10.1177/1525822X02239569 ARTICLEFIELD METHODSRyan, Bernard / TECHNIQUES TO IDENTIFY THEMES
Techniques to Identify Themes
GERY W. RYAN
H. RUSSELL BERNARD
University of Florida
Theme identification is one of the most fundamental tasks in qualitative research. It
also is one of the most mysterious. Explicit descriptions of theme discovery are rarely
found in articles and reports, and when they are, they are often relegated to appendi
ces or footnotes. Techniques are shared among small groups of social scientists, but
sharing is impeded by disciplinary or epistemological boundaries. The techniques
described here are drawn from across epistemological and disciplinary boundaries.
They include both observational and manipulative techniques and range from quick
word counts to laborious, in-depth, line-by-line scrutiny. Techniques are compared
on six dimensions: (1) appropriateness for data types, (2) required labor, (3)
required expertise, (4) stage of analysis, (5) number and types of themes to be gener-
ated, and (6) issues of reliability and validity.
Keywords: theme identification; qualitative analysis; text analysis; open coding;
qualitative research methods
Analyzing text involves several tasks: (1) discovering themes and
subthemes, (2) winnowing themes to a manageable few (i.e., deciding which
themes are important in any project), (3) building hierarchies of themes or
code books, and (4) linking themes into theoretical models.
We focus here on the first task: discovering themes and subthemes in
texts—and in other qualitative data, like images or artifacts, for that matter.
We outline a dozen techniques, drawn from across the social sciences and
from different theoretical perspectives. The techniques range from simple
word counts that can be done by a computer to labor-intensive, line-by-line
analyses that, so far, only humans can do.
Each technique has advantages and disadvantages. Some methods are
more suited to rich, complex narratives, while others are more appropriate for
short responses to open-ended questions. Some require more labor and
expertise on behalf of the investigator, others less.
Making explicit the techniques we use for discovering themes in qualita
tive data is important for three reasons. First, discovering themes is the basis
Field Methods, Vol. 15, No. 1, February 2003 85–109
© 2003 Sage Publications
of much social science research. Without thematic categories, investigators
have nothing to describe, nothing to compare, and nothing to explain. If
researchers fail to identify important categories during the exploratory phase
of their research, what is to be said of later descriptive and confirmatory
Second, being explicit about how we establish themes allows consumers
of qualitative research (including those who fund it) to assess our method
Third, qualitative researchers need an explicit and jargon-free vocabulary
to communicate with each other across disciplines and across epistemo
logical positions. As we see it, theme discovery is practiced by avowed
positivists and interpretivists alike. In fact, some of the techniques we
describe are drawn from the interpretivist tradition, while others reflect the
efforts of positivists who analyze qualitative data. We see nothing wrong
with this. All the techniques we describe can help researchers see their data in
a new light. Each has its advantages and disadvantages.
We rarely see descriptions (even in footnotes or appendices) of how
researchers came to discover the themes they report in their articles. The
techniques we use for finding themes are, of course, shared within invisible
colleges, but wider sharing is impeded by disciplinary or epistemological
boundaries. “Many researchers,” said Renata Tesch (1990:115), “read only
certain authors and remain quite ignorant of analysis purposes and proce-
dures different from the ones their favorite methodological writers describe.”
More than a decade later, little appears to have changed.
WHAT IS A THEME?
This problem has a long history. Seventy years ago, Thompson ([1932-
1936] 1993) created an index of folktale motifs that filled six volumes.
Anthropologist Morris Opler (1945) saw the identification of themes as a key
step in analyzing cultures. “In every culture,” he said,
are found a limited number of dynamic affirmations, called themes, which con
trol behavior or stimulate activity. The activities, prohibitions of activities, or
references which result from the acceptance of a theme are its expressions. . . .
The expressions of a theme, of course, aid us in discovering it. (pp. 198-99)
Opler (1945) established three principles for thematic analysis. First, he
observed that themes are only visible (and thus discoverable) through the
manifestation of expressions in data. And conversely, expressions are mean
ingless without some reference to themes.
86 FIELD METHODS
Second, Opler (1945) noted that some expressions of a theme are obvious
and culturally agreed on, while others are subtler, symbolic, and even
Third, Opler (1945) observed that cultural systems comprise sets of inter
related themes. The importance of any theme, he said, is related to (1) how
often it appears, (2) how pervasive it is across different types of cultural ideas
and practices, (3) how people react when the theme is violated, and (4) the
degree to which the number, force, and variety of a theme’s expression is
controlled by specific contexts.
Today, social scientists still talk about the linkage between themes and
their expressions but use different terms to do so. Grounded theorists talk
about “categories” (Glaser and Strauss 1967), “codes” (Miles and Huberman
1994), or “labels” (Dey 1993:96). Opler’s (1945) “expressions” are called
“incidents” (Glaser and Strauss 1967), “segments” (Tesch 1990), “thematic
units” (Krippendorf 1980), “data-bits” (Dey 1993), and “chunks” (Miles and
Huberman 1994). Lincoln and Guba (1985) referred to expressions as “units”
(p. 345). Strauss and Corbin (1990) called them “concepts.”
For Strauss and Corbin (1990), the links between expressions and themes
are “conceptual labels placed on discrete happenings, events, and other
instances of phenomena.” Themes, or categories, are the classification of
more discrete concepts. “This classification is discovered when concepts are
compared one against another and appear to pertain to a similar phenomenon.
Thus, the concepts are grouped together under a higher order, more abstract
concept called a category” (p. 61).
Here, we follow Agar’s (1979, 1980) lead and remain faithful to Opler’s
(1945) terminology. To us, the terms “theme” and “expression” more natu-
rally connote the fundamental concepts we are tying to describe. In everyday
language, we talk about themes that appear in texts, paintings, and movies
and refer to particular instances as expressions of anger and evil. In selecting
one set of terms over others, we surely ignore subtle differences, but the basic
ideas are just as useful under many glosses.
HOW DO YOU KNOW A THEME WHEN YOU SEE ONE?
To us, themes are abstract (and often fuzzy) constructs that link not only
expressions found in texts but also expressions found in images, sounds, and
objects. You know you have found a theme when you can answer the ques
tion, What is this expression an example of? Themes come in all shapes and
sizes. Some themes are broad and sweeping constructs that link many differ
ent kinds of expressions. Other themes are more focused and link very spe
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 87
cific kinds of expressions. When we describe themes as the conceptual link
ing of expressions, it is clear that there are many ways in which expressions
can be linked to abstract constructs.
WHERE DO THEMES COME FROM?
Themes come both from the data (an inductive approach) and from the
investigator’s prior theoretical understanding of the phenomenon under
study (an a priori approach). A priori themes come from the characteristics of
the phenomenon being studied; from already agreed on professional defini
tions found in literature reviews; from local, commonsense constructs; and
from researchers’ values, theoretical orientations, and personal experiences
(Bulmer 1979; Strauss 1987; Maxwell 1996). Strauss and Corbin (1990:41–
47) called this theoretical sensitivity. Investigators’ decisions about what
topics to cover and how best to query informants about those topics are a rich
source of a priori themes (Dey 1993:98). In fact, the first pass at generating
themes often comes from the questions in an interview protocol (Coffey and
Atkinson 1996:34). Unlike pure literature reviews, these themes are partly
Mostly, though, themes are induced from empirical data—from texts,
images, and sounds. Even with a fixed set of open-ended questions, one can-
not anticipate all the themes that arise before analyzing the data (Dey
1993:97–98). The act of discovering themes is what grounded theorists call
open coding and what classic content analysts call qualitative analysis
(Berelson 1952) or latent coding (Shapiro and Markoff 1997).
There are many variations on these methods, and individual researchers
have different recipes for arriving at the preliminary set of themes (Tesch
1990:91). We next describe eight observational techniques—things to look
for in texts—and four manipulative techniques—ways of processing texts.
These twelve techniques are not exhaustive and are often combined in
SCRUTINY TECHNIQUES—THINGS TO LOOK FOR
Looking for themes in written material typically involves pawing through
texts and marking them up with different colored pens. Sandelowski
(1995:373) observed that analysis of texts begins with proofreading the
material and simply underlining key phrases “because they make some as yet
inchoate sense.” For those who tape their interviews, the process of identify
88 FIELD METHODS
ing themes probably begins with the act of transcribing the tapes. Bogdan and
Biklen (1982:165) suggested reading over the text at least twice. Whether the
data come in the format of video, audio, or written documents, handling them
is always helpful for finding themes. Here is what researchers look for.
Repetition is one of the easiest ways to identify themes. Some of the most
obvious themes in a corpus of data are those “topics that occur and reoccur”
(Bogdan and Taylor 1975:83) or are “recurring regularities” (Guba 1978:53).
“Anyone who has listened to long stretches of talk,” said D’Andrade (1991),
“knows how frequently people circle through the same network of ideas” (p.
287). Claudia Strauss (1992), for example, did several in-depth interviews
with Tony, a retired blue-collar worker in Connecticut, and found that Tony
repeatedly referred to ideas associated with greed, money, businessmen, sib
lings, and “being different.” Strauss concluded that these ideas were impor-
tant themes in Tony’s life. She displayed the relationships among these ideas
by writing the concepts on a piece of paper and connecting them with lines to
Tony’s verbatim expressions, much as researchers today do with text analy-
sis software. The more the same concept occurs in a text, the more likely it is a
theme. How many repetitions are enough to constitute an important theme,
however, is an open question and one only the investigator can decide.
Indigenous Typologies or Categories
Another way to find themes is to look for local terms that may sound unfa-
miliar or are used in unfamiliar ways. Patton (1990:306, 393–400) referred to
these as “indigenous categories” and contrasted them with “analyst-
constructed typologies.” Grounded theorists refer to the process of identify-
ing local terms as in vivo coding (Strauss 1987:28; Strauss and Corbin
1990:61–74). Ethnographers call this the search for typologies or classifica
tion schemes (Bogdan and Taylor 1975:83) or cultural domains (Spradley
Spradley (1972) recorded conversations among tramps at informal gath
erings, meals, and card games. As the men talked to each other about their
experiences, they made many references to making a flop. Spradley searched
through his recorded material and notes looking for verbatim statements
made by informants about this topic. He found that he could categorize most
statements into subthemes such as kinds of flops, ways to make flops, ways to
make your own flop, kinds of people who bother you when you flop, ways to
make a bed, and kinds of beds. Spradley then returned to his informants and
sought additional information from them on each of the subthemes. For other
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 89
examples of coding for indigenous categories, see Becker’s (1993) descrip
tion of medical students’ use of the word “crock” and Agar’s (1973) descrip
tion of drug addicts’ understandings of what it means to shoot up.
Metaphors and Analogies
In pioneering work, Lakoff and Johnson (1980) observed that people
often represent their thoughts, behaviors, and experiences with analogies and
metaphors. Analysis, then, becomes the search for metaphors in rhetoric and
deducing the schemas or underlying themes that might produce those meta
phors (D’Andrade 1995; Strauss and Quinn 1997).
Naomi Quinn (1996) analyzed hundreds of hours of interviews to dis
cover fundamental themes underlying American marriages and to under
stand how these themes are tied together. She found that people talk about
their surprise at the breakup of a marriage by saying that they thought the cou-
ple’s marriage was “like the Rock of Gibraltar” or that they thought the mar-
riage had been “nailed in cement.” People use these metaphors because they
assume that their listeners know that cement and the Rock of Gibraltar are
things that last forever.
Quinn (1996) reported that the hundreds of metaphors in her corpus of
texts fit into eight linked classes that she labeled lastingness, sharedness,
compatibility, mutual benefit, difficulty, effort, success (or failure), and risk
of failure. For example, when informants said of someone’s marriage that “it
was put together pretty good” or was a “lifetime proposition,” Quinn saw
these metaphors as exemplars of the expectation of lastingness in marriage.
Other examples of the search for cultural schemas in texts include Hol-
land’s (1985) study of the reasoning that Americans apply to interpersonal
problems, Kempton’s (1987) study of ordinary Americans’ theories of home
heat control, and Strauss’s (1997) study of what chemical plant workers and
their neighbors think about the free-enterprise system.
Naturally occurring shifts in content may be markers of themes. In written
texts, new paragraphs may indicate shifts in topics. In speech, pauses,
changes in voice tone, or the presence of particular phrases may indicate tran
sitions. Agar (1983) examined transcripts of arguments presented by inde
pendent truckers at public hearings of the Interstate Commerce Commission.
He noticed that each speech was divided into topical sections that were often
demarcated by metaphors. In semistructured interviews, investigators steer
the conversation from one topic to another, creating transitions, while in two-
party and multiparty natural speech, transitions occur continually. Analysts
90 FIELD METHODS
of conversation and discourse examine features such as turn taking and
speaker interruptions to identify these transitions. (For an overview, see
Similarities and Differences
What Glaser and Strauss (1967:101–16) called the “constant comparison
method” involves searching for similarities and differences by making sys
tematic comparisons across units of data. Typically, grounded theorists
begin with a line-by-line analysis, asking, What is this sentence about? and
How is it similar to or different from the preceding or following statements?
This keeps the researcher focused on the data rather than on theoretical
flights of fancy (Glaser 1978:56–72; Charmaz 1990, 2000; Strauss and
Another comparative method involves taking pairs of expressions—from
the same informant or from different informants—and asking, How is one
expression different from or similar to the other? The abstract similarities and
differences that this question generates are themes. If a particular theme is
present in both expressions, then the next question to ask is, Is there any dif-
ference, in degree or kind, in which the theme is articulated in both of the
expressions? Degrees of strength in themes may lead to the naming of
subthemes. Suppose an investigator compares two video clips and finds that
both express the theme of anxiety. On careful scrutiny, the researcher notices
that the two instances of anxiety are both weak, but one is expressed verbally
and the other through subtle hand gestures. The investigator codes these as
two new subthemes.
Researchers also compare pairs of whole texts, asking, How is this text
different from the preceding text? and What kinds of things are mentioned in
both? They ask hypothetical questions such as, What if the informant who
produced this text had been a woman instead of a man? and How similar is
this text to my own experience? Bogdan and Biklen (1982:153) recom
mended reading through passages of text and asking, “What does this remind
me of?” Just as a good journalist would do, investigators compare answers to
questions across people, space, and time. (For more formal techniques of
identifying similarities and differences among segments of text, see the dis
cussion below on cutting and sorting.)
Another approach is to look carefully for words and phrases such as “be
cause,” “since,” and “as a result,” which often indicate causal relations. Words
and phrases such as “if” or “then,” “rather than,” and “instead of” often sig
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 91
nify conditional relations. The phrase “is a” is often associated with taxo
nomic categories, as in “a lion is a kind of cat.” Time-oriented relationships
are expressed with words such as “before,” “after,” “then,” and “next.”
Typically, negative characteristics occur less often than do positive ones.
Simply searching for the words “not,” “no,” “none,” or the prefix “non-” (and
its allomorphs, “un-,” “in-,” “il-,” “im-,” etc.) may be a quick way to identify
some themes. Investigators can discover themes by searching for such
groups of words and looking to see what kinds of things the words connect.
What other kinds of relationships might be of interest? Casagrande and
Hale (1967) suggested looking for attributes (e.g., X is Y), contingencies
(e.g., if X, then Y), functions (e.g., X is a means of affecting Y), spatial orien
tations (e.g., X is close to Y), operational definitions (e.g., X is a tool for
doing Y), examples (e.g., X is an instance of Y), comparisons (e.g., X resem
bles Y), class inclusions (X is a member of class Y), synonyms (e.g., X is
equivalent to Y), antonyms (e.g., X is the negation of Y), provenience (e.g., X
is the source of Y), and circularity (e.g., X is defined as X). (For lists of other
kinds of relationships that may be useful for identifying themes, see Lindsay
and Norman 1972; Burton and Kirk 1980:271; and Werner and Schoepfle
Metaphors, transitions, and connectors are all part of a native speaker’s
ability to grasp meaning in a text. By making these features more explicit, we
sharpen our ability to find themes.
The next scrutiny-based approach works in reverse from typical theme-
identification techniques. Instead of asking, What is here? we can ask, What
is missing? Researchers have long recognized that much can be learned from
qualitative data by what is not mentioned. Bogdan and Taylor (1975) sug
gested being “alert to topics that your subjects either intentionally or uninten
tionally avoid” (p. 82).
For instance, women who have strong religious convictions may fail to
mention abortion during discussions of birth control. In power-laden inter
views, silence may be tied to implicit or explicit domination (Gal 1991). In a
study of birth planning in China, Greenhalgh (1994) reported that she could
not ask direct questions about resistance to government policy but that
respondents “made strategic use of silence to protest aspects of the policy
they did not like” (p. 9). Obviously, themes that are discovered in this manner
need to be carefully scrutinized to ensure that investigators are not finding
only what they are looking for.
92 FIELD METHODS
In fact, lacunae in texts may indicate primal cultural assumptions.
Spradley (1979:57–58) observed that when people tell stories, they leave out
information that “everyone knows.” He called this process abbreviating. The
statement “John was broke because it was the end of the month” requires a
great deal of cultural understanding. It requires knowing that there is abso
lutely no causal relationship between financial solvency and dates, that peo
ple are often paid at the end of the month, and that people sometimes spend all
their money before getting their next paycheck. Price (1987) suggested look
ing for missing information by translating people’s narratives into the
worldview of a different audience. When she finds herself filling in the gaps,
she knows she has found fundamental themes.
Searching for missing information is not easy. People may not trust the
interviewer, may not wish to speak when others are present, or may not
understand the investigator’s questions. Distinguishing between when infor-
mants are unwilling to discuss a topic and when they assume the investigator
already knows about the topic requires a lot of familiarity with the subject
A variant on the missing data technique is to scrutinize any expressions
that are not already associated with a theme (Ryan 1999). This means reading
a text over and over. On the first reading, salient themes are clearly visible
and can be quickly and readily marked with highlighters. In the next stage,
the researcher searches for themes in the data that remain unmarked. This tac-
tic—marking obvious themes early and quickly—forces the search for new
and less obvious themes in the second pass.
In addition to identifying indigenous themes—themes that characterize
the experience of informants—researchers are interested in understanding
how qualitative data illuminate questions of importance to social science.
Spradley (1979:199–201) suggested searching interviews for evidence of
social conflict, cultural contradictions, informal methods of social control,
things that people do in managing impersonal social relationships, methods
by which people acquire and maintain achieved and ascribed status, and
information about how people solve problems. Bogdan and Biklen
(1982:156–62) suggested examining the setting and context, the perspec
tives of the informants, and informants’ ways of thinking about people,
objects, processes, activities, events, and relationships. Strauss and Corbin
(1990:158–75) urged investigators to be more sensitive to conditions,
actions/interactions, and consequences of a phenomenon and to order these
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 93
conditions and consequences into theories. “Moving across substantive
areas,” said Charmaz (1990), “fosters developing conceptual power, depth,
and comprehensiveness” (p. 1163).
There is a trade-off, of course, between bringing a lot of prior theorizing to
the theme-identification effort and going at it fresh. Prior theorizing, as
Charmaz (1990) said, can inhibit the forming of fresh ideas and the making of
surprising connections. And by examining the data from a more theoretical
perspective, researchers must be careful not to find only what they are look
ing for. Assiduous theory avoidance, on the other hand, brings the risk of not
making the connection between data and important research questions.
The eight techniques described above can all be used with pencil and
paper. Once you have a feel for the themes and the relations among them, we
see no reason to struggle bravely on without a computer. Of course, a com
puter is required from the onset if the project involves hundreds of inter-
views, or if it is part of a multisite, multi-investigator effort. Even then, there
is no substitute for following hunches and intuitions in looking for themes to
code in texts (Dey 1993).
Next, we describe four techniques that require more physical or computer-
based manipulation of the text itself.
Some techniques are informal—spreading texts out on the floor, tacking
bunches of them to a bulletin board, and sorting them into different file fold-
ers—while others require special software to count words or display word-
Cutting and Sorting
After the initial pawing and marking of text, cutting and sorting involves
identifying quotes or expressions that seem somehow important and then
arranging the quotes/expressions into piles of things that go together. Lincoln
and Guba (1985:347–51) offered a detailed description of the cutting and
sorting technique. Their method of constant comparison is much like the
pile-sorting task used extensively in cognitive research (e.g., Weller and
There are many variations on this technique. We cut out each quote (mak
ing sure to maintain some of the context in which it occurred) and paste the
material on a small index card. On the back of each card, we write down the
quote’s reference—who said it and where it appeared in the text. Then we lay
94 FIELD METHODS
out the quotes randomly on a big table and sort them into piles of similar
quotes. Then we name each pile. These are the themes.
Clearly, there are many ways to sort the piles. Splitters, who maximize the
differences between passages, are likely to generate more fine-grained
themes. Lumpers, who minimize the differences, are likely to identify more
overarching or metathemes. As the first exploratory step in the data analysis,
investigators are most concerned with identifying as wide a range of themes
as possible. In later steps, they will need to address the issue of which themes
are the most important and worthy of further analysis.
In another variation, the principal investigator on a large project might ask
several team members to sort the quotes into named piles independently.
This is likely to generate a longer list of possible themes than would be pro
duced by a group discussion. And if the people sorting the quotes are unaware
of whom the quotes came from, this is an unbiased way of comparing themes
across different groups.
In really large projects, investigators might have pairs of team members
sort the quotes together and decide on the names for the piles. Ryan (1995)
has found it particularly helpful to audiotape the conversations that occur
when pairs of people perform pile-sorting tasks. The conversations often pro-
vide important insights into the underlying criteria and themes people use to
Barkin, Ryan, and Gelberg (1999) provided yet another variation. They
interviewed clinicians, community leaders, and parents about what physi-
cians could do and did to prevent violence among youth. These were long,
complex interviews, so Barkin, Ryan, and Gelberg broke the coding process
into two steps. They started with three major themes that they developed
from theory. The principal investigator went through the transcripts and cut
out all the quotes that pertained to each of the major themes. Then, four other
coders independently sorted the quotes from each major theme into piles.
For each major theme, Barkin, Ryan, and Gelberg (1999) converted the
pile sort data into a quote-by-quote similarity matrix. The numbers in the
cells, which ranged from 0 to 4, indicated the number of coders who had
placed the quotes in the same pile. The researchers analyzed each matrix with
multidimensional scaling (MDS) and cluster analysis. The MDS displayed
the quotes in a map, where pairs of quotes that were sorted into the same pile
by all four coders appeared closer together than did pairs of quotes that were
never placed together. The cluster analysis identified groups of quotes shared
across coders. Barkin, Ryan, and Gelberg used these results to identify
subthemes. (See Patterson, Bettini, and Nussbaum 1993 for another
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 95
Jehn and Doucet (1997) used a similar approach but skipped the first steps
of cutting the data into individual expressions. They asked seventy-six U.S.
managers who had worked in Sino-American joint ventures to describe
recent interpersonal conflicts with business partners. Each person described
two conflicts: one with a same-culture manager and another with a different-
culture manger. The descriptions were usually short paragraphs. From these
152 texts, Jehn and Doucet identified the 30 intracultural and the 30 inter
cultural scenarios that they felt were the most clear and pithy. They recruited
fifty more expatriate managers to assess the similarities (on a five-point
scale) of 60–120 randomly selected pairs of scenarios. When combined
across informants, the managers’ judgments produced two aggregate,
scenario-by-scenario similarity matrices—one for the intracultural conflicts
and one for the intercultural conflicts. Jehn and Doucet analyzed each with
Jehn and Doucet (1997) found they needed four dimensions in the MDS to
explain the intercultural data. They interpreted these dimensions as (1) open
versus resistant to change, (2) situational causes versus individual traits, (3)
high- versus low-resolution potential based on trust, and (4) high- versus
low-resolution potential based on patience. In the scaling of the intracultural
similarity data, they identified four different dimensions: (1) high versus low
cooperation, (2) high versus low confrontation, (3) problem solving versus
accepting, and (4) resolved versus ongoing.
The Jehn-Doucet technique for finding themes is quite novel. Unlike
other investigators, they chose not to break up their textual data into smaller
expressions or quotes. Furthermore, they asked fifty expert informants,
rather than one or two members of the research team, to sort the data. They
did not have sorters identify themes but simply asked them to evaluate how
similar pairs of responses were to each other. They then used the results of
MDS to interpret the larger, overarching themes.
Word Lists and Key Words in Context (KWIC)
Word lists and the KWIC technique draw on a simple observation: If you
want to understand what people are talking about, look closely at the words
they use. To generate word lists, researchers first identify all the unique
words in a text and then count the number of times each occurs. Computer
programs perform this task effortlessly.
Ryan and Weisner (1996) told fathers and mothers of adolescents,
“Describe your children. In your own words, just tell us about them.” Ryan
and Weisner transcribed the verbatim responses and produced a list of all the
96 FIELD METHODS
unique words (not counting 125 common English words, including mostly
prepositions, articles, and conjunctions). Ryan and Weisner counted the
number of times each unique word was used by mothers and by fathers. They
found that mothers were more likely than fathers to use words such as
“friends,” “creative,” “time,” and “honest”; fathers were more likely than
were mothers to use words such as “school,” “good,” “lack,” “student,”
“enjoys,” “independent,” and “extremely.” The words suggested that parents
were concerned with themes related to their children’s independence and to
their children’s moral, artistic, social, athletic, and academic characteristics.
Ryan and Weisner used this information as clues for themes that they would
use later in actually coding the texts.
Word-counting techniques produce what Tesch (1990:139) called data
condensation or data distillation, which helps researchers concentrate on the
core of what might otherwise be a welter of confusing data. But concentrated
data such as word lists and counts take words out of their original context. A
KWIC approach addresses this problem. In this technique, researchers iden-
tify key words or phrases and then systematically search the corpus of text to
find all instances of each key word or phrase. Each time they find an instance,
they make a copy of it and its immediate context. Themes get identified by
physically sorting the examples into piles of similar meaning.
Word-based techniques are fast and are an efficient way to start looking
for themes, particularly in the early stages of research. Word lists and KWIC
techniques can, of course, be combined and are particularly helpful when
used along with ethnographic sources of information.
This approach, also known as collocation, comes from linguistics and
semantic network analysis and is based on the idea that a word’s meaning is
related to the concepts to which it is connected. As early as 1959, Charles
Osgood (1959) created word co-occurrence matrices and applied factor anal
ysis and dimensional plotting to describe the relation of major themes to one
another. The development of computers has made the construction and anal
ysis of co-occurrence matrices much easier and has stimulated the develop
ment of this field (Danowski 1982, 1993; Barnett and Danowski 1992).
Jang and Barnett (1994) examined whether a national culture—U.S. or
Japanese—was discernible in the annual letters to stockholders of CEOs in
U.S. and Japanese corporations. Jang and Barnett selected thirty-five For
tune 500 companies, including eighteen U.S. and seventeen Japanese firms,
matched by their type of business. For example, Ford was matched with
Honda, Xerox with Canon, and so on. All of these firms are traded on the New
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 97
York Stock Exchange, and each year, stockholders receive an annual mes
sage from the CEO or president of these companies. (Japanese firms that
trade on the New York Exchange send the annual letters in English to their
Jang and Barnett (1994) read through the 1992 annual letters to sharehold
ers and (ignoring a list of common words such as “the,” “because,” “if,” and
so on) isolated ninety-four words that occurred at least eight times across the
corpus of thirty-five letters. This produced a 94 (word) × 35 (company)
matrix, where the cells contained a number from 0 to 25, 25 being the largest
number of times any word ever occurred in one of the letters.
Next, Jang and Barnett (1994) created a 35 (company) × 35 (company)
similarity matrix, based on the co-occurrence of words in their letters. In this
case, they used the correlation coefficient to measure similarity among com-
panies. They could have used a number of other measures, including first
dichotomizing the original matrix based on whether the word was mentioned
and then calculating the percentage of times that each company used the
same words. It is unclear to what degree such choices affect outcomes, and
this is clearly an area that needs further research.
Next, Jang and Barnett (1994) analyzed the company-by-company matrix
with MDS and found that the companies divided into two clearly distinct
styles of corporate reporting to stockholders, one American and one Japa-
nese. Next, Jang and Barnett asked, “Which words were important in distin-
guishing the groups, and what were their relationships to the two groups?”
Discriminant analysis indicated that twenty-three words had a significant
effect on differentiating between the groups, so Jang and Barnett (1994) used
correspondence analysis to analyze the 35 (company) × 23 (word) matrix.
Correspondence analysis clusters row and column items simultaneously. In
this case, then, the analysis showed clusters of words and clusters of compa-
nies. The analysis showed that thirteen words were close to the American
group and were tightly clustered together: “board,” “chief,” “leadership,”
“president,” “officer,” “major,” “position,” “financial,” “improved,”
“good,” “success,” “competitive,” and “customer.” To Jang and Barnett,
these words represented two themes: financial information and organiza
Six words were close to the Japanese companies: “income,” “effort,”
“economy,” “new,” “development,” and “quality.” To Jang and Barnett
(1994), these words represented organizational operations and reflected Jap
anese concern for the development of new quality products in order to com
pete in the American business environment. The remaining four words
(“company,” “marketplace,” “people,” and “us”) fell between the American
98 FIELD METHODS
and Japanese clusters. Jang and Barnett felt that these words represented a
more neutral category and did not consider them a theme.
For other examples of how word co-occurrences can be used to identify
themes, see Kirchler’s (1992) examination of business obituaries,
Danowski’s (1982) analysis of Internet-based conferences, Nolan and
Ryan’s (2000) analysis of students’ descriptions of horror films, and
Schnegg and Bernard’s (1996) analysis of German students’ reasons for
studying anthropology. What is so appealing about word-by-word co-
occurrence matrices is that they are produced by computer programs and
there is no coder bias introduced other than to determine which words are
examined. (See Borgatti 1992 and Doerfel and Barnett 1996 for computer
programs that produce word-by-word co-occurrence matrices.)
There is, of course, no guarantee that any analysis of a word co-occurrence
matrix will be meaningful, and it is notoriously easy to read pattern (and thus
meaning) into any set of items.
Metacoding examines the relationship among a priori themes to discover
potentially new themes and overarching metathemes. The technique requires
a fixed set of data units (paragraphs, whole texts, pictures, etc.) and a fixed set
of a priori themes. For each data unit, the investigator asks which themes are
present and, possibly, the direction and valence of each theme. The data are
recorded in a unit-by-theme matrix. This matrix can then be analyzed statisti-
cally. Factor analysis, for example, indicates the degree to which themes
coalesce along a limited number of dimensions. Correspondence analysis,
cluster analysis, or MDS show graphically how units and themes are distrib
uted along dimensions and into groups or clusters.
This technique tends to produce a limited number of large metathemes.
Jehn and Doucet (1996, 1997) used metacoding in their analysis of
intracultural and intercultural conflicts. First, two coders read the 152 con
flict scenarios (76 intracultural and 76 intercultural) and evaluated those sce
narios (on a five-point scale) for twenty-seven different themes they had
identified from the literature on conflict.
This produced two 76 × 27
scenario-by-theme profile matrices—one for the intracultural conflicts and
one for the intercultural conflicts. The first three factors from the inter
cultural matrix reflect (1) interpersonal animosity and hostility, (2) aggrava
tion, and (3) the volatile nature of the conflict. The first two factors from the
intracultural matrix reflect (1) hatred and animosity with a volatile nature and
(2) conflicts conducted calmly with little verbal intensity.
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 99
Themes like these are often not readily apparent, even after a careful and
exhaustive scrutinizing of the text. Because metacoding involves analyzing
fixed units of texts for a set of a priori themes, it works best when applied to
short, descriptive texts of one or two paragraphs.
SELECTING AMONG TECHNIQUES
Given the variety of methods available for coding texts, the obvious ques
tion is, When are the various techniques most appropriate? Clearly, there is
no one right way to find themes, but some techniques are more effective
under some conditions than others. Below, we evaluate the techniques on
five dimensions: (1) kind of data types, (2) required expertise, (3) required
labor, (4) number and types of themes to be generated, and (5) issues of reli-
ability and validity.
Kind of Data
Qualitative researchers work with many kinds of data—textual and
nontextual, verbatim and nonverbatim, long and short. Although all the tech-
niques we have described are appropriate for discovering themes in some
kinds of textual data, only half are useful for nontextual data. For pictures,
sounds, and objects, investigators are limited to looking for repetitions, simi-
larities and differences, missing data, and theory-related material and to
using sorting or metacoding techniques.
In writing field notes, the researcher acts as a kind of theme filter, choos-
ing (often subconsciously) what data are important to record and what data
are not. In this sense, producing field notes is a process of identifying themes.
This inherent filtering process poses a particular set of problems for analyz
ing field notes. When applying techniques that use informant-by-variable
matrices, researchers need to remember that patterns discovered in such data
may come from informants as well as from investigators’ recording biases.
With the exception of metacoding, all twelve techniques can be applied to
rich narrative data. As texts become shorter and less complex, looking for
transitions, metaphors, and linguistic connectors becomes less efficient. Dis
covering themes by looking for what is missing is inappropriate for very
short responses to open-ended questions because it is hard to say whether
missing data represent a new theme or are the result of the data elicitation
technique. Though not impossible, it is inefficient to look for theory-related
material in short answers, so we do not recommend metacoding for this kind
100 FIELD METHODS
Not all techniques are available to all researchers. One needs to be truly
fluent in the language of the text to use techniques that rely on metaphors, lin
guistic connectors, and indigenous typologies or that require spotting subtle
nuances such as missing data. Researchers who are not fluent in the language
should rely on cutting and sorting and on the search for repetitions, transi
tions, similarities and differences, and etic categories (theory-related mate
rial). Word lists and co-occurrences, as well as metacoding, also require less
language competence and so are easier to apply.
Investigators who plan to use word co-occurrence or metacoding need to
know how to manipulate matrices and how to use methods for exploring and
visualizing data—methods such as MDS, cluster analysis, factor analysis,
and correspondence analysis. Those without these skills should use the scru-
tiny techniques, such as looking for repetitions, similarities and differences,
indigenous typologies, metaphors, transitions, or linguistic connectors, and
the process techniques, such as cutting and sorting, word lists, and KWIC,
which do not require skills in handling matrix analysis.
Figure 1 offers suggestions on how to select among the various theme-
identification techniques. Clearly, looking for repetitions and similarities
and differences as well as cutting and sorting techniques are by far the most
versatile techniques for discovering themes. Each can be applied to any type
of qualitative data. Not surprisingly, it is these techniques that are most often
described in texts about qualitative methods.
A generation ago, scrutiny-based techniques required less effort and
resources than did process techniques. Today, computers have made count-
ing words and co-occurrences of words much easier. Software also has made
it easier to analyze larger corpora of texts.
Still, some of the scrutiny-based techniques (searching for repetitions,
indigenous typologies, metaphors, transitions, and linguistic connectors) are
best done by eyeballing, and this can be quite time consuming.
Of all the techniques, we find that using software to generate a common
word list is an efficient way to start looking for themes. (Use packages like
TACT, ANTHROPAC, or Code-A-Text to generate frequency counts of key
) A careful look at a word frequency list and perhaps some quick pile
sorts are often enough to identify quite a few themes. Word co-occurrence
and metacoding require more work and produce fewer themes, but they are
excellent for discovering big themes hidden within the details and nuances of
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 101
Yes No (e.g., sounds, images, objects)
Yes No (e.g., field notes)
5. Similarities & Differences
9. Cutting & Sorting
5. Similarities & Differences
9. Cutting & Sorting
2. Indigenous Typologies
6. Linguistic Connectors
7. Missing Data
8. Theory-Related Material
10. Word Lists & KWIC
11. Word Co-Occurrence
5. Similarities & Differences
9. Cutting & Sorting
7. Missing Data
8. Theory-Related Material
5. Similarities & Differences
9. Cutting & Sorting
2. Indigenous Typologies
7. Missing Data
8. Theory-Related Material
10. Word Lists & KWIC
11. Word Co-Occurrence
5. Similarities & Differences
9. Cutting & Sorting
2. Indigenous Typologies
10. Word Lists & KWIC
11. Word Co-Occurrence
Selecting among Theme-Identification Techniques
NOTE: KWIC = key words in context.
Number and Kinds of Themes
In theme discovery, more is better. It is not that all themes are equally
important. Investigators must eventually decide which themes are most
salient and how themes are related to each other. But unless themes are first
discovered, none of this additional analysis can take place.
We know of no research comparing the number of themes that each tech
nique generates, but our experience suggests that there are differences.
Looking for repetitions, similarities and differences, and transitions and lin
guistic connectors that occur frequently in qualitative data will likely pro
duce more themes than will looking for indigenous metaphors and indige
nous categories that occur less frequently. Of all the scrutiny techniques,
searching for theory-related material or for missing data will likely produce
the least number of new themes. Of the process techniques, we find that cut-
ting and sorting and word lists yield an intermediate number of themes, while
word co-occurrence and metacoding produce only a few metathemes. If the
primary goal is to discover as many themes as possible, then the best strategy
is to apply several techniques.
Cutting and sorting is the most versatile technique. By sorting expressions
into piles at different levels of abstraction, investigators can identify themes,
subthemes, and metathemes. Searching for indigenous typologies and com-
bining word lists and KWIC is particularly useful for identifying subthemes.
In contrast, techniques that analyze aggregated data such as word co-
occurrences and metacoding are particularly good at identifying more
Reliability and Validity
Theme identification does not produce a unique solution. As Dey (1993)
noted, “there is no single set of categories [themes] waiting to be discovered.
There are as many ways of ‘seeing’ the data as one can invent” (pp. 110–11).
Jehn and Doucet (1996, 1997) used three different discovery techniques on
the same set of data, and each produced a different set of themes. All three
emically induced theme sets have some intuitive appeal, and all three yield
analytic results that are useful. Jehn and Doucet might have used any of the
other of the techniques we describe to discover even more themes.
How do investigators know if the themes they have identified are valid?
There is no ultimate demonstration of validity, but we can maximize clarity
and agreement and make validity more, rather than less, likely.
identification involves judgments on the part of the investigator. If these
judgments are made explicit and clear, then readers can argue with the
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 103
researcher’s conclusions (Agar 1980:45). This is one of our motivations for
outlining in detail the techniques investigators use.
Second, we see validity as hinging on the agreement across coders, meth
ods, investigations, and researchers. Intercoder reliability refers to the degree
to which coders agree with each other about how themes are to be applied to
qualitative data. Reliability is important in that it indicates that coders are
measuring the same thing. Strong intercoder agreement also suggests that the
concept is not just a figment of the investigator’s imagination and adds to the
likelihood that a theme is also valid (Sandelowski 1995). Agreement across
techniques gives us further confidence that we have identified appropriate
themes in the same way that finding similar themes across multiple investiga
Bernard (1994) argued that ultimately, the validity of a concept depends
on the utility of the device that measures it and the collective judgment of the
scientific community that a construct and its measure are valid. “In the end,”
he said, “we are left to deal with the effects of our judgments, which is just as
it should be. Valid measurement makes valid data, but validity itself depends
on the collective opinion of researchers” (p. 43). Denzin (1970) assigned
even greater significance to the role of the research community in establish-
ing validity. “Rules for establishing a sound sample, a reliable test, or a valid
scale,” he said, “are only symbolic—they have no meaning other than that
given by the community of scientists” (p. 106).
Patton (1990:468) referred to such an agreement among investigators as
“triangulation through multiple analysts.” It is what makes Lincoln and
Guba’s (1985) team approach to sorting and naming piles of expressions so
appealing. Agreement need not be limited to members of the core research
team. Recall that Jehn and Doucet (1997) asked local experts to sort word
lists into thematic categories, and Barkin, Ryan, and Gelberg (1999) had both
experts and novices sort quotes into piles. The more agreement among team
members, the more confidence we have in themes being valid.
Some investigators also recommend that respondents be given the oppor
tunity to examine and comment on themes and categories (e.g., Lincoln and
Guba 1985:351; Patton 1990:468–69). This is appropriate when one of the
goals of research is to identify and apply themes that are recognized or used
by the people whom one studies, but this is not always possible. The discov
ery of new ideas derived from a more theoretical approach may involve the
application of etic rather than emic themes—that is, understandings held by
outsiders rather than those held by insiders. In such cases, researchers would
not expect their findings necessarily to correspond to ideas and beliefs held
by study participants.
104 FIELD METHODS
We still have much to learn about finding themes. Further research is
needed in five broad areas:
1. How reliable is each technique? To what degree do the same coders find simi
lar themes when performing the task at different points in time? To what
degree do different coders find the same themes on the same data sets?
2. How do identification techniques compare when applied to the same data
sets? For example, do some techniques systematically produce significantly
more themes or subthemes than others? And to what extent do the different
techniques produce overlapping or similar themes? Jehn and Doucet (1996,
1997) have already provided a model for addressing such questions that can
now be applied to other techniques as well.
3. How do identification techniques compare when applied to different data
sets? How much of an effect does the size and complexity of the qualitative
data corpus have on the number, kind, and organization of themes that coders
4. To what extent is theme identification dependent on the number and expertise
of coders? For instance, under what conditions can we expect novices to find
the same number and kinds of themes as novices? And to what extent does
increasing or decreasing the number of coders affect the size and composition
5. Finally, to what extent can we develop automated procedures for finding
themes? Can we create word- and grammar-based algorithms to identify
themes that mirror the processes used and the themes found by human coders?
Only by addressing such issues directly will we be able to explicitly justify
our methodological choices.
1. For thorough overviews of linking themes to specific expressions, see Carey, Morgan, and
Oxtoby (1996). For suggestions about how to describe themes, see Miles and Huberman (1994)
and Ryan and Bernard (2000). For building thematic hierarchies and code books, we recommend
Dey (1993), Carey, Morgan, and Oxtoby (1996), and MacQueen et al. (1998). For identifying
“important” themes and linking them to theoretical models, Strauss and Corbin (1990), Dey
(1993), and Miles and Huberman (1994) are quite helpful.
2. To ensure interrater reliability, the two raters coded thirty-five scenarios in common. The
final rating used in these thirty-five common scenarios was the agreement reached when the rat
ers met together to discuss discrepancies. Rater 1 coded seventy scenarios, rater 2 coded forty
scenarios, and they coded thirty-five scenarios in common (70 + 40 + 35 = 152).
3. TACT(CHASS), ANTHROPAC (Analytic Technologies), and Code-A-Text (Cart
wright) are software packages that have the capacity to convert free-flowing texts into word-by-
document matrices. TACT is a powerful DOS program created by the University of Toronto and
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 105
available free on the Web at http://www.chass.utoronto.ca/cch/tact.html. Code-A-Text is dis
tributed in the United States by Scolari, Sage Publications. ANTHROPAC is created and distrib
uted by Analytic Technologies, Inc., 11 Ohlin Lane, Harvard, MA 01451; phone: (978) 456-
7372; fax: (978) 456-7373; e-mail: email@example.com; Web: www.analytictech.com.
4. For reviews of key issues related to reliability and validity, see Campbell (1957), Camp
bell and Stanley (1963), Cook and Campbell (1979); Guba (1981), Guba and Lincoln (1982);
Lincoln and Guba (1985), Fielding and Fielding (1986), Kirk and Miller (1986), Hammersley
(1992), and Denzin (1997).
Agar, M. 1973. Ethnography and the addict. In Cultural illness and health, edited by L. Nader
and T. W. Maretzki, 35–42. Washington, DC: American Anthropology Association.
. 1979. Themes revisited: Some problems in cognitive anthropology. Discourse Pro
. 1980. Getting better quality stuff: Methodological competition in an interdisciplinary
niche. Urban Life 9:34–50.
. 1983. Political talk: Thematic analysis of a policy argument. Policy Studies Review 2
Barkin, S., G. Ryan, and L. Gelberg. 1999. What clinicians can do to further youth violence pri-
mary prevention: A qualitative study. Injury Prevention 5:53–58.
Barnett, G., and J. Danowski. 1992. The structure of communication: A network analysis of the
international communication association. Human Communication Resources 19:164–285.
Becker, H. 1993. How I learned what a crock was. Journal of Contemporary Ethnography
Berelson, B. 1952. Content analysis in communication research. Glencoe, IL: Free Press.
Bernard, H. R. 1994. Research methods in anthropology: Qualitative and quantitative
approaches. 2d ed. Walnut Creek, CA: AltaMira.
Bogdan, R., and S. K. Biklen. 1982. Qualitative research for education: An introduction to the
ory and methods. 2d ed. Boston: Allyn & Bacon.
Bogdan, R., and S. J. Taylor. 1975. Introduction to qualitative research methods. New York:
Borgatti, S. P. 1992. ANTHROPAC version 4.0. Columbia, SC: Analytic Technologies. Avail
able from www.analytictech.com.
Bulmer, M. 1979. Concepts in the analysis of qualitative data. Sociological Review 27 (4): 651–77.
Burton, M. L., and L. Kirk. 1980. Ethnoclassification of body parts: A three-culture study. In
Indigenous knowledge systems and development, edited by D. Brokensha, D. M. Warren,
and O. Werner, 269–96. Lanham, MD: University Press of America.
Campbell, D. 1957. Factors relevant to the validity of experiments in social settings. Psychologi
cal Bulletin 54:297–312.
Campbell, D., and J. C. Stanley. 1963. Experimental and quasi-experimental designs for
research on teaching. In Handbook of research on teaching, edited by N. Gage, 171–246.
Chicago: Rand McNally.
Carey, J. W., M. Morgan, and M. J. Oxtoby. 1996. Intercoder agreement in analysis of responses
to open-ended interview questions: Examples from tuberculosis research. Cultural Anthro
pology Methods Journal 8 (3): 1–5.
106 FIELD METHODS
Casagrande, J. B., and K. L. Hale. 1967. Semantic relations in Papago folk definitions. In Studies
in Southwestern ethnolinguistics: Meaning and history in the languages of the American
Southwest, edited by D. H. Hymes and W. E. Bittle, 165–93. Paris: Mouton & Co.
Charmaz, K. 1990. “Discovering” chronic illness: Using grounded theory. Social Science and
. 2000. Grounded theory: Objectivist and constructivist methods. In Handbook of quali
tative research, 2d ed., edited by N. Denzin and Y. Lincoln, 509–36. Thousand Oaks, CA:
Coffey, A., and P. Atkinson. 1996. Making sense of qualitative data: Complementary research
strategies. Thousand Oaks, CA: Sage.
Cook, T. D., and D. Campbell. 1979. Quasi-experimentation: Design and analysis issues for
field settings. Chicago: Rand McNally.
D’Andrade, R. 1991. The identification of schemas in naturalistic data. In Person schemas and
maladaptive interpersonal patterns, edited by M. J. Horowitz, 279–301. Chicago: Univer
sity of Chicago Press.
. 1995. The development of cognitive anthropology. Cambridge, UK: Cambridge Uni
Danowski, J. 1982. Computer-mediated communication: A network-based content analysis
using a CBBS conference. In Communication yearbook, edited by R. Bostrom, 905–25. New
Brunswick, NJ: Transaction Books.
. 1993. Network analysis of message content. In Progress in communication science,
volume 12, edited by W. D. Richards and G. A. Barnett, 197–221. Norwood, NJ: Ablex.
Denzin, N. K. 1970. The research act: A theoretical introduction to sociological methods. Chi-
. 1997. Interpretive ethnography. Thousand Oaks, CA: Sage.
Dey, I. 1993. Qualitative data analysis: A user-friendly guide for social scientists. London:
Routledge Kegan Paul.
Doerfel, M. L., and G. A. Barnett. 1996. The use of Catpac for text analysis. Cultural Anthropol-
ogy Methods Journal 8 (2): 4–7.
Fielding, N. G., and J. L. Fielding. 1986. Linking data. Newbury Park, CA: Sage.
Gal, S. 1991. Between speech and silence: The problematics of research on language and gender.
In Gender at the crossroads of knowledge: Feminist anthropology in the postmodern era,
edited by Michaela di Leonardo, 175–203. Berkeley: University of California Press.
Glaser, B. G. 1978. Theoretical sensitivity. Mill Valley, CA: Sociology Press.
Glaser, B. G., and A. Strauss. 1967. The discovery of grounded theory: Strategies for qualitative
research. New York: Aldine.
Greenhalgh, S. 1994. Controlling births and bodies. American Ethnologist 21 (1): 3–30.
Guba, E. G. 1978. Toward a methodology of naturalistic inquiry in educational evaluation.
Monograph 8. Los Angeles: UCLA Center for the Study of Evaluation.
. 1981. Criteria for assessing the trustworthiness of naturalistic inquiries. Educational
Communication and Technology Journal 29:75–91.
Guba, E. G., and Y. S. Lincoln. 1982. Epistemological and methodological bases of naturalistic
inquiry. Educational Communication and Technology Journal 30:233–52.
Hammersley, M. 1992. What’s wrong with ethnography? New York: Routledge.
Holland, D. 1985. From situation to impression: How Americans get to know themselves and
one another. In Directions in cognitive anthropology, edited by J. W. D. Dougherty, 389–
412. Urbana: University of Illinois Press.
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 107
Jang, H.-Y., and G. Barnett. 1994. Cultural differences in organizational communication: A
semantic network analysis. Bulletin de Méthodologie Sociologique 44 (September): 31–59.
Jehn, K. A., and L. Doucet. 1996. Developing categories from interview data: Text analysis and
multidimensional scaling. Part 1. Cultural Anthropology Methods Journal 8 (2): 15–16.
. 1997. Developing categories for interview data: Consequences of different coding and
analysis strategies in understanding text. Part 2. Cultural Anthropology Methods Journal 9
Kempton, W. 1987. Two theories of home heat control. In Cultural models in language and
thought, edited by D. Holland and N. Quinn, 222–42. Cambridge, UK: Cambridge Univer
Kirchler, E. 1992. Adorable woman, expert man: Changing gender images of women and men in
management. European Journal of Social Psychology 22:363–73.
Kirk, J., and M. L. Miller. 1986. Reliability and validity in qualitative research. Beverly Hills,
Krippendorf, K. 1980. Content analysis: An introduction to its methodology. Beverly Hills, CA:
Lakoff, G., and M. Johnson. 1980. Metaphors we live by. Chicago: University of Chicago Press.
Lincoln, Y. S., and E. G. Guba. 1985. Naturalistic inquiry. Beverly Hills, CA: Sage.
Lindsay, P. H., and D. A. Norman. 1972. Human information processing: An introduction to
psychology. New York: Academic Press.
MacQueen, K., E. McLellan, K. Kay, and B. Milstein. 1998. Code book development for team-
based qualitative analysis. Cultural Anthropology Methods Journal 10 (2): 31–36.
Maxwell, J. 1996. Qualitative research design: An interactive approach. Thousand Oaks, CA:
Miles, M. B., and A. M. Huberman. 1994. Qualitative data analysis: An expanded sourcebook.
Thousand Oaks, CA: Sage.
Nolan, J., and G. Ryan. 2000. Fear and loathing at the cinemaplex: Gender differences in
descriptions and perceptions of slasher films. Sex Roles 42 (1/2): 39–56.
Opler, M. E. 1945. Themes as dynamic forces in culture. American Journal of Sociology 5:198–
Osgood, C. 1959. The representational model and relevant research methods. In Trends in con
tent analysis, edited by I. de Sola Pool, 33–88. Urbana: University of Illinois Press.
Patterson, B. R., L. Bettini, and J. F. Nussbaum. 1993. The meaning of friendship across the life-
span: Two studies. Communication Quarterly 41 (2): 145–60.
Patton, M. Q. 1990. Qualitative evaluation and research methods. Thousand Oaks, CA: Sage.
Price, L. 1987. Ecuadorian illness stories. In Cultural models in language and thought, edited by
D. Holland and N. Quinn, 313–42. Cambridge, UK: Cambridge University Press.
Quinn, Naomi. 1996. Culture and contradiction: The case of Americans reasoning about mar
riage. Ethos 24:391–425.
Ryan, G. W. 1995. Medical decision making among the Kom of Cameroon: Modeling how char
acteristics of illnesses, patients, caretakers, and compounds affect treatment choice in a rural
community. Ph.D. diss., University of Florida, Gainesville.
. 1999. Measuring the typicality of text: Using multiple coders for more than just reliabil
ity and validity checks. Human Organization 58 (3): 313–22.
Ryan, G. W., and H. R. Bernard. 2000. Data management and analysis methods. In Handbook of
qualitative research, 2d ed., edited by N. Denzin and Y. Lincoln, 769–802. Thousand Oaks,
Ryan, G. W., and T. Weisner. 1996. Analyzing words in brief descriptions: Fathers and mothers
describe their children. Cultural Anthropology Methods Journal 8 (3): 13–16.
108 FIELD METHODS
Sandelowski, M. 1995. Qualitative analysis: What it is and how to begin. Research in Nursing
and Health 18:371–75.
Schnegg, M., and H. R. Bernard. 1996. Words as actors: A method for doing semantic network
analysis. Cultural Anthropology Methods Journal 8 (2): 7–10.
Shapiro, G., and J. Markoff. 1997. A matter of definition. In Text analysis for the social sciences:
Methods for drawing statistical inferences from texts and transcripts, edited by C. Roberts,
9–34. Mahwah, NJ: Lawrence Erlbaum.
Silverman, D. 1993. Interpreting qualitative data: Methods for analysing talk, text and interac
tion. London: Sage.
Spradley, J. P. 1972. Adaptive strategies of urban nomads. In Culture and cognition: Rules,
maps, and plans, edited by J. P. Spradley, 235–78. New York: Chandler.
. 1979. The ethnographic interview. New York: Holt, Rinehart & Winston.
Strauss, A. 1987. Qualitative analysis for social scientists. Cambridge, UK: Cambridge Univer
Strauss, A., and J. Corbin. 1990. Basics of qualitative research: Grounded theory procedures
and techniques. Thousand Oaks, CA: Sage.
Strauss, C. 1992. What makes Tony run? Schemas as motive reconsideration. In Human motives
and cultural models, edited by R. D’Andrade and C. Strauss, 191–224. Cambridge, UK:
Cambridge University Press.
. 1997. Research on cultural discontinuities. In A cognitive theory of cultural meaning,
edited by C. Strauss and N. Quinn, 210–51. Cambridge, UK: Cambridge University Press.
Strauss, C., and N. Quinn. 1997. A cognitive theory of cultural meaning. Cambridge, UK: Cam-
bridge University Press.
Tesch, R. 1990. Qualitative research: Analysis types and software tools. New York: Falmer.
Thompson, S. [1932–1936] 1993. Motif-index of folk-literature: A classification of narrative
elements in folktales, ballads, myths, fables, mediaeval romances, exempla, fabliaux, jest-
books, and local legends. Bloomington: Indiana University Press.
Weller, S. C., and A. K. Romney. 1988. Systematic data collection. Newbury Park, CA: Sage.
Werner, O., and G. M. Schoepfle. 1987. Systematic fieldwork. 2 vols. Newbury Park, CA: Sage.
GERY W. RYAN (Ph.D., University of Florida) is a behavioral scientist at RAND. He has
conducted fieldwork on health care choices in United States, Latin America, and Africa.
He also has written and lectured on qualitative data collection and analysis techniques
and was the associate director of the Fieldwork and Qualitative Data Laboratory at
UCLA Medical School. Before joining RAND, he was an assistant professor of anthro
pology at the University of Missouri–Columbia. He was a coeditor of Cultural Anthro
pology Methods Journal (1993–1998) and is currently on the editorial board of Field
Methods. He has published in Social Science and Medicine, Human Organization, and
Archives of Medical Research.
H. RUSSELL BERNARD is a professor of anthropology at the University of Florida. His
research interests include the consequences of literacy in previously nonliterary lan
guages and various aspects of social networks analysis. He is the author of Social
Research Methods (2000, Sage) and the editor of Field Methods.
Ryan, Bernard / TECHNIQUES TO IDENTIFY THEMES 109