear in a text document, audio
recording, or video and then counting how many times they appear and in what context. For
example, a researcher might be interested in how many times the word “democracy” is used in
newspaper articles from a particular country or she/he might be interested in how places are
portrayed in a television program. Frequently in content analysis, sampling is used in similar
ways to quantitative analysis of populations; perhaps only front-page newspaper stories are
included in the analysis, or a television program is sampled for five minutes out of each hour.
Similarly, researchers using content analysis typically subject their coded findings to standard
statistical analysis to determine frequencies, correlations, variations, and so on. There are many
good guidebooks and instructions for conducting content analysis, including some available on
the Internet and broader methods texts. While content analysis has its place, the practices of
coding qualitative data go well beyond merely quantifying them; we consider those practices
next.
There are many ways to approach coding (in geography, see Crang 2005 for additional
strategies), but to simplify, I discuss two main types of codes—descriptive and analytic.
Descriptive codes reflect themes or patterns that are obvious on the surface or are stated directly
by research subjects. Descriptive codes can be thought of as category labels because they often
answer “who, what, where, when, and how” types of question. Examples of descriptive codes
that might interest geographic researchers include demographic categories (male, female, young,
elderly), site categories (home, school, work, public space), or even scale identifiers (local,
regional, national, global).
One special type of descriptive code is called in vivo codes; these are descriptive codes
that come directly from the statements of subjects or are common phrases found in the texts
being examined (Strauss and Corbin 1990). For example, Jacquie Housel did interviews with
elderly women about their daily routines and spaces in Buffalo, NY, and they repeatedly
mentioned concern with crime in their neighbourhoods (see Housel 2009). In this instance,
“crime” became an in vivo descriptive code because the term is used by and describes something
important to the subjects. In vivo codes are a good way to get started in coding, particularly in
projects that are designed to be inductive (moving from data to theory) or exploratory. Other
descriptive codes are generated by the researchers’ interactions with the data as they sort through
the sources, but they tend to—as their name implies—be fairly superficial.
Qualitative researchers also develop analytic codes to code text (and other forms of data)
that reflect a theme the researcher is interested in or that has already become important in the
project. Analytic codes typically dig deeper into the processes and into the context of phrases or
actions. For example, in Housel’s work it became apparent that the elderly white women just
mentioned were experiencing an change of status in the neighbourhood as its demographics
shifted to younger Black families: while in the past the women had been seen as the matriarchs
of the local community, they increasingly experienced what they perceived as being ignored,
disrespected, or even threatened by “newcomers” marked by race and cultural differences. Based
on this shift, “erosion of white privilege” served as an analytic code (Housel 2009, 134). This
code might then be applied to the rest of the data to identify other instances of fear and loss,
perceptions of particular social actors (e.g., young men, Black families moving into the area),
and other negatively perceived experiences in the disinvested and deteriorating public spaces of
the neighbourhood.
Often, descriptive codes bring about analytic codes by revealing some important theme or
pattern in the data or by allowing a connection to be made (for example, crime, fear of youth in
public, shifting meanings of “race”), while other times analytic codes are in place from the
beginning of the coding process because they are embedded in the research questions. For
instance, if Housel was interested from the start in how elderly women navigate urban spaces,
their personal mobility and perception of their neighborhoods would be themes reflected in the
analytic codes from the project’s very beginning. The recursive strength of coding lies in its
being open to new and unexpected connections, which can sometimes generate the most
important insights.

The Purposes of Coding

Three main purposes for coding qualitative material can be readily identified: data reduction,
organization and the creation of searching aids, and analysis. As the prolific French theorist
Henri Lefebvre noted, “Reduction is a scientific procedure designed to deal with the complexity
and chaos of brute observations” (Lefebvre 1991, 105). Qualitative research often produces
masses of data in forms that are difficult to interpret or digest all at once, whether the data are in
the form of interview transcripts, hours of video, or pages of observation notes. Therefore, some
form of reduction, or abstracting, is desirable to facilitate familiarity, understanding, and
analysis. Coding helps to reduce data by putting them into smaller “packages.” These packages
could be arranged by topic, such as “instances in which environmental degradation was
mentioned,” or by characteristics of the participants such as “interviews with women working
part-time,” or by some other feature of the research context or subjects such as “observations in
public spaces.” By reducing the “chaos of brute observations,” data reduction helps us get a
handle on what we have and allows us to start paying special attention to the contents of our
data.
The second purpose of coding is to create an organizational structure and finding aid that
will help us make the most of qualitative data. Similar to data reduction, the organizational
process mitigates the overwhelming aspects of minutiae and allows analysis to proceed by
arranging the data along lines of similarity or relationship. An important background step is
constructing and maintaining a complete record of sources, dates (of participant observation,
interviews, or focus groups, for example), subject contact information, and other relevant
information. While this database is not part of the coding process per se, it is a valuable part of
organizing qualitative material for coding and analysis and also allows the researcher to find
specific data more easily. For example, interview transcripts might be coded not only for their
content but also by their circumstances—was the interview conducted in the participant’s home?
were others present? did the subject seem nervous?—which can help organize information. With
better computer-assisted qualitative data analysis software (CAQDAS) available now (see below),
organizing and searching within electronic documents is greatly simplified. Additionally, coding
itself is also an important aspect of organizing and searching because it is essentially a process of
categorizing and qualifying data. “The organizing part will entail some system for categorizing
the various chunks [words, phrases, paragraphs], so the researcher can quickly find, pull out, and
cluster the segments relating to a particular research question, hypothesis, construct, or theme”
(Miles and Huberman 1994, 57). While the development of the coding structure is by no means
a simple process, it is one that—if done well—enables the data to be organized in such a way
that patterns, commonalities, relationships, correspondences, and even disjunctures are identified
and brought out for scrutiny. Good organization also means the process will be more rigorous, an
important consideration for defending one’s work as reliable, transparent, and trustworthy
(Baxter and Eyles 1997), even if it is not “generalizable” in the statistical sense.
The final, and principal, purpose of coding is analysis. While strategies for analytical
coding will be examined in greater detail in the next section, at this point it is sufficient to note
that the process of coding is an integral part of analysis. Rather than imagining that analysis of
the data is something that begins after the coding is finished, we should recognize that coding is
analysis (and is probably never truly “finished”!). Coding is in many ways a recursive juggling
act of starting with initial codes that come from the research questions, background literature,
and categories inherent in the project and progressing through codes that are more interpretive as
patterns, relationships, and differences arise.
Coding, as with memoing, also opens the opportunity for reflexivity, that critical self-
evaluation of the research process (see Chapter 2). By recursively reviewing data and the
connections between codes, researchers can also come to see elements of their own research
practice, subjects’ representations, and broader strategies of knowledge construction that had not
previously been apparent. Even if it is difficult to be self-critical in the midst of fieldwork or data
collection, the process of coding is inherently more contemplative and analytical and thus offers
a moment ripe for reflection.

Getting Started with Coding

The preceding discussion of types of codes addressed two main approaches to coding, which
may be seen as descriptive and analytic codes, although other terms are also used in the literature
(for example, initial codes and interpretive codes). The key distinction is that one type of code
is fairly obvious and superficial and is often what the researcher begins with, such as simple
category labels. The other type of code is interpretive, analytic, and has more connections to the
theoretical framework of the study; it tends to come later in the coding process after some initial
patterns have been identified. When coding is done manually, researchers develop a codebook—
a list of codes that are categorized and organized repeatedly. Although current qualitative
software packages typically do not use the term codebook, it is a useful concept that has
relevance whether the codebook is actually a tangible item in manual coding or merely an
abstraction in electronic coding.
 To start a codebook, it is easiest to begin with the most obvious qualities, conditions,
actions, and categories seen in your data and use them as descriptive codes. These elements will
emerge quite rapidly from background literature, your own proposal, other research-planning
documents, and the themes that stick out for you from gathering qualitative data (for example,
memorable statements in interviews, notable actions seen while doing participant observation,
key words that jump out in first readings of historical documents). For example,1 in my work on
how urban children conceptualized city spaces in Buffalo, New York, one of my original
interests that was heavily present in my grant proposal was how children in the 8- to 12-year-old
age group define “neighbourhood” and “community” (see Box 18.1). These terms are obvious
starting points for my codebook.

<CATCH insert Box 18.1>

29 November 2003. After-School Program Observation Notes, Children’s Urban
Geographies Research

Text: Field Notes Descriptive and Analytic Codes and Notes

from the Quilt Project Category Codes Themes
As I was setting up, Jakob Early release day
Jakob*, Mariana, Mariana from school–kids
and Ari came over Ari seemed wild and
and then Izzy and Izzy bored.
Salomé (a new girl I Salomé
hadn’t met before). Relations: Izzy and
We set up at a round Salomé are friends
table in between the
bench and the
“café,” near the pool
table. The noise
level was very high Tape recording would
and I had a hard not have worked here!
time hearing the
children at my table.
Next to us, three or
four younger boys
(Stefan et al.) were Stefan Relationship between
playing a war board gender and violent
game and making play?
lots of terrible
noises (at one point Research setting
I asked them to be conditions: loud
quieter).
After I explained
what I had planned Gave very loose
(to use the materials instructions to allow
to show your house children freedom
or apartment within the project’s
building and scope.
family), I asked the
children what a “Neighbourhood”
“neighbourhood” is.
There were varying
answers
immediately, mostly
around the idea of
“a bunch of houses
next to each other.”
Izzy said, “It’s when Play
you have one house
and then another
one and you all get
together to play.”
Mariana said, “I
don’t live in a Mariana seems proud
neighbourhood, I’m “Community” of her West Side
part of the West identification.
Side Community”(!) Difference between
I couldn’t hear very “neighbourhood” and
well, so I got out my “community”
notebook and went
around the group to
write down answers.
Ari’s answer was Tactic: attention
very long and Ari (age 5) seems to
complicated with crave attention.
something to do
with your “home
friend.” I’d like to
revisit the question
of what is a
neighbourhood in
video interviews. Future work–video
Then I got out my Technology interviews??
digital camera and
took pictures of the Nate
group (all five gave The children seem to
full permission for love technology and
this). Nate came up the gizmos we bring
 and wanted to “see” in get a lot of
the camera, which I Interactions: Nate’s Children’s attention.
didn’t want to let bullying influences identification with a
him do because he my perception of him community or
is so volatile and neighbourhood
unpredictable.
Reluctantly, I let
him take a picture of
me with the children
working on the quilt
and retrieved the
camera from him
immediately. Review and code
[Ironically, the photos.
photo Nate took is Our relationships with
one of my favourites specific children
of this project!]

*All children’s names have been changed

<end CATCH>

However, codes can also be too general and become cumbersome. Because much of my
children’s urban geography research was centred around issues of neighbourhood and
community, I found that I needed to break each of them into more specific codes, such as codes
for the particular neighbourhoods the children referred to, the use of both “neighbourhood” and
“community” to mean “local” (such as in reports from the city newspaper), and the way that
school curriculum materials defined “community.” This is a frequent characteristic of coding: an
initial category becomes overly broad and must be refined and partitioned into multiple codes.
Bear in mind that the opposite also occurs—some codes die a natural death through lack
of use. For instance, in my project I had naively expected the children, who were for the most
part in low-income families, to talk about a lack of money or not being able to afford something
they wanted. However, after four years in the project, I found little evidence of children
discussing their own poverty (though that absence is itself an interesting research question). So,
while other materials in the project necessitated that I keep ‘low income’ as a code, it was rarely
used in analyzing direct quotes from children. As Miles and Huberman said, “some codes do not
work; others decay. No field material fits them, or the way they slice up the phenomenon is not
the way the phenomenon appears empirically. This issue calls for doing away with the code or
changing its level” (1994, 61).
Thus, the first step is to make a list of what you think are the most important themes up
front, with the understanding that some of them will be split into finer specifics while others will
remain largely unused. But how do you know what is important? Anselm Strauss, one of the
founders of grounded theory, had a helpful system for beginning this awesome task (best
represented in Strauss and Corbin 1990). He suggested paying attention to four types of themes:

<CATCH bullet list>

• conditions
• interactions among actors
• strategies and tactics
• consequences

<end list>

“Conditions” might include geographical context (both social and physical), the circumstances of
individual participants, or specific life situations that are mentioned or observed (for example,
losing a job, becoming a parent, a child changing schools). By thinking along the lines of
“conditions” and coding only for them, the coding process is easily started, and you may learn a
lot about your data in a short time.
The same is true for limiting your scope to “interactions among actors”—if you focus on
relationships, encounters, conflicts, accords, and other types of interactions, a series of powerful
codes will emerge that will be helpful throughout the research. For example, in her research on
adolescent girls in the southern United States, Mary Thomas (2004) found that young (14-year-
old) African-American girls’ interactions with peers were strongly implicated in the type and
level of their sexual activity. Thus, Thomas might have coded her interview transcripts regarding
peer factors by whether, how, and where girls engaged in sexual activity, as well as whom they
were influenced by or interacted with.
“Strategies and tactics”2 is a little more complicated than Strauss’s first two types of
themes because it requires a deeper understanding of the things (events, actions, statements) you
observe and how they relate to broader phenomena and it suggests a certain level of purposeful
intent among the research subjects that may demand additional inquiry on your part. For
example, feminist geographers are often interested in women’s survival or “livelihood” strategies
in different areas of the world (see, for example, Oberhauser and Yeboah 2011; Jones and
Murphy 2011). Noting that women in certain economic contexts tend to use particular types of
financial survival tactics (for example, growing food products for sale in a local market) can
begin to illuminate broader economic, social, and political processes that shape women’s options
and actions, which is a valuable insight for geographic research. Other types of strategies or
tactics might involve education and career decisions, political activism, housing choices, family
negotiations, or even subversion.
Coding for strategies and tactics can be straightforward (and descriptive) in instances
when respondents say something like “I moved in with my mother so that she could care for my
baby while I finished job training” or “I got involved with a local group of residents to raise
awareness of environmental contamination in our neighbourhood because I was concerned about
property values.” Note the words so and because in these statements, which are good tip-offs that
a strategy or tactic is embedded in the text.
Other times, coding for strategies and tactics may be more subtle—and more analytical—
as when respondents do not explicitly state their reasons for certain actions but a connection
emerges through observation, review of interview text, or other data. For instance, many
geographers (for example, Blumen 2002; Lee, Kim and Wainwright 2010; Lees 2012; Secor
2004) have paid attention to ways that people engage in resistance against diverse forms of
oppression, which may be seen as strategies for empowerment, rights, or merely survival. For
example, Orna Blumen (2002, 133) took “dissatisfaction articulated in subtle terms” by ultra-
orthodox Jewish women as small but significant indicators of the women’s resistance to their
families’ economic circumstances and, more broadly, to the status and roles of women in that
community. For the women in Blumen’s study, then, referring to fatigue, hoping their husbands
would soon find paying work, and “minor, personal, nonconformist remarks suggestive of
ambivalence” (2002, 140) could all be coded as tactics of resistance, in part because Blumen—
through careful qualitative work—had sufficiently analyzed the broader context of the women’s
lives and goals to recognize them as such.
Similar to the above, “consequences” is a slightly more complicated code. On the surface,
there are descriptive indicators for consequences, including terms such as “then,” “because,” “as
a result of,” and “due to” that may be used in subjects’ statements and can be good clues to
consequences and as a first run could certainly be used in this way. Again, however, there are
also more analytically sophisticated ways of discovering and coding consequences that are
dependent on the unique empirical settings and events of each study. Some consequences will be
matters of time passing and actions taking place that result in a particular outcome—the passage
of a law, a change in rules or practices, and so on. However, other consequences are more subtle
and personal, or they are not the result of changes over time and therefore may be trickier to
identify and code as such. For example, when Anna Secor (2004) heard from young Kurdish
women living in Istanbul that they felt uncomfortable in some areas of the city, she might have
coded her focus group transcripts for the consequences of feeling out of place due to the
women’s identity as an oppressed minority in Turkey. Coding for “consequences” of this kind
requires sensitivity to both the subjects and their community context but is potentially a rich
source of analysis and insight if done with care.
As an example of what a sample of coded material looks like with both descriptive and
analytic codes, Box 18.1 demonstrates a small selection of field notes from my project on
children’s urban geographies along with codes, themes, and notes. Even this fairly short piece of
text reveals several relationships (friends, cousins, bullies), tactics (ways of getting attention),
and conditions (disorganization, noise level) that stimulated further examination in other project
analysis. Additionally, several analytic themes or questions are seen emerging here: the possible
relationship between gender and violent play, some children’s pride in perceived community
membership (despite living in a disinvested physical environment), and the importance of play in
defining what a neighbourhood is among the children. Subsequent to the quilt project represented
here and in combination with other Children’s Urban Geographies Project data, I generated a
theory of how children define and ascribe meaning to the idea of “neighbourhood” (Cope 2008);
theory-building, after all, is an important goal of most qualitative research.

Developing the Coding Structure

Using the four types of themes reviewed here will take you a long way towards constructing a
codebook, and you may find other types of themes that are helpful to you, such as “meanings,”
“processes,” or “definitions.” Using the combination of descriptive and analytic codes, you may
well have more than 100 codes at some point, which is unwieldy at best and counterproductive at
worst. Lists of codes that have not been categorized, grouped, and connected will be hard to
remember, have too much overlap and/or leave uneven gaps, and will not enable productive
analysis. Therefore, the next step is to develop a coding structure whereby codes themselves are
grouped together according to their similarities, substantive relationships, and conceptual links.
This process requires some amount of work but is well worth the effort, both for ease of coding
your material and for discerning significant results from your findings.
Developing the coding structure can proceed in various ways, and there are many
resources available that demonstrate different approaches (see Denzin and Lincoln 2011; Miles
and Huberman 1994; Silverman 2001), but the main purpose is to organize the codes—and
therefore the data and the analysis process. Some codes will automatically cluster; for example,
codes relating to the setting of interviews (for example, home, office, public space, clinic),
characteristics of subjects (for example, age, gender), or other categories (for example,
occupations, leisure activities, life events). Other codes seem to fit together because of their
common issues; for example, you might have a group of codes related to people’s goals or
intentions or a group of codes related to people’s experiences of oppression. Finally, codes based
on the substantive content of text or actions—and most likely related to the analytic themes you
are developing—will create another cluster of codes, for example: perceptions, meanings, places,
identities, memories, difference, representations, and associations.
Once the codebook is relatively comfortable (I hesitate to say “complete”) and the coding
structure is devised, you will want to go through much of your data again to capture connections
that may have been missed the first (or second or third) time around. Remember that coding is an
iterative process that feeds back on itself—only you can decide when it is time to move on. As
Miles and Huberman (1994) point out, it is sometimes simpler when time or money pressures put
some finality on projects that otherwise could always benefit from “one more case study” or
endless additional tweaking of the coding structure.

Coding with Others

Depending on the size and resources of the research project, there may be a case for using
multiple coders for the data, which can add considerable complexity to the process. There is an
inherent tension in using multiple coders on a project: is the goal to make everyone code as
consistently as possible, or is the goal to allow each coder to interpret data in her or his way
within the bounds of the coding structure in order to capture many diverse meanings? The
answer will depend on the project and the epistemological leanings of the lead researcher(s), but
in fact both of these goals are important. In the first instance, reliability of the data is
undoubtedly enhanced when several coders independently code a piece of data the same way—a
common interpretation of data means that there is agreement on its meaning. For the sake of time
and data reduction, having multiple coders can certainly be helpful, but only if they are truly
consistent in their coding, which is rare but could be accomplished by achieving conformity on
the meanings of codes and providing thorough definitions for each code. On the other hand, text
and video—as social data sources—are inherently subject to multiple interpretations and
understandings, all of which may be correct or “true.” While there may be some interpretations
that are farfetched or extreme, in general we as social researchers will be interested in capturing
diverse understandings, and having multiple coders can be a great benefit for the project to make
deeper and broader connections from the data.

Computer-Aided Qualitative Data Analysis Software (CAQDAS) and

Qualitative GIS

Anyone who has coded research material “by hand” can understand the attraction of computer
assistance in this endeavour—the idea that one’s codes could be organized and employed using
software holds many potential benefits in terms of time-savings, consistency, and data security,
particularly when compared to stacks of notecards. However, there are also some cautions that
should be kept in mind before embarking on the investment of time and resources to use
CAQDAS. This brief overview is not intended as instructional or comprehensive, but rather,
identifies some key considerations and points the reader towards further resources.
There are numerous CAQDAS packages on the market today, ranging in price,
sophistication, functions, and utility. However, the basic premise of all of them is the same: to
bring coding and analysis into a computerized operation. In fact, some of the most basic
functions of coding can be completed in word processors or using spreadsheets, and for smaller
or fairly basic analyses, these might be preferable. For example, a first-run content analysis is
easily accomplished using the “find” function in word processors, and pieces of text can be
highlighted and commented upon using review functions such as “track changes” in MS-Word.
Lists of codes and memos, as well as notes about when to use which codes, are easily stored and
managed in spreadsheet programs or “table” options in word processors. For more complex
projects with more codes, more data, and/or multiple researchers investing in learning how to use
a CAQDAS (and, sometimes, investing in purchasing a licence, which many universities already
have available) can be well worth it.
I highly recommend reading the review and self-critique Bettina van Hoven (2010) wrote
about her experiences learning and using CAQDAS in her dissertation project. She reveals several
“advantages” of using CAQDAS, including managing large quantities of data, convenient coding
and retrieving, and quick identification of deviant cases (advantages) as well as “concerns” that
include obsessions with volume of data, mechanistic data analysis and a “taken-for-granted”
mode of data handling, alienation from one’s data, and an over-emphasis on grounded theory
(van Hoven 2010, 462). No matter how much packages change, these advantages and concerns
are likely to be timeless.
Some examples of this type of software being used by geographers include NVivo (a
cousin of QSR N4), Atlas.ti, MaxQDA, and HyperRESEARCH. All of these have demo versions and
online manuals in PDF on their respective websites, and most have lower-cost student versions;
therefore one can easily “test-drive” a package before committing fully to it. Because packages
are constantly being updated and changing in their functionality, an assessment of the relative
strengths and merits of these would be quickly outdated, and therefore they will not be reviewed
here per se; I recommend searching the web and online forums for reviews of the latest versions
of CAQDAS to see what might suit a given project, the capacities of the researchers, and the
available time and budget. However, there are some features that are worth mentioning here that
are relevant to consider, regardless of what new offerings emerge.
First, CAQDAS has three primary functions: text retrieval (especially helpful for the
“constant comparison” approach to analysis), coding, and theory-building. It is, of course,
essential to realize that the program does not do the coding for you, except in the most
rudimentary way based on a set of instructions you provide; the researcher still must engage in
thoughtful and thorough building of codes and applying them to text segments, photo elements,
or video footage clips. However, the CAQDAS is an excellent organizational tool for storage,
retrieval, and interpretation of data, and all packages have at least some “memoing” capacity.
Targeted searches or browsing by coded material can reveal connections that would have been
difficult to see with paper index cards, and the theory-building functions (such as hierarchies or
concept maps that show relations between data, or even between codes) are useful for analysis
and the generation of original interpretations and conceptual understanding.
The second feature to consider is the degree to which a CAQDAS can be geographically
linked (if this is something that is relevant for your research). These capabilities are pushing
forward the practice of qualitative GIS (Cope and Elwood, 2009) by increasingly developing
the integrated analysis of qualitative and geographically referenced data (Gilbert, Jackson, and di
Gregorio, 2013; Jung, 2014; Verd Pericas and Porcel, 2012). For example, Fielding and
Cisneros-Puebla (2009) review the capacities of both Atlas.ti and MaxQDA for integrating
geographical maps (that is, not just concept maps!) into analysis. Atlas.ti, for instance, allows the
importing of KML files and Google maps can be opened from within the program; MaxQDA
allows hyperlinks to Google Earth images. In what Fielding and Cisneros-Puebla identify as a
“technological convergence” these developments should be promising for many geography
scholars and students. Further, with the rapid growth of online mapping (both vernacular and
professional), volunteered geographic information, crowdsourcing, and spatially referenced
social media, the possibilities for qualitative research of, with, and across “big data” are growing
as fast as the Internet itself (see, for example, Crampton, Graham, Poorthuis, Shelton, Stephens,
Wilson, and Zook 2013). These raise the important question of how the massive saturation of
spatially referenced digital data will affect qualitative research in geography and beyond.
New work is emerging that pushes the boundaries of coding and leads to creative
analyses and representations of qualitative geographic data. With the rise of participatory
research in geography (Kesby, Kindon, and Pain 2005; see also Chapter 17 in this volume),
qualitative GIS (Cope and Elwood 2009; Kwan and Knigge 2006, Preston and Wilson, 2014),
participatory mapping, mixed methods, and other integrative practices, we need to stay attuned to
how coding can keep up with new research processes and technologies. One example of this is
Jin-Kyu Jung’s (2009) experiments with using codes as a bridge between qualitative analysis
using CAQDAS and spatial data analysis using GIS. In his work, the code literally serves as a
software-level link between databases, allowing analysis programs to “speak” to each other in a
platform he calls CAQ-GIS (computer-aided qualitative geographical information systems).
At a conceptual level in Jung’s work, the code also serves as an analytical connection between
social contextual data and spatially referenced data, allowing researchers to develop new
understandings of social–spatial relations. Further, these innovations have been increasingly
taken up outside of geography, for example in urban sociology and in public health research,
where the potential for integrating spatially referenced numerical data with qualitative data is
especially relevant. Along these lines, Fielding and Cisneros-Puebla (2009, 349) argue that
“there is an emergent convergence of methodologies and analytical purposes between qualitative
geography and qualitative social science” that represents an exciting new realm of inquiry. With
the many possibilities for mixed-methods and qualitative research facilitated by better and faster
digital technologies, the necessity for researchers to be well-trained, rigorous, reflexive, and
ethical has never been greater.

Being in the World, Coding the World

By way of conclusion, let me point out that interpreting and analyzing the social world is not a
mysterious process that must be learned from scratch but rather is one that we are all already
actively practising in our everyday lives. The recognition that we are all constantly “coding” and
making sense of the world around us may be a helpful realization for getting started in a research
project and can also assist us in critiquing our own practices of data reduction, organization, and
analysis. As Silverman (1991, 293) points out, there are many ways of “seeing” and interpreting
the world, and—as social beings—we never really shut those lenses off, so why not embrace
diverse perspectives and turn our gaze to the process of interpretation?

<CATCH block quote>

How we code or transcribe our data is a crucial matter for qualitative researchers. Often,
however, such researchers simply replicate the positivist model routinely used in
quantitative research. According to this model, coders of data are usually trained in
procedures with the aim of ensuring a uniform approach. . . . However,
ethnomethodology reminds us that “coding” is not the [sole] preserve of research
 scientists. In some sense, researchers, like all of us, “code” what they hear and see in the
world around them [all the time]. . . . The ethnomethodological response is to make this
everyday “coding” (or “interpretive practice”) the object of inquiry. [Silverman 1991,
293]

<end bq>

Being in the world requires us to categorize, sort, prioritize, and interpret social data in all of our
interactions. Analyzing qualitative data is merely a formalization of this process in order to apply
it to research and to provide some structure as a way of conveying our interpretations to others.

Key Terms

abstracting
analytic code
CAQDAS (computer-aided qualitative data analysis software)
CAQ-GIS (computer-aided qualitative geographical information systems)
codebook
coding
coding structure
concept mapping
content analysis
data reduction
descriptive code
ethnography
grounded theory
initial codes
interpretive codes
in vivo code
memo
qualitative GIS
reflexivity
theme

Review Questions
1. What is the difference between descriptive and analytic codes, how do they relate to one
another, and what are their respective uses in coding qualitative data?
2. Why does the author state that coding is analysis?
3. What are some potential benefits and potential problems with using a CAQDAS in a project?
4. In what ways do we “code” events, processes, and other phenomena in everyday life? How
might thinking about these ways help us to become better qualitative researchers?

Review Exercises

1. Memoing: Have the class write a short, first-person account of some type of experience (“my
journey from home to the classroom today” works well!) and ask each person to go back over
their own account making memos about why certain events, conditions, interactions, or
thoughts were important.
2. Concept mapping: Collect all the first-person accounts from Exercise 1 and print them
single-sided, then lay them out on a large table or tack them to a wall. Have students review
each account and identify key statements, copying each key statement onto a note card
(remember to mark which original document it came from). Then use the “sort and pile”
method to identify common statements, build themes, and identify connections between
concepts. The notecards can then be arranged in clouds, networks, hierarchies, or other visual
patterns to further analyze relationships.
3. Practise coding: Take a newspaper story or a short video clip (5–10 minutes) and work singly
or in small groups to identify some basic in vivo (present in the original text/video) and
descriptive codes that would suit an initial coding of that material. Each person writes her/his
codes on large note cards or sticky notes. Collect all the codes from class members, tack
them to the wall, and collectively construct a coding structure by eliminating duplicates,
grouping similar codes together, and identifying outliers or gaps.

Useful Resources

For an excellent step-by-step guide to coding, see The Coding Manual for Qualitative
Researchers by Johnny Saldaña (2012). Additionally, there are several examples of
coding and “making sense” of data by geographers, including collections by Kitchin and
Tate (2000); Clifford, French, and Valentine (2010); Flowerdew and Martin (2005);
Limb and Dwyer (2001); and Moss (2002).
 Notes

1. While it is always difficult to convey examples of coding without recounting the entire scope
of the research, it is hoped that these examples from a real research project are sufficiently
illustrative to demonstrate different coding approaches.
2. Despite the similar pairing of these two words, I am not referring here to Michel de Certeau’s
(1984) notion of “strategy” (a technique of spatial organization employed by “the powerful”) and
“tactic” (an everyday means of “making do,” typically used by those with few options), although
there are certainly potential connections. Rather, I am using the terms in their most literal sense
as they are employed in Strauss and Corbin (1990) to convey ideas about how people
conceptualize what they want and what they do to try to arrive at those goals.

View publication stats