H.

RUSSELL BERNARD
GERY RYAN

Sixteen

Text Analysis:
Qualitative and Quantitative Methods

Introduction
There is growing interest across the social sciences in the systematic analysis of
“text.”1 Little wonder: Most of the recoverable data about human thought and human
behavior is text of one kind or another. In this chapter, we survey methods of text
analysis in the social sciences and particularly how anthropologists have used those
methods to look for meaning and pattern in written text.2
We cover two broad types of text analysis: the linguistic tradition, which treats text
as an object of analysis itself, and the sociological tradition,which treats text as a
window into human experience. For the linguistic tradition, we review how
anthropologists have collected and produced texts, analyzed indigenous literatures,
discovered patterns and structures in performance styles, and compared the pro-
duction of narratives within and across cultures. For the sociological tradition, we
review the methods of schema analysis, grounded theory, classical content analysis,
semantic network analysis, cognitive mapping, and Boolean analysis.
Throughout, we focus on methods for collecting and analyzing written texts such
as political speeches, song lyrics, personal diaries, transcriptions of interviews,
newspaper editorials, and so on. Many of these methods serve just as well in dealing
with images, such as photographs, home movies, video tape, commercial movies,
kinescopes of old television shows, etc.
In the study of text, some scholars use methods identified with the humanist
tradition while others use methods identified with the positivist tradition. The former

595
596 BERNARD / RYAN

involves interpretation and the search for meaning. The latter involves the reduction
of texts to codes that represent themes or concepts and the application of quantitative
methods to find patterns in the relations among the codes. Nowhere is Eric Wolf’s
aphorism that “anthropology is the most humanistic of the sciences and the most
scientific of the humanities” (1964:88) better demonstrated than in the practice of text
analysis.

Anthropology: A Passion for Collecting Texts

Debates about the value of structuralism, functionalism, historical particularism,
materialism, and postmodernism come and go, but the value of faithfully produced
texts is undisputed. Among Franz Boas’s lasting contributions is the corpus of texts
he collected, translated. and published (or deposited in archives) from speakers of
Bella Bella (1928b), Sahaptin and Salishan (1917), Keresan (1928a) and, with
George Hunt, Kwakiutl (Boas and Hunt 1902–1905, 1906; Boas 1910, 1935–1943).
In 1893, Boas taught Hunt to write Kwakiutl, Hunt’s native language. By the time
Hunt died in 1933, he had produced 5,650 pages of text—a corpus from which Boas
produced most of his reports about Kwakiutl life (Rohner 1966).
To the extent possible before the invention of voice-recording devices, Boas
trained his students and collaborators to collect verbatim texts.3 Following Boas’s
example with Hunt, Paul Radin worked with Sam Blowsnake, a Winnebago.
Blowsnake wrote the original manuscript (in Winnebago) that became (in translation)
Crashing Thunder: An Autobiography of a Winnebago Indian (Blowsnake and Radin
1983 [1920, 1926]). Among Boas’s other students, Swanton (1909), Goddard (1911),
Kroeber (1907), Lowie (1930), and Sapir (Sapir and Dixon 1910; Sapir and Curtain
1974; Sapir and Swadesh 1978) collected and analyzed indigenous language text, and
Margaret Mead produced hours and hours of cinema verité about Bali dance—a rich,
textual record that can be turned to again and again as new insights and new methods
of analysis become available.
The concern for the collection and archiving of text remains undiminished in
anthropology. In the 1970s, Eric Hamp edited the Native American Texts Series of
the International Journal of American Linguistics (see, for example, Furbee-Losee
1976; Bernard and Salinas Pedraza 1976). As the literary language of a pre–Colonial
Era civilization, Maya has attracted particular attention. Munro Edmonson’s
translations of the books of Chilam Balam (1982, 1986) and Dennis Tedlock’s
translation of the Popol Vuh (1985) are outstanding examples.
Following the example of Boas with Hunt and Radin with Blowsnake, Bernard
(Salinas Pedraza and Bernard 1978; Bernard and Salinas Pedraza 1989), El Guindi
and Hernández Jiménez (1986), Lurie (1961), and Sexton (Bizarro and Sexton
1981, 1985), among others, have helped indigenous people create narratives
themselves in indigenous languages. Bernard (1997) has provided indigenous people
with the computer technology and training to produce their own books in their own
 Text Analysis 597

languages. Here again, the emphasis is on the production of texts, not on their
analysis. (See Salinas Pedraza [1997] and González Ventura [1997] for indigenous
perspectives on the production of indigenous-language text by computer.)

Analysis of Indigenous Literature

There are about 240 languages in the world spoken by more than a million people.
Some of those languages have very long literary histories, and while smaller
languages are disappearing, other indigenous ones are developing new literary tradi-
tions. Postel-Coster (1977) studies an indigenous tradition of novels in Indonesia that
goes back to around 1920 when a thriving production of new literature appeared in
Malay by western Sumatran, or Minangkabau, writers.
Postel-Coster does not see such novels as a source of factual information, but as
the modern continuation of the myth in nonliterary societies. Although novels can not
be taken as documentaries, he says, important elements of social reality can be
exposed in novels without allusion to actual events. In novels about young couples
who must struggle against social rules and disapproval to marry, Postel-Coster finds
that “ many problems of Minangkabau culture are explicitly dealt with: matrilineal
succession, polygynous marriage, the enormous impact of the extended family on an
individual’s life, and the question of merantau—the traditional emigration of young
people, mostly men, to other areas” (1977:137).
Besnier (1995) has studied an indigenous literacy tradition on the Polynesian atoll
of Nukulaelae. The texts produced by the people of Nukulaelae include letters,
sermons, and announcements of events. Besnier analyzes correspondence by identify-
ing and presenting exemplars of structural regularities. For example, the main body
of letters on Nukulaelae “usually begins with a greeting identical to the greeting used
in face-to-face interactions (taalofa “hello,” a Samoan borrowing). This is followed
by references to the health of everyone at the writer’s and recipient’s ends, and
sometimes a very long series of invocations to God’s grace and kindness” (Besnier
1995:86). Besnier follows this with a series of exemplars.
Besnier noticed that letter writers “always adhere to a religious reference scheme
in opening themes, Christian for the majority, or Baha’i, etc., for the handful of
religious converts. In letters written by younger people, the introduction is usually
much shorter and more predictable in content than that of letters written by older
individuals. These introductions bear many similarities to the beginning of formal
speeches” (1995:87).
Using literary methods of analysis, then, Besnier identifies covariation between
certain elements of style in text and independent variables like religious affiliation
and age.
598 BERNARD / RYAN

Patterns in Performance
The discovery of regularities in narrative performance is achieved mostly through the
analysis of written text (for a review, see Hanks 1989). The work of Dell Hymes is of
singular importance. In 1977, Hymes reported that “the narratives of the Chinookan
peoples of Oregon and Washington can be shown to be organized in terms of lines,
verses, stanzas, scenes, and what many call acts.” Hymes felt that this discovery
might be relevant to many indigenous languages of the Americas (1977:431).
Hymes looked at texts in Shoalwater Chinook and Kathlamet Chinook. The texts
had been collected by Boas between 1890–1894 from one informant who happened
to speak both mutually unintelligible languages. Hymes also examined texts from
Clackamas Chinook (collected in 1930 and 1931 by Melville Jacobs) and in Wasco
(Wishram) Chinook (collected by Sapir in 1905, by Hymes in the 1950s. and by
Michael Silverstein in the 1960s and ’70s). Hymes found that features of Chinook
that might have seemed idiosyncratic to the speakers of Shoalwater, Kathlamet, and
Clackamas Chinook “are in fact part of a common fabric of performance style” so that
the three languages “share a common form of poetic organization” (Hymes
1977:431).
This was a truly important discovery, for it made clear that Native North American
texts have something to contribute to a general theory of poetics and literature. Hymes
discovered verses, not by counting lines of text “but by recognizing repetition within
a frame. . . . Covariation between form and meaning, between units with a recurrent
Chinookan pattern of narrative organization, is the key” (Hymes 1977:438).
In some texts, Hymes found recurrent linguistic elements that made the task easy.
Linguists who have worked with precisely recorded texts in Native American
languages have noticed the recurrence of elements like “Now,” “Then,” “Now then,”
and “ Now again.” These often signal the separation of verses and “once such
patterning has been discovered in cases with such markers, it can be discerned in
cases without them” (1977:439). The method is to look for “abstract features that
co-occur with the use of initial particle pairs in the narratives” of other speakers
who use initial particle pairs. The method, then, is a form of controlled comparison.
In a series of articles and books (1976, 1977, 1980a, 1980b, 1981) Hymes showed
that most Native American texts of narrative performance (going back to the early
texts collected by Boas and his students and continuing in today’s narrative
performance as well) are organized into verses and stanzas that are aggregated into
groups of either fives and threes or fours and twos. Boas and his students organized
the narratives of American Indians into lines. This hid from view “a vast world of
poetry waiting to be released by those of us with some knowledge of the languages”
(Virginia Hymes 1987:65).
According to Virginia Hymes (1987:67–68), Dell Hymes’s method involves
“working back and forth between content and form, between organization at the level
of the whole narrative and at the level of the details of lines within a single verse or
even words within a line.” Gradually, an analysis emerges that reflects the analyst’s
 Text Analysis 599

understanding of the narrative tradition and of the particular narrator. Virginia

Hymes emphasizes that it is “only through work with many narratives by many
narrators that the analyst builds up a knowledge of the range of narrative devices used
in the language and the variety of uses to which they may be put” (1987:67–68).
In his study of Zuni narratives, Tedlock (1972:221) found that such paralinguistic
features as voice quality, loudness, and pausing are key indicators of performance
(and see Woodbury [1987] for an examination of prosodic features in Yup’ik Eskimo
narrative). Hymes points out that the texts recorded by people like Sapir and Jacobs
often had paralinguistic features such as nonphonemic vowel length and nonpho-
nemic stress (for emphasis) marked. Some linguists of the time emphasized phonemic
transcription under “the unquestioned assumption that linguistic structure should be
built up on the basis of referential function alone, to the exclusion of features serving
stylistic function” (Hymes 1977:452–453). This wasn’t, of course, how Boas and his
students did things. They used all manner of conventions for indicating expressive
as well as strictly phonemic features of narratives that they recorded.
Thus, says Hymes, the view of the Boasians that “one should regard and present
conventional phonetic habits, expressive as well as referential, comes to seem in the
context of ethnopoetics, not old hat, but a cause for gratitude” (1977:453).
Tedlock (1987) showed the exegetical power that linguistic methods can bring to
text analysis. He had translated the Popol Vuh, a sixteenth-century Quiché Maya
manuscript that had been written out by Francisco Ximénez, a missionary of the time.
Suppose, Tedlock asked, we have an ancient text, one that was meant to have been
narrated. Can we narrate it today as performers would have done in ancient times?
In achieving his translation of the Popol Vuh, Tedlock had relied on Andrés Xiloj,
a modern speaker of Quiché. Xiloj had not been trained to read Maya, but he was
literate in Spanish and made the transition very quickly. “ When given his first chance
to look at the Popol Vuh text, he produced a pair of spectacles and began reading
aloud, word by word” (Tedlock 1987:145).
As was true of many medieval manuscripts in Europe, Ximénez’s rendition of the
Popol Vuh was more or less an undifferentiated mass of text with almost no
punctuation—that is, no clues on how a performer might have “emphasized or elided
the boundaries of . . . segments of discourse through the use of intonational contours,
or how he might have varied his timing through the placement of pauses” (Tedlock
1987:147).
The solution to this problem was to capture oral narratives (not just casual speech)
from modern speakers of the language—speeches, prayers, songs, stories—and to
look for “patterns in the wording that have analogs in the ancient text” and how these
patterns are enunciated (Tedlock 1987:147). Tedlock devises conventions for marking
pauses, accelerations, verse endings, and so on. It is in the very use of such written
marks that we see Tedlock’s analysis—that is, his understanding of how a
performance went. He can apply the written conventions to ancient texts once the
analysis of performance is done. Then, by using techniques from linguistics (such as
systematic comparison to look for recurrent sound patterns that signify variations in
600 BERNARD / RYAN

meaning), Tedlock found that Quiché verse has the same structure as ancient Middle
Eastern texts—texts that predate Homer. Indeed, he concluded it is the same structure
found in all living oral traditions that have not yet been influenced by writing. This
is a contribution to a general theory of poetics and literature of the sort that Hymes
had envisioned a decade earlier for the methods of ethnopoetics.
Sherzer (1994) presents a detailed analysis of a two-hour performance by Chief
Olopinikwa of a traditional San Blas Kuna chant. The chant was recorded in 1970.
Like many linguistic anthropologists, Sherzer had taught an assistant, Alberto
Campos, to use a phonetic transcription system. After the chant, Sherzer asked
Campos, to transcribe and translate the tape. Campos put Kuna and Spanish on left-
and right-facing pages (1994:907).
By studying Campos’s translation against the original Kuna, Sherzer was able to
pick out certain recurrent features. Campos left out the chanted utterances of the
responding chief (usually something like “so it is”), which turned out to be markers
for verse endings in the chant. Campos also left out so-called framing words and
phrases (like “Thus” at the beginning of a verse and “it is said, so I pronounce” at the
end of a verse). These contribute to the line and verse structure of the chant. Finally,
“instead of transposing metaphors and other figurative and allusive language into
Spanish” Campos “explains them in his translation” (Sherzer 1994:908).
A key method of text analysis in ethnopoetics is text presentation. It turns out that
verse breaks are determined by the regular turn-taking between Chief Olopinikwa and
the responding chief and that verses and lines have a regular melodic shape. In his
presentation of Chief Olopinikwa’s performance, Sherzer breaks the work into lines
and verses, using the convention of beginning verses and lines flush on the left of
each page and indenting the lines of the responding chief. Earlier, in his presentation
of The Hot Pepper Story, Sherzer (1990:178) used a highly literal translation. The
text repeats a small number of words and themes, and Sherzer felt that a more liberal
translation would fail to capture the poetics of performance. So, Sherzer describes the
thematic elements he sees in the text but uses the device of literalness in the
translation to draw the reader’s attention to those elements.
Text analysis produces new text, which in turn can be analyzed. Hanks (1988)
reviewed Edmunson’s (1982) translation of The Book of Chilam Balam of Tizimin.
Edmunson had translated and annotated the original Mayan corpus into 5,514 lines
of text, changing the format of presentation in the process. In the original, the lines
had run clear across the page, but Edmonson presented the text in short lines to
emphasize what he considered to be the verse structure. Hanks analyzes not only the
Mayan text, but the literary style that Edmonston used in his presentation.
In translating Ñähñu (Otomí) parables, folk tales, and jokes, Bernard and Salinas
(1976) presented a fully literal translation and a fully liberal translation, in addition
to a transcription of the Ñähñu. At the time, Bernard felt that there was no way to
mediate between the characteristics of the original, free Ñähñu and a free English
translation. Later, in translating Salinas’s four-volume ethnography of the Ñähñu,
Bernard tried a middle course—one in which the English is grammatical but also one
 Text Analysis 601

which makes clear from the style that it is a translation (see Bernard and Salinas
1989).
Anthropologists are still experimenting with methods for presenting text of
indigenous performance that capture the subtleties of performance. How can one
know if a particular presentation does, in fact, capture regular features of narrative?
Tedlock’s work with Andrés Xiloj, Sherzer’s with Alberto Campos, and Bernard’s
with Jesús Salinas are, we think, experiments in method.
Sherzer and Woodbury (1987) observed that highly artistic, creative performance
may be based on an underlying cognitive representation. These representations are
knowable, they said, by systematically comparing texts across performances. Thus,
they pose the possibility that there are schemas for performance—schemas that go
beyond the lexical and syntactic levels of grammar. We will return to the methods of
schema analysis.

Inter- and Intracultural Comparisons of Narratives

Comparing how people produce narratives in different cultures has long been of
interest to many social scientists. In 1975, Wallace Chafe and five colleagues received
a grant from the National Institutes for Mental Health to conduct research on how
people store knowledge in the mind. In the spirit of the pioneering work by F. C.
Bartlett (1967 [1932]) on how people remembered and retold folk tales, Chafe and
his colleagues hired a professional filmmaker and produced the Pear Story film.4
They showed it to more than 50 speakers of English in California, and to 20 or more
speakers of Chinese, Japanese, Malay, Greek, Thai, German, Haitian Creole, Persian,
and Sapultec (a Mayan language in Guatemala). They asked their informants for
narratives about the film within half an hour of viewing and transcribed the tapes
verbatim (with lots of attention to pauses, pause filters, stutterings, and such).
Examining the Pear Stories, Chafe (1980) identified the existence of idea units
(about six seconds long and containing about six words) commonly marked by an
intonation contour that involves a rise in pitch or a fall. Chafe suggests “that these
idea units, these spurts of language, are linguistic expressions of focuses of
consciousness” (1980:15) that are packaged into sentences. Children often package
entire narratives into a single unit, but so do adults sometimes.
Downing (1980) compared the 20 English and 20 Japanese Pear story narratives,
marking all nominal references to concrete entities (1,363 in the English narratives;
786 in the Japanese). She found that speakers of both languages make extensive use
of a similar body of “basic” lexemes, but that cognitive, stylistic, and textual
constraints may cause individual speakers to substitute other words and phrases for
the basic terms at a given point in the narrative.
Herzfeld (1977) analyzed multiple renditions of the khelidonisma, or swallow
song, sung in modern Greece as part of the welcoming of spring. He collected texts
of the song from ancient, medieval, and modern historical sources and recorded texts
602 BERNARD / RYAN

of current-day renditions in several locations across Greece. His purpose was to show
that inconsistencies in the texts come not from “some putative irrationality in the
processes of oral tradition” but are, in fact, reflections of structural principles that
underlie the rite de passage for welcoming spring in rural Greece. To make his point,
Herzfeld looks for anomalies across renditions—like “ March, my good March” in one
song compared to “ March, terrible March” in another. Herzfeld claims that the word
“good” is used ironically in Greek where the referent is a source of anxiety.
Is March a subject of symbolic anxiety for Greek villagers? Yes, says, Herzfeld, it
is, as evidenced by widely observed practices such as avoidance of certain activities
during the drimata (the first three days of March). Herzfeld supports his analysis by
referring to the drimes, a word that denotes the first three days of August, which are
associated with malevolent spirits. Since March is the transition from winter to
summer and August is the transition from summer to winter, Herzfeld concludes that
there is symbolic danger associated with these mediating months. He finds support
for this analysis in the fact that February is never referred to with an unequivocally
good epithet.
This is symbolic analysis—the search for symbols and their interconnection in the
expression of culture. The method for doing this kind of analysis requires deep
involvement with the culture, including an intimate familiarity with the langauge, so
that the symbolic referents emerge during the study of those expressions—as in the
study of texts here. You can’t see the connections among symbols if you don’t know
what the symbols are and what they are supposed to mean.
Furbee (1996) is doing an ongoing study of a new cult in and around Lomantán,
a Tojolabal Maya village in the state of Chiapas, Mexico. According the the local
story, Dominga Hernández was cutting wood on April 30, 1994, when God appeared
and gave her images to care for. The images included the Christ Child, the Virgin
Mary, Saint Joseph, and animals of the crèche. Hernández was to keep the images in
her house and have a church built in Lomantán. Then she could turn over care of the
the images to the community. Within 43 days, Hernández had mobilized support for
purchase of materials and for the donated labor that went into building the church
that now houses the images.
In 1996, Furbee and Jill Brody collected 26 versions of the Lomantán miracle. The
tellers of these stories come from eighteen different villages across the region,
including ten that are loyal to the PDR (an opposition party) and eight that are loyal
to the PRI (the ruling political party).
Spanish loan words in the texts run from 1.9%–12.5% (the high end is the tale
told by Hernández herself), but the number of Spanish loan words in the texts from
the PRI-affiliated villages is 22% greater than the number of loan words in the texts
from the PDR-affiliated villages (the difference is statistically significant). In other
words “the greater Spanish loan usage is where one would expect to find it, with
people from villages sympathetic to prevailing power.” The PDR-affiliated villages
are more sympathetic to the Zapatista cause and the speech of those villages contains
fewer Spanish loan words—“just what one might predict from those who oppose the
 Text Analysis 603

prevailing hegemony and who are engaged in a revitalization movement” (Furbee

1996:13).
La Llorona (the weeping woman) is a morality tale told across Mexico. Mathews
(1992) collected 60 tellings of it. Here is one telling, which Mathews says is typical:

La Llorona was a bad woman who married a good man. They had children and all was
well. Then one day she went crazy and began to walk the streets. Everyone knew but
her husband. When he found out he beat her. She had much shame. The next day she
walked into the river and drowned herself. And now she knows no rest and must forever
wander the streets wailing in the night. And that is why women must never leave their
families to walk the streets looking for men. If they are not careful they will end up like
La Llorona. (1992:128)

In another telling, La Llorona kills herself because her husband becomes a drunk
and loses all their money. In yet another, she kills herself because her husband is seen
going with other women and La Llorona, in disbelief, finally catches him paying off
a woman in the streets.
Borrowing from Mandler’s (1984) notion that stories are composed of an ordered
series of constituent units, Mathews builds a grammar of the La Llorona stories. It is
this grammar, says Mathews, this schema, that accounts for the success of the tale’s
motivational force. The morality tale succeeds in shaping people’s behavior because
“the motives of the main characters draw upon culturally shared schemas about
gendered human nature” (1992:129). Men, according to Mathews’s understanding
of the cultural model in rural mestizo Oaxaca, view women as sexually uncontrolled.
Unless they are controlled, or control themselves, their true nature will emerge and
they will begin (as the story says) to “walk the streets” in search of sexual
gratification. Men, for their part, are viewed by women as sexually insatiable. Men
are driven, like animals, to satisfy their desires, even at the expense of family
obligations. In her grammar of La Llorona tales, Mathews shows that women have
no recourse but to kill themselves when they cannot make their marriages work.
Mathews, however, goes beyond simply identifying the schema; she offers an
explanation of where key parts of the schema come from. Most marriages in the
village where Mathews did her research are arranged by parents and involve some
exchange of resources between the families. Consequently, “natal families are usually
unwilling to take back a daughter permanently and thereby contribute to the break-up
of a marriage. So the only option perceived to be open to a woman who wants to
terminate her marriage is suicide” (1992:150). Thus, Mathews offers a materialist
explanation of how structural features in the society effect superstructural outcomes
(perceptions) and consequent behavior—the inclusion of suicide by the woman in
virtually all tellings of the La Llorona morality tale, despite significant variations in
the tellings by men and women.
We now turn to the sociological analysis of text, beginning with the search for
schemas.
604 BERNARD / RYAN

Schemas, Models, and Metaphors

Schema analysis combines elements of the linguistic and sociological traditions. It
pays particular attention to linguistic and paralinguistic features such as metaphors,
proverbs, repetitions, pauses, speaker transitions, turn taking, and interruptions in the
search for mental models that motivate action (D’Andrade 1984, 1987, 1995). (For
a review of the concept of mental models, see Johnson-Laird [1983, 1989].)
Schema analysis is based on the idea that there is too much information about
reality for people to deal with and that people must carry around some simplifications
that help make sense of the welter of information to which they are exposed (Casson
1983:430). Early work by Bartlett (1967 [1932]) on how people remember things
suggested the existence of such simplifications or “building blocks of cognition”
(Rumelhart 1980).
In an influential book, Schank and Abelson (1977) postulated that schemas—or
scripts, as they called them—enable culturally skilled people to fill in details of a
story. We often say things like “Fred lost his data because he forgot to save his work.”
We know that Fred’s forgetting to save his work did not actually cause him to lose
his data. A whole set of links are left out, but they are easily filled in by any listeners
who have the background to do so. It is, says Wodak (1992:525), our schemas that
lead us to interpret Mona Lisa’s smile as evidence of her perplexity or of her
desperation.
Some schemas may be universal. Piaget (1970) studied cognitive schemas in
children that he thought were universal in human development. Some are surely
idiosyncratic (each person speaks a language that is, in some ways, like no other).
Somewhere between universal and idiosyncratic schemas are cultural schemas: They
are developed through experience but are held by a population (Rice 1980:154;
D’Andrade 1995:130).
Rice (1980) developed what she called the American schema for telling a story,
which she contrasted with the Eskimo schema for doing the same thing. Using an
experimental design, she took two Eskimo stories, adjusted them to about 20 phrases
each, and presented one of five, systematically distorted versions to 12 Americans.
She also presented the complete, original Eskimo story (in English) or an
Americanized version of the story to 12 people. Then she asked informants to recall
and write down, in their own words, certain passages from the stories.
When the passages fit the American story schema (as they did in the Americanized
versions of the stories), subjects agreed about which events they remembered.
Furthermore, subjects recalled vastly more exactly worded phrases from the
Americanized versions of the stories than from the Eskimo ones. Thus, people distort
stories in recall to fit their cultural expectations (their schemas) about what stories
ought to be like.
Cognitive scientists, including anthropologists like Rice, often study schemas by
setting up experiments so that they can observe the act of reasoning under uncom-
plicated conditions. Hutchins takes a naturalist’s perspective. “If what we want to
 Text Analysis 605

know about is how people reason in the real world,” he says, “let’s look at them
doing that” (1980:123). Hutchins (1980) recorded and transcribed a formal dispute
in the Trobriands. Two men claim the right to cultivate a particular garden plot.
The antagonists have different views of the facts but, as Hutchins shows, they share
an underlying logic—a schema—for how land claims are to be understood
(1980:128).
Of course, it is not always possible to record people who are reasoning about
important issues in their lives. The intermediate step, between experiments on
reasoning and recording natural discourse on reasoning, is to collect texts. In 1979,
Naomi Quinn and her students collected and transcribed interviews about marriage
from 11 North American couples. Some of the couples were recently married; others
had been married a long time. The couples came from different parts of the country
and represented various occupations, educational levels, and ethnic and religious
groups. Each of the 22 people were interviewed separately for 15–16 hours, and the
interviews were transcribed.
In a series of articles, Quinn (1982, 1987, 1992, 1996, 1997) has analyzed this
body of text to discover and document the concepts underlying American marriage
and to show how these concepts are tied together—how they form a cultural model
shared by people from different backgrounds about what constitutes success and
failure in marriage.
Quinn’s method is to “exploit clues in ordinary discourse for what they tell us
about shared cognition—to glean what people must have in mind in order to say the
things they do” (1997:140). She begins by looking at patterns of speech and the
repetition of key words and phrases, paying particular attention to informants’ use of
metaphors and the commonalities in their reasoning about marriage. For example,
Nan, one of her informants, uses a popular metaphor, that “ marriage is a manu-
factured product”—something that has properties, like strength and staying power,
and that requires work to produce. Some marriages are “put together well,” while
others “fall apart” like so many cars or toys or washing machines (Quinn 1987:174).
Quinn’s emphasis on metaphor owes much to the pioneering work by Lakoff and
Johnson (1980). The object is to look for metaphors in rhetoric and deduce the
schemas, or underlying principles, that might produce patterns in those metaphors.
For instance, Quinn found that people talk about their surprise at the breakup of a
marriage by saying that they thought the couple’s marriage was “like the Rock of
Gibraltar” or that they thought the marriage 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.
But Quinn reasons that if schemas or scripts are what make it possible for people
to fill in around the bare bones of a metaphor, then the metaphors must be surface
phenomena and cannot themselves be the basis for shared understanding. Quinn
found that the hundreds of metaphors in her corpus of texts fit into just eight linked
classes that she calls: lastingness, sharedness, compatibility, mutual benefit, dif-
ficulty, effort, success (or failure), and risk of failure. For example, Quinn’s
606 BERNARD / RYAN

informants often compared marriages (their own and those of others) to manufactured
and durable products (“it was put together pretty good”) and to journeys (“we made
it up as we went along; it was a sort of do-it-yourself project”). Quinn sees these
metaphors, as well as references to marriage as “a lifetime proposition,” as exemplars
of the overall expectation of lastingness in marriage.
The classes of metaphors, the underlying concepts, are linked together in a schema
that guides the discourse of ordinary Americans about marriage:

Marriages are ideally lasting, shared and mutually beneficial. Marriages that are not
shared will not be mutually beneficial and those not mutually beneficial will not last.
Benefit is a matter of fulfillment. Spouses must be compatible in order to be able to fill
each other’s [emotional] needs so that their marriages will be fulfilling and hence
beneficial. Fulfillment and, more specifically, the compatibility it requires, are difficult
to realize but this difficulty can be overcome, and compatibility and fulfillment
achieved, with effort. Lasting marriages in which difficulty has been overcome by effort
are regarded as successful ones. Incompatibility, lack of benefit, and the resulting
marital difficulty, if not overcome, put a marriage at risk of failure. (Quinn 1997:164)

Quinn presents extended excerpts from eight informants to illustrate the relation-
ship between lastingness and success. She hopes that the examples she gives will
familiarize readers with her mode of analysis and “convince them of the pattern
exemplified” in the cases she presents. “Finding this structure,” Quinn says, “was a
methodological challenge” (1997:167).
Other examples of the search for cultural schemas in texts include Holland’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
Claudia Strauss’s (1997) study of what chemical plant workers and their neighbors
think about the free enterprise system.
Examining metaphors and proverbs are not the only linguistic features used to
infer meaning from text. D’Andrade notes that “perhaps the simplest and most direct
indication of schematic organization in naturalistic discourse is the repetition of
associative linkages” (1991:294). He observes that “indeed, anyone who has listened
to long stretches of talk—whether generated by a friend, spouse, workmate, infor-
mant, or patient—knows how frequently people circle through the same network of
ideas” (1991:287).
In a study of blue-collar workers in Rhode Island, Claudia Strauss (1992) refers
to these ideas as “personal semantic networks.” She describes such a network from
one of her informants. On rereading her intensive interviews with one of the workers,
Strauss found that her informant repeatedly referred to ideas associated with greed,
money, businessmen, siblings, and “being different.” She displays the relationships
among these ideas by writing the concepts on a page of paper and interconnected with
lines and explanations.
Price (1987) observes that when people tell stories, they assume that their listeners
share many assumptions about how the world works and so they leave out
 Text Analysis 607

information that “everyone knows.” Thus, in her study of 14 narratives of illness and
misfortune in a Mestizo community in Ecuador, Price looks for what is not said in
order to identify underlying cultural assumptions (1987:314).
If underlying schemas exist, then, with a native speaker’s command of the
language and a deep understanding of one another’s metaphors (about marriage and
so many other things), we can recognize the surface representations of those schemas.
Understanding the complete lexicon of a language, then, makes it possible to do text
analysis. Language competence is nine-tenths of method.
We turn next to the two methods most widely used across the social sciences for
analyzing text: grounded theory and classical content analysis. Grounded theory
emphasizes the discovery and labeling of concepts (variables) and the building of
models based on a close reading of the text. Classic content analysis emphasizes the
formal description of concepts and the testing of models and hypotheses.

Grounded Theory
Grounded theory is a set of techniques that: (1) brings the researcher close to
informants’ experiences; (2) provides a rigorous and detailed method for identifying
categories and concepts that emerge from text; and (3) helps the researcher link the
concepts into substantive and formal theories (Glaser and Strauss 1967; Lincoln and
Guba 1985; Strauss 1987; Lonkila 1995; Charmaz 1990; Strauss and Corbin 1990;
Wilson and Hutchinson 1996). Miles and Huberman (1994) refer to their own brand
of text analysis as “soft-nosed positivism”—a good characterization of most work in
grounded theory as well, in our view.
Grounded theory has been used to examine topics in public health (Hitchcock and
Wilson 1992; Sohier 1993; Kearney et al. 1994, 1995; Irurita 1996; Wright 1997),
social welfare (Silverberg et al. 1996), and business (Hunt and Ropo 1995; Locke
1996). It also has a long history in ethnographic case studies (Becker et al. 1961;
Agar 1979, 1980, 1983). Journals such as Nursing Research, Qualitative Health
Research, and Qualitative Sociology have been outlets for this type of research.
The mechanics of grounded theory are deceptively simple: produce verbatim
transcripts of interviews and read through a small sample of text (usually line by
line). Identify potential themes that arise. As analytic categories emerge, pull all the
data (that is, exemplars) from those categories together and compare them, con-
sidering not only what text belongs in each emerging category but also how the
categories are linked together. Use the relationships among categories to build theo-
retical models, constantly checking the models against the data—particularly against
negative cases. Throughout the process, keep running notes about the coding and
about potential hypotheses and new directions for the research. This is called
“ memoing” in the vocabulary of grounded theory. Grounded theory is an iterative
process by which the analyst becomes more and more “grounded” in the data and
develops increasingly richer concepts and models of how the phenomenon being
studied really works.
608 BERNARD / RYAN

Identifying Themes

Many researchers offer specific advice and schemes for inductive or “open” coding
of text (Taylor and Bogdan 1984; Lincoln and Guba 1985; Strauss and Corbin 1990;
Bogdan and Biklen 1992; Bernard 1994; Lofland and Lofland 1995; Agar 1996).
Sandelowski (1995a:373) observes that analysis of texts begins with proofreading the
material and simply underlining key phrases “because they make some as yet inchoate
sense.” Identifying the categories and terms used by informants themselves is called
“in vivo coding” (Strauss and Corbin 1990). Spradley (1979:199–201) advised
searching texts for evidence of social conflict, cultural contradictions, informal
techniques of social control, methods that people use in managing impersonal social
relationships, the methods by which people acquire and maintain achieved and
ascribed status, and information about how people solve problems. Each of these
arenas, he said, is likely to yield major themes in cultures.
Others suggest that coders start with some general themes derived from reading
the literature and add more themes and subthemes as they go (Willms et al. 1990;
Miles and Huberman 1994). Regardless of which strategy is used, by the time one
identifies the themes and refines them to the point where they can be applied to the
whole text, a lot of the interpretive analysis has been done. Miles and Huberman say
simply: “Coding is analysis” (1994:56).

Building Models

The next step is to identify how themes are linked to each other in a theoretical
model. Memoing is one of the principal techniques for recording relationships among
themes. When reviewing the text, you continually write down your thoughts about
what you’re reading. These thoughts become a set of information on which to develop
theory. We think of memoing as taking field notes on observations about texts.
Strauss and Corbin discuss three kinds of memos: code notes, theory notes, and
operational notes (1990:18, 73–74, 109–129, 197–219). Code notes describe the
concepts that are being discovered in “the discovery of grounded theory.” In theory
notes, the researcher tries to summarize his or her ideas about what’s going on in the
text. Operational notes are about practical matters.
Once a model starts to take shape, researchers specifically look for negative
examples that do not fit the pattern. Negative case analysis is discussed in detail by
Becker et al. (1961:37–45), Strauss and Corbin (1990:108–109), Lincoln and Guba
(1985:309–313), Dey (1993:226–233), and Miles and Huberman (1994:271). Nega-
tive cases either disconfirm parts of the model or suggest new connections that need
to be made. In either case, these negative examples need to be accommodated when
results are presented.
When the steps of the grounded theory approach are followed, models or
“theories” are produced that are, indeed, “grounded” in the text. These models,
however, are not the final product of the grounded theory approach. In their
 Text Analysis 609

original formulation, Glaser and Strauss (1967) emphasized that the building of
grounded theory models is a step in the research process. The next, of course, is
to confirm the validity of a model by testing it on an independent sample of data.
The grounded theory approach, including iterative coding and analysis by constant
memoing, has been the inspiration for several of the most widely used software
packages in text analysis, including Atlas/ti (Muhr 1991), NUD*IST (Richards and
Richards 1991), and Kwalitan (Peters and West 1990). In fact, 17 of the 24 text
analysis packages reviewed by Weitzman and Miles (1995:316–325) have some
provision for writing memos on the fly and retrieving them during analysis.5

Displaying Concepts and Models

Much of grounded theory involves presenting segments of text—verbatim quotes

from informants—as exemplars of concepts and theories. These illustrations may be
prototypical examples of central tendency or they may represent exceptions to the
norm. Grounded theory researchers may display their theoretical results in maps of
the major categories and the relationships among them (Miles and Huberman
1994:134–137, Kearney et al. 1995). These “concept maps” are similar to the
personal semantic networks described by C. Strauss (1992) and D’Andrade (1991)
(see below).

An Example of Grounded Theory

Kearney et al. (1995) interviewed 60 women who reported using crack cocaine an
average of at least once weekly during pregnancy. The semistructured interviews
lasted from one–three hours and covered childhood, relationships, life context,
previous pregnancies, and actions during the current pregnancy related to drug use,
prenatal care, and self-care. Transcripts were coded and analyzed as soon as they
became available so that data collection and data analysis were intricately linked. As
new topics emerged, investigators asked about the topics in subsequent interviews.
Kearney et al. coded the data first for the general topics they used to guide the
interviews. Later, they would use these codes to search for and retrieve examples of
text related to various interview topics. Next, team members reread each transcript
searching for examples of social psychological themes in the women’s narratives.
Each time they found an example, they considered “ What is this an example of?”.
The answers suggested substantive categories that were refined with each new
transcript.
Kearney et al. (1995) looked at how substantive categories were related. They
recorded their ideas about these interactions in the forms of memos and developed a
preliminary model. With each subsequent transcript, they looked for negative cases
and pieces of data that challenged their emerging model. They adjusted the model to
include the full range of variation that emerged in the transcripts.
To begin with, Kearney et al. identified five major categories, which they
called: VALUE, HOPE, RISK, HARM REDUCTION, and STIGMA MAN-
610 BERNARD / RYAN

AGEMENT. (Capital letters are often for code names in grounded theory research,
just as in statistical research.) Women valued their pregnancy and the baby-to-be
in relation to their own life priorities (VALUE); women expressed varying degrees
of hope that their pregnancies would end well and that they could be good
mothers (HOPE) and they were aware that cocaine use posed risks to their fetus
but they perceived that risk differently (RISK). Women tried in various ways to
minimize the risk to the fetus (HARM REDUCTION) and they used various
stratagems to reduce social rejection and derision (STIGMA MANAGEMENT).
By the time they had coded 20 interviews, Kearney et al. realized that the
categories HARM REDUCTION and STIGMA MANAGEMENT were components
of a more fundamental category that they labeled EVADING HARM. After about 30
interviews had been coded, they identified and labeled an overarching psychological
process they called SALVAGING SELF that incorporated all five of the major
categories. “Theoretical saturation” was reached at approximately 40 interviews and
Kearney et al. conducted another 20 without discovering any new categories or
relationships.
Kearney et al. (1995) present their model graphically with ties to supporting
textual evidence. They describe in rich detail each of the major categories that they
discovered. Finally, they checked the validity of their model by presenting it to
knowledgeable informants (pregnant drug users), members of the project staff, and
health and social service professionals who were familiar with the population under
study.

Classical Content Analysis

While grounded theory is concerned with the discovery of data-induced hypotheses,
content analysis comprises techniques for (1) reducing the symbol-laden artifacts
produced by human behavior (including, but not limited to texts) to a unit-by-variable
matrix and (2) analyzing that matrix quantitatively in order to test hypotheses. The
matrix is produced by applying a set of codes to artifacts and checking the reliability
of human coders against one another. When the artifacts are texts, the coding may be
done by a computer, using a content analysis dictionary. In addition to written text,
the symbol-laden artifacts of human effort include television sit-coms, political
cartoons, advertisements, song lyrics, and clay pots. The object is to test hypotheses
about the producers of the symbolic artifacts, the consumers, or both.6
Cowan and O’Brien (1990), for example, tested whether males or females were
more likely to be survivors in slasher films. The corpus of “text” in this case was 56
slasher films. The films contained a total of 474 victims, who were coded for gender
and survival. Conventional wisdom about slasher films holds that victims are mostly
women and slashers are mostly men. While slashers in these films were, in fact,
mostly men, it turned out that victims were equally likely to be male or female.
Surviving as a female slasher victim, however, was strongly associated with the
absence of sexual behavior and with being less physically attractive than
 Text Analysis 611

nonsurviving females. The male nonsurvivors were cynical, egotistical, and dic-
tatorial. Cowan and O’Brien conclude that, in slasher films, sexually pure women
survive and that “unmitigated masculinity” ends in death (1990:195).
The methodological issues associated with content analysis are all evident here.
Does the sample of 56 films used by Cowan and O’Brien justify generalizing to
slasher films in general? Did the coders who worked on the project make correct
judgments in deciding things like the physical attractiveness of female victims or the
personality and behavioral characteristics of the male victims? These two issues in
particular, sampling and coding, are at the heart of content analysis.

Sampling

There are two components to sampling. The first is the identification of the corpus
of texts; the second is the identification of the units of analysis within the texts. If one
collects 40 or 50 life histories, then the entire set of texts is analyzed. When the units
of data run into the hundreds or even thousands (i.e., all television commercials that
ran during prime time in August 1997, all front-page stories of The New York Times
from 1887–1996, all campaign speeches by Bill Clinton and George Bush during the
1996 presidential campaign), then a representative sample of records must be made.
Gilly (1988) did a cross-cultural study of gender roles in advertising. She vide-
otaped a sample of 12 hours of programming in Los Angeles (U.S.), Monterrey
(Mexico), and Brisbane (Australia), from 8 a.m. to 4 p.m. on Tuesday and 7 p.m. to
11 p.m. on Wednesday. To control for seasonal variation between the hemispheres,
the U.S. and Mexico samples were taken in September 1984, while the Australia
sample was taken in February 1985. There were 617 commercials: 275 from the U.S.,
204 from Mexico, and 138 from Australia. Because of her research question, Gilly
used only adult men and women who were on camera for at least three seconds or
who had at least one line of dialogue. There were 169 women and 132 men in the
U.S. ads; 120 women and 102 men in the Mexican ads; and 52 women and 49 men
in the Australian ads.
Cohen (1990) wanted to know whether the unpredictability of the environment
(floods, drought, etc.) would be reflected in a society’s folktales. He selected a sample
of 19 societies using two criteria: (1) two independent coders had to agree on the
presence of three variables about resources in a cross-cultural study of warfare done
by Ember and Ember (1992); and (2) there had to be at least 9 folktales for each
society in the Human Relations Area Files (as of 1988). Cohen then selected up to 10
folktales from each society. If there were more than that, he numbered them and
selected them randomly. In all, Cohen had 187 different folktales to code and
analyze.
Waitzkin and Britt (1993:1121) did an interpretive analysis of 50 encounters
between patients and doctors by randomly selecting texts from 336 audiotaped
encounters. Nonquantitative text analysis is often based on purposive sampling. Trost
(1986) thought the relationship between teenagers and their families might be
612 BERNARD / RYAN

affected by five different dichotomous variables. To test this idea, he intentionally

selected five cases from each of the 32 possible combinations of the five variables and
conducted 160 interviews.
Nonquantitative studies in content analysis may also be based on extreme or
deviant cases, cases that illustrate maximum variety on variables, cases that are
somehow typical of a phenomenon, or cases that confirm or disconfirm a hypothesis.
A single case may be sufficient to display something of substantive importance, but
Morse (1994) suggests using at least six participants in studies in which one is trying
to understand the essence of experience. Morse also suggests 30–50 interviews for
ethnographies and grounded theory studies. Patton (1990: 169–186) and Sandelowski
(1995b) provide useful reviews of the nonrandom strategies for sampling texts.
Finding themes and building theory may require fewer cases than comparing across
groups and testing hypotheses or models.
Once a sample of texts is established, the next step (called “unitizing”
[Krippendorf 1980] or “segmenting” [Tesch 1990]) is to identify the basic, nonover-
lapping units of analysis. The units may be the entire texts (books, interviews,
responses to an open-ended question on a survey) or segments (words, word-senses,
sentences, themes, paragraphs). Where the object is to compare across texts—to see
whether or not certain themes occur—the whole text (representing an informant or
an organization) is the appropriate unit of analysis. When the object is to compare the
number of times a theme occurs across a set of texts, then what Kortendick (1996)
calls a context unit—a chunk of text that reflects a theme—is likely to be the
appropriate unit of analysis.
With a set of texts in hand, the next steps are to develop a codebook and mark
(actually code) the text. This is the heart and soul of sociological text analysis,
whether it is schema analysis, grounded theory, or classic content analysis. How to
develop a codebook is covered in detail by Dey (1993:95–151), Crabtree and Miller
(1992), Miles and Huberman (1994:55–72). Richards and Richards (1991) discuss
the theoretical principles related to hierarchical coding structures that emerge out of
the data. Araujo (1995) uses an example from his own research on the traditional
British manufacturing industry to describe the process of designing and refining
hierarchical codes. The development and refinement of coding categories has long
been a central task in classical content analysis (for example, Bereleson 1952:
147–168; Holsti 1969:95–126) and is particularly important in the construction of
concept dictionaries (Stone et al. 1966:134–168; Deese 1969). Krippendorff (1980:
71–84) and Carey et al. (1996) note that much of codebook refinement comes during
the training of coders to mark the text and in the act of checking for intercoder
agreement.
The word “code” has two quite different meanings. Codes are sometimes simply
tags that mark off sections of text in a corpus and sometimes are values of a variable.
When codes are used as tags, they are reference markers, like an index in the back of
a book. When they are used as values of a variable, they identify nominal, ordinal, or
even ratio characteristics of episodes, cases or persons (Bernard 1991, 1994:193–194;
Seidel and Kelle 1995). Below we look more closely at these distinctions.
 Text Analysis 613

Inductive Coding

In hypothesis-testing content analysis, codes are typically formulated first and

the codebook is then tested for reliability, using multiple coders and modified if
necessary. In exploratory content analysis, inductive coding (open coding in
grounded theory terms) is appropriate. Kortendick (1996), for example, interviewed
43 members of a three-generation extended family in Holland, all of whom were
related to three sisters (the first generation) who had immigrated from Indonesia.
Some of the interviews were done with individuals, others with groups of several
people. After reading eight interviews carefully, Kortendick found that he was no
longer identifying any new themes and he constructed a codebook.
Kortendick used a numerical coding scheme. (Miles and Huberman [1994:65]
warn against this, but it’s really a matter of taste—whatever works for individual
researchers.) Kortendick applied his coding scheme to all the interview texts,
tagging each text for things like the level of competence in Dutch versus
Indonesian that people evinced, whether they mentioned any of 21 family myths
that he had identified, and whether each person felt fully Dutch, fully Indonesian,
or ambivalent about their identity. Kortendick built a matrix of informants-by-
themes so he could count and compare references to these myths across all 43
informants.
Ryan and Weisner (1996) asked fathers and mothers to describe their adolescent
children. Ryan and Weisner generated a list of the unique words in the corpus and the
number of times each word was used. They used the list to look for differences
between fathers and mothers and to look for themes that might be coded in a content
analysis. Mothers, for example, were more likely to use words like friends, creative,
time, and honest; fathers were more likely to use words like school, good, lack,
student, enjoys, independent, and extremely.

Deductive Coding

Deductive coding is appropriate in confirmatory research. From her reading of

literature on the theory of resources, Hirschman (1987) thought that she would find
ten kinds of resources in personal ads: love, physical characteristics, educational
status, intellectual status, occupational status, entertainment services (nonsexual),
money status, demographic information (age, marital status, residence), ethnic char-
acteristics, and personality info (not including sexual or emotional characteristics).
Hirschman formulated and tested hypotheses about which resources men and women
would offer and seek in personal ads.
She selected 20 test ads at random from the New York Magazine and The
Washingtonian and checked that the ten kinds of resources were, in fact, observable
in the ads. Sexual traits and services were less than 1% of all resources coded. This
was 1983–84, but even then, ads with explicit references to sexual traits and services
were more common in other periodicals than in The Washingtonian and New York
Magazine.
Hirschman next gave 10 men and 11 women the list of resource categories and a
614 BERNARD / RYAN

list of 100 actual resources (“young,” “attractive,” “fun loving,” “divorced,” “32-year-
old,” etc.) gleaned from the 20 test ads. The respondents were asked to match the 100
resources with the resource category that seemed most appropriate. This exercise
demonstrated that the resource items were mutually exclusive and exhaustive: No
resource items were left over and all could be categorized into just one of the ten
resource categories.
When she was confident her codebook worked, Hirschman tested her hypotheses.
She sampled approximately 100 female-placed ads and 100 male-placed ads from
each magazine. A male and a female coder, working independently (and unaware of
the hypotheses of the study) coded 3,782 resource items taken from the 400 ads as
belonging to one of the ten resource categories. The coding took three weeks. This is
not easy work.
A third coder was given the data and identified discrepancies between the first
two coders. Of 3,782 resource items coded, there were theme contrasts on 636
(16.8%) and one of the coders failed to code 480 items (12.7%). The theme contrasts
were resolved by Hirschman. The omissions were checked against the ads to see if,
in fact, the one coder who had made an assignment had done so because the resource
was in the ad. This was always the case, so the 480 resource items omitted by one
coder were counted as if they had been assigned to the ad by both coders.
Hirschman found that men were more likely than women to offer monetary
resources, whereas women were more likely than men to seek monetary resources.
Women were more likely than men to offer physical attractiveness. It would be very
interesting to repeat this study today with the same magazines. After all, Washington,
DC and New York City are supposed to be hip places. Are the stereotypes of how men
and women market themselves to one another today very different from what they
were in 1983–84?
Confirmatory hypothesis testing and deductive coding are also used by
anthropologists and other scholars who examine the Human Relations Area Files.
The “codes” in the Outline of Cultural Materials are tags, however, not measure-
ments. “Tagging” a paragraph in a text on the Yanomamo with “warfare” indicates
textual material on the topic. It says nothing about how much the Yanomamo engage
in warfare, or how intense their battles are, or whether Yanomamo men are at high
or low risk of dying in battle. These kinds of measurements require a close reading
of the segments of text that deal with warfare. Absent any prior knowledge about the
Yanomamo, the tag “warfare” could just as well indicate that the Yanomamo are a
peaceful people who despise and preach actively against violence. [Methodological
issues associated with cross-cultural research are dealt with at length in Chapter 17
of this volume. Ed.]

Intercoder Agreement

The marking of text often involves multiple coders. The idea is to see whether the
constructs being investigated are shared and whether multiple coders see the same
 Text Analysis 615

constructs as applying to the same chunks of text. Carey et al. (1996) asked 51 newly
arrived Vietnamese refugees in New York State 32 open-ended questions about
tuberculosis. Topics included knowledge and beliefs about TB symptoms and causes,
as well as beliefs about susceptibility to the disease, prognosis for those who contract
the disease, skin testing procedures, and prevention and treatment methods. The
investigators read the responses and created a code list based simply on their
judgment. The initial code book contained 171 codes.
Next, the researchers broke the text into 1,632 segments (each segment repre-
senting a single informant’s response to each of the 32 questions) and two coders
independently coded 320 segments. Text segments could be marked with multiple
codes. Segments were counted as reliably coded if both coders used the same codes
on it. If one coder left off a code or assigned an additional code, then this was
considered a coding disagreement. On their first try, only 144 (45%) out of 320
responses were coded the same by both coders. The coders discussed their
disagreements and found that some of the 171 codes were redundant, some were
vaguely defined, and some were not mutually exclusive. In some cases, coders simply
had different understandings of what a code meant. When these problems were
resolved, a new, streamlined codebook was issued and the coders marked up the data
again. This time they were in agreement 88.1% of the time.
This seems like high reliability, but analysts typically apply a correction formula
to take account of the fact that some fraction of agreement will always occur by
chance. Cohen’s Kappa, or K is a popular measure for taking these chances into
account (Jacob Cohen 1960). When K is zero, agreement is what might be expected
by chance.7 When K is negative, the observed level of agreement is less than one
expects by chance. Of the 152 codes in the new code list that had been applied to the
320 sample segments, the coders agreed perfectly (kappa = 1.0) for 82.9% of the
codes. Only 17 (11.2%) of the codes had final K values # 0.89. As senior investigator,
Carey resolved any remaining intercoder discrepancies himself (Carey et al. 1996).
How much intercoder agreement is enough? The standards are still evolving, but
Krippendorf (1980:147–148) advocates agreement of at least .70 and notes that some
scholars (Brouwer et al. [1969] use a cutoff of .80. In developing software to create
psychological scales from texts, Gottschalk and Bechtel (1993) ensured that the
reliability of the computer scores were greater than .80 when compared to human
coders. Including several measures of intercoder agreement in popular text analysis
packages would encourage much needed research on this issue.8

Dictionaries

Computer-based, general-purpose content analysis dictionaries are a kind of auto-

mated codebook. To build such dictionaries, words are assigned, by hand, to one or
more categories (there are typically 60–50 categories in computerized content
analysis dictionaries), according to a set of rules. The rules are part of a computer
program that parses new texts, assigning words to categories.
616 BERNARD / RYAN

Work began in the 1960s on the best-known system, the General Inquirer and
continues to this day (Stone et al. 1966; Kelly and Stone 1975; Zuell et al. 1989). The
system comprises a computer program (the General Inquirer), which uses a
dictionary (the Harvard Psychosocial Dictionary) in parsing text and categorizing
text. An early version of the system was tested on 66 suicide notes—33 written by
men who had actually taken their own lives, and 33 written by men who were asked
to produce simulated suicide notes. The control group men were matched with the
men who had written actual suicide notes on age, occupation, religion, and ethnicity.
The General Inquirer program parsed the texts and picked the actual suicide notes
91% of the time (Ogilvie et al. 1966).
The 1975 update of the Harvard Psychosocial Dictionary, called Harvard IV,
initially contained about 4,000 entries (Kelly and Stone 1975:47), and the Dartmouth
adaptation of Harvard IV now contains about 8,500 (Rosenberg et al. 1990). The next
version will contain around 10,000 (Philip Stone, personal communication).9 Unlike
its predecessors, the Harvard IV dictionaries can distinguish among multiple
meanings of many words. If the program runs into the word “broke,” for example, it
looks at the context and determines whether the meaning is “fractured,” or
“destitute,” or “stopped functioning,” or (when the word is paired with “out”)
“escaped” (Rosenberg et al. 1990:303). Of course, dictionaries do not include all the
words in a text, so investigators must still look at words that are not tagged and
decide how to tag them independently.10
Content dictionaries do not need to be very big to be useful. In his study of
Navaho and Zuni responses to thematic apperception tests, Colby’s (1966) initial
impression was that the Navajo regarded their homes as havens and places of
relaxation, whereas the Zuni depicted their homes as places of discord and tension.
To test this idea, Colby created a special-purpose dictionary that contained two
word groups that he and his colleagues had developed before looking at the data.
One word group, the “relaxation” group, comprised the words assist, comfort,
easy, affection, happy, and play. The other, the “tension” group, comprised the words
destruction, discomfort, difficult, dislike, sad, battle and anger. Colby examined the
35 sentences that contained the word “home” and one of the words in either of the
two word groups. The Navajos were more than twice as likely to use words from the
relaxation group when they were talking about home as they were to use words from
the tension group. The Zuni were almost twice as likely to use tension words as they
were to use relaxation words.
Colby (1966) also found that the Navajo were more likely to use words associated
with exposure such as storm, cold, freezing, hot, heat, and windy. Colby was not
surprised at the results; he noted that the Navajo were sheep herders and were
concerned about protecting their sheep from the elements, whereas the Zuni were
crop growers and were concerned about the water they need to grow their corn. What
was surprising was that the texts were generated from pictures that had nothing to do
with sheep or crops. (See Jehn and Werner [1993] and Furbee [1996] for other
examples of the application of a special-purpose content dictionary.)11
Content dictionaries are attractive because they are entirely reliable but, as Shapiro
 Text Analysis 617

(1997) argues, this may be offset by a decrease in validity. In many cases, only
humans can parse the subtleties of meaning reflected in context (Viney 1983), but
dictionary-based markup of text is producing better and better results as time goes on,
particularly in well-defined domains. For example, texts are now scored by computer
for the Gottschalk-Gleser psychological scales (measuring various forms of anxiety
and hostility) with greater than .80 reliability (Gottschalk and Bechtel 1993).12 We
expect to see increasing use of computer-based dictionaries in text analysis.
For additional examples of special-purpose dictionaries in content analysis, see
Fan and Shaffer (1990), Holsti (1966), Laffal (1990, 1995), McTavish and Pirro
(1990), and Schnurr et al. (1986).

Combining Grounded Theory and

Content Analysis Approaches
Techniques from grounded theory and content analysis can be fruitfully combined.
Jehn and Doucet (1996, 1997) asked 76 U.S. managers who had worked in Sino-
American joint ventures to describe recent interpersonal conflicts with business
partners. Each person described a situation with a same-culture manager and a
different-culture manger. They made sure that the manager included the relationship
to the person, who was involved, what the conflict was about, what caused the
conflict, and how the conflict was resolved.
Jehn and Doucet used three quite different methods to identify themes in these
data: (1) word counts combined with consensual coding by expert judges; (2) factor
analysis of a traditional content-analysis (text unit-by-variable) matrix; and (3)
multidimensional scaling of similarity judgments by experts of the scenarios.

1. Word counts and consensual coding. There were 7,479 unique words in the
corpus on intercultural conflicts compared to 2,747 in the corpus on the
intracultural conflicts. Jehn and Doucet wanted to identify words that were
related to conflict, but did not want to impose their own definitions. They asked
three expatriate managers who had experiences similar to those of their 76
informants and who were blind to the conditions of the study to go through both
word lists and select the words that seemed related to conflict. The three judges
went over the lists together and settled on a list of 542 conflict words from the
intercultural list and 242 words from the intracultural list. Two out of three
judges had to agree before a word was put on the final lists.
In an open discussion to reach consensus, the judges assigned each conflict
word to a category. The judges developed 15 subcategories for the intercultural
data—things like conflict, expectations, rules, power, and volatile—and 15
categories for the intracultural data—things like conflict, needs, standards,
power, contentious, and lose. Taking into consideration the total number of
words in each corpus, conflict words were used more in intracultural interviews
and resolution terms were more likely to be used in intercultural interviews.
2. Traditional content analysis. Next, two coders read the 152 conflict scenarios
(76 intracultural and 76 intercultural) and evaluated (on a 5-point scale) each on
618 BERNARD / RYAN

27 different themes. This produced two 76x27 scenario-by-theme profile

matrices—one for the intracultural conflicts and one for the intercultural
conflicts. The first three factors from the intercultural matrix reflect: (1)
interpersonal animosity and hostility; (2) aggravation; 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.
3. Scenario comparisons. Finally, Jehn and Doucet identified the 30 intracultural
and the 30 intercultural scenarios that they felt were the most clear and pithy.
Fifty more expatriate managers assessed the similarities (on a 5-point scale) of
60–120 randomly selected pairs of scenarios (each pair of scenarios was seen and
judged by eight respondents) and described the basis for their judgments. When
combined across informants, this produced two aggregate, scenario-by-scenario,
similarity matrices—one for the intracultural conflicts and one for the
intercultural conflicts. Multidimensional scaling of the intercultural similarity
data identified four dimensions: (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. Scaling of the intracultural similarity data 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 research by Jehn and Doucet illustrates the rich combination of qualitative and
quantitative methods now available for text analysis. Jehn and Doucet collected
narratives from their informants and asked their informants to help identify the emic
themes in the narratives. Informants sorted key words, coded each scenario for
potential themes, and compared scenarios to each other. The analysis of the data from
these tasks produced different sets of themes. All three emically induced theme sets
have some intuitive appeal and all three yield analytic results that are useful.
In a series of articles on young adult “occasional” drug users, Agar (1979, 1980,
1983) described his grounded methods for content analysis. Agar conducted and
transcribed three interviews with each of his three informants. In the 1979 article,
Agar describes his initial, intuitive analysis. He pulled all the statements that
pertained to informants’ interactions or assessments of other people. He then looked
at the statements and sorted them into piles based on their content. He named each
pile as a theme and assessed how the themes interacted. Agar found that he had three
piles. The first contained statements where the informant was expressing negative
feelings for a person in a dominant social position. The second pile emphasized the
other’s knowledge or awareness. The third small cluster emphasized the importance
of change or an openness to new experiences.
From this intuitive analysis, Agar felt that his informants were telling him that
those in authority were only interested in displaying their authority unless they had
knowledge or awareness; knowledge or awareness comes through openness to new
experience; most in authority are close to new experience or change.
In his second article (1983), Agar systematically tested his intuitive understanding
of the data. He used all the statements from a single informant and coded the
 Text Analysis 619

statements for their role type (kin, friend/acquaintance, educational, occupational, or

other), power (dominant, symmetrical, subordinate, or undetermined), and affect
(positive, negative, ambivalent, or absent). Agar realized that he could analyze the
covariations among role type, power, and affect, or he could examine the distribution
of the themes as they occur throughout the text, or he could simply count the number
of statements in the different categories.
Agar restricted his analysis to check his earlier primary finding: that for a given
informant, a particular negative sentiment is expressed toward those in dominant
social roles. He found that out of 40 statements coded as dominant, 32 were coded
negative and 8 were coded positive. For the 36 statements coded as symmetrical, 20
were coded positive and 16 negative, lending support to his original theory.
Next, Agar looked closely at the deviant cases—the eight statements where the
informant expressed positive affect toward a person in a dominant role. These
counterexamples suggested that the positive affect was expressed toward a dominant
social other when the social other possessed, or was communicating to the informant,
knowledge that the informant valued.
Finally, Agar (1980) developed a more systematic questionnaire to further test
his hypothesis with an independent set of data for one of his informants. He
selected 12 statements, four from each of the control, knowledge, and change
themes identified earlier. Some statements came directly from the informant.
Others he made up. Then he selected eight roles from the informant’s transcript
(father, mother, employer, teacher, friend, wife, co-worker, and teammate). Each
role term was matched with each statement and the informant was asked if the
resulting statement was true, false, or irrelevant. (In no case, did the informant
report “irrelevant.”) Agar then took the responses and compared them to what his
hypotheses suggested. The results both met and did not meet his expectations. On
balance, there seemed to be general support for his hypothesis, but discrepancies
between Agar’s expectations and his results suggested areas for further research.
Increasingly, text analysis is about extracting models or schemas from
transcriptions of human discourse. In the next section, we review some of the main
currents: semantic network analysis, cognitive mapping analysis, Boolean analysis,
and schema analysis.

Structural Analysis and Semantic Networks

The fundamental point of structuralism is that abstractions, called “structures,” are
in some important way responsible for, or govern, human action. Modern network
analysis is based on this principle. Traditional sociological analysis involves a profile
matrix—a matrix of things (rows) characterized by values on variables (columns) that
characterize the things. Structural analysis, or network analysis, examines the
properties that emerge from relations among a set of things. Here the matrix is one
of similarities, where rows and columns represent the same things.
The relations among even a small a set of things (the relations among, say, the 30
620 BERNARD / RYAN

households in a village, for example, or the 15 countries of the European Union) can
be wildly complex. Seeing patterns, if they exist, in such complexity is made possible
by applying graph-theoretic methods that enable the visualization of structure through
the messiness of surface reality. Hage and Harary (1983) provide other examples of
structuralist analysis in anthropology.
The application of graph-theoretic principles and methods to the study of meaning
in text is sometimes called semantic network analysis. As early as 1959, Charles
Osgood (1959) created word co-occurrence matrices and applied factor analysis and
dimensional plotting to describe the relation of major factors to one another. The
development of computers has made the construction and analysis of co-occurrence
matrices much easier and has stimulated the development of this field (Danowski
1982, 1993; Barnett and Danowski 1992).
Jang and Barnett (1994) examined whether there was a national culture—U.S. or
Japanese—discernible in the annual letters to stockholders of CEOs in U.S. and
Japanese corporations. Jang and Barnett selected 35 Fortune 500 companies,
including 18 U.S. and 17 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 York Stock Exchange, and each year
stockholders receive an annual message from the CEO or president of these com-
panies. (Japanese firms that trade on the New York Exchange send the annual letters
in English to their U.S. stockholders.) Jang and Barnett read through the 1992 annual
letters to shareholders and (ignoring a list of common words like “the,” “because,”
“if,” and so on) isolated 94 words that occurred at least eight times across the corpus
of 35 letters. This produced a 94(word)-by-35(company) matrix, where the cells
contained a number from 0–25, 25 being the largest number of times any word ever
occurred in one of the letters.
Next, Jang and Barnett created a 35(company)-by-35(company) similarity matrix
of companies, based on the co-occurrence of words in their letters. They analyzed the
matrix by multidimensional scaling. Figure 1 shows the result—a twodimensional
plot of similarities between companies.
 Text Analysis 621

Figure 1. Multidimensional scaling, in two dimensions, of Jang’s word co-occurrence matrix

for 35 Japanese and American companies. J=Japanese company. A=American company.
(Adapted from Jang and Barnett [1994:48]. Reprinted with permission of Bulletin de
Méthodologie Sociologique.)

It’s clear that there are two styles of corporate reporting to stockholders, one
American and one Japanese. From a close reading of the texts, Jang concludes that
U.S. executives discuss financial information and the structure of their organizations
in their letters and Japanese executives focus more on organizational operations.
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.
See Schnegg and Bernard [1996] for another example of their use.) There is,
however, no guarantee that the output of any word co-occurrence matrix will be
meaningful, and it is notoriously easy to read pattern (and, thus, meaning) into any
set of items.

Cognitive Maps
Cognitive map analysis combines the intuition of human coders with the quantitative
methods of network analysis. Carley’s work is instructive. If mental models, or
schemas, are in there, she says, they are expressed in the texts of people’s speech and
can be represented as networks of concepts (Carley and Palmquist 1992:602), an
approach also suggested by D’Andrade (1991). To the extent that mental models are
widely shared, she asserts, even a very small set of texts will contain the information
required for describing the models, especially for narrowly defined arenas of life.
In one study, Carley (1993) asked students some questions about the work of
scientists. Here are two brief texts that address questions about the motivation of
scientists and their collaboration with colleagues:

Student A: I found that scientists engage in research in order to make discoveries and
generate new ideas. Such research by scientists is hard work and often involves
collaboration with other scientists which leads to discoveries which make the scientists
famous. Such collaboration may be informal, such as when they share new ideas over
lunch, or formal, such as when they are coauthors of a paper.

Student B: It was hard work to research famous scientists engaged in collaboration and
I made many informal discoveries. My research showed that scientists engaged in
collaboration with other scientists are coauthors of at least one paper containing their
new ideas. Some scientists make formal discoveries and have new ideas. (Carley
1993:89)
622 BERNARD / RYAN

Carley compares these texts (one 64 words long, the other 48 words) by counting
11 concepts: I, scientists, research, hard work, collaboration, discoveries, new ideas,
formal, informal, coauthors, paper. Each concept occurs exactly the same number of
times in both texts, yet the texts clearly have different meanings. To analyze the
differences in meaning, Carley produces maps of the relation between and among
concepts. Concepts are coded for their strength, sign (positive or negative), and
direction (whether one concept is logically prior to others), not just for their
existence. Figure 2 shows Carley’s maps of the two texts.

Figure 2. Coded maps of two student’s texts (from Carley [1993:104]. Reprinted with the
permission of the American Sociological Association).

This approach to text analysis holds a lot of promise, combining, as it does, the
sensitivity of human intuition and interpretation with the labor-saving characteristics
of automation. As Carley recognizes, though, a lot depends on who does the
 Text Analysis 623

coding—just as in classical content analysis. Different coders (with more or less

knowledge of the implicit culture lurking in a text) will produce different maps by
making different coding choices. In the end, as with the search for schemas,
competency in the native language is one of the fundamental methodological
requirements for analysis (see also Carley and Palmquist 1992; Carley 1993, 1997;
Carley and Kaufer 1993; Palmquist et al. 1997).

Analytic Induction and Boolean Tests

Analytic induction is a formal, nonquantitative method for building up causal
explanations of phenomena from a close examination of cases. It was proposed as
an alternative to statistical analysis by Znaniecki (1934:249–331), and is discussed
by Denzin (1978) and Manning (1982), among others. The method involves the
following steps: First, define a phenomenon that requires explanation and propose
an explanation. Examine a case to see if the explanation fits. If it does, then
examine another case. An explanation is accepted until a new case falsifies it.
When a case is found that doesn’t fit, then, under the rules of analytic induction,
the alternatives are to change the explanation to include the new case or redefine the
phenomenon to exclude the nuisance case. Ideally, the process continues until a
universal explanation for all known cases of a phenomenon is attained. Explaining
cases by declaring them all unique is not an option of the method. Classic examples
of analytic induction include Lindesmith’s (1947) study of drug addicts, Cressey’s
(1953) study of embezzlers, and McCleary’s (1978) study of how parole officers
decide when one of their charges is in violation of parole.
Ragin (1987, 1994) formalized the logic of analytic induction, offering a Boolean
approach, and Romme (1995) applies the approach to textual data. Boolean algebra
involves two states, true and false (present and absent). With just three dichotomous
causal conditions (A and not A, B and not B, and C and not C) and one dependent
variable (D and not D), there are 16 possible outcomes (A, B, C, D; A, not B, C, D;
A, B, not C, D; and so on).
Schweizer (1991, 1996) applied Boolean logic in his analysis of conflict and social
status in Chen Village, China. In the 1950s, the village began to prosper with the
application of technology to agriculture. The Great Leap Forward and the Cultural
Revolution of the 1960s, however, reversed the village’s fortunes. Chan et al. (1984)
reconstructed the recent history of Chen Village, focusing on the political fortunes of
key actors there.
Schweizer coded the Chan et al. text for whether each of 13 actors experienced an
increase or a decrease in status after each of 14 events (such as the Great Leap
Forward, land reform and collectivization, the collapse of Brigade leadership, and an
event known locally as “the great betrothal dispute”). Thus, he has a 13-actor-by-14-
event matrix, where a 1 means that an actor had success in a particular event and a
0 means failure (loss of status). When Schweizer looked at the actor-by-event matrix
he found that, over time, 9 of the actors consistently won or lost, but 4 of the actors
lost sometimes and won other times. This produces 17 unique combinations of actors
624 BERNARD / RYAN

and outcomes.
Schweizer then partitioned the 17 unique cases according to three binary
independent variables (urban versus rural origin, proletarian versus nonproletarian
background, presence versus absence of external ties) and one dependent variable
(whether the actor was a success overall). Table 1 shows the 16 possible outcomes,
given four binary variables.
By setting up the logical possibilities, Schweizer was able to discern and test
several hypotheses about success and failure in Chen Village. People from an urban
background have an advantage, but inspection of Table 1 shows that it’s not enough.
To ensure success, you should come from a proletarian family or have good external
ties and access to information and power at the regional level. Failure is predicted
even better: If an actor has failed in the Chen Village disputes, then he or she is of
rural origin (comes from the village) OR comes from a nonproletarian family AND
has no ties to authorities beyond the village. The Boolean formula for this statement
is: Lack of success 6 nonurban v (nonproletarian & lack of ties).13 The substantive
conclusions from this analysis are intuitively appealing: In a communist revolutionary
environment, it pays over the years to have friends in high places; people from urban
areas are more likely to have those ties; and it helps to have been born into a
politically correct (that is, proletarian) family.

TABLE 1
The Outcome of 17 Cases from Schweizer’s (1996) Text Analysis
Success External Proletarian Urban No. of
Ties Background Origin Cases
0 0 0 0 2
0 0 0 1 2
0 0 1 0 1
0 0 1 1 0
0 1 0 0 0
0 1 0 1 0
0 1 1 0 2
0 1 1 1 0
1 0 0 0 1
1 0 0 1 3
1 0 1 0 0
1 0 1 1 0
1 1 0 0 0
1 1 0 1 1
1 1 1 0 4
1 1 1 1 1

Analytic induction helps identify the simplest model that logically explains the
data. Like classic content analysis and cognitive mapping, it requires that human
coders read and code the text into an event-by-variable matrix. The object of the
analysis, however, is not to show the relationships between all codes, but to find the
minimal set of logical relationships among the concepts that accounts for a single
 Text Analysis 625

dependent variable. With three binary independent variables, as in Schweizer’s data,

two logical operators (OR and AND), and three implications (“if A then B,” “if B
then A,” and “if A, then and only then, B”), there are 30 multivariate hypotheses: 18
when all three independent variables are used, plus 12 when two variables are used.
With more variables, the analysis becomes much more difficult. Computer programs
like QCA (Drass 1980) and ANTHROPAC (Borgatti 1992) test all possible
multivariate hypotheses and find the optimal solution. (QCA is reviewed in
Weitzman and Miles 1995.)

Computer-Assisted Text Analysis

One problem with text as data, as all field workers know, is that it piles up quickly.
The sheer volume and the problems of handling and sorting through so much
information has made text analysis less popular in the past than it is now. What has
changed is the presence of computers and of programs that help researchers code and
analyze text.
A useful review of the features of currently popular programs is given by
Weitzman and Miles (1995). Each of the most popular programs has a devoted group
of users who communicate regularly with one another (and with the authors of the
programs) on Internet lists. Academic centers specializing in text analysis have also
emerged, though mostly outside the United States. Centers at universities in Surrey
(Centre for Computer Assisted Qualitative Data Analysis Software) and Kent,
England (Centre for Social Anthropology and Computers), as well as Mannheim,
Germany (Zentrum für Umfragen Methoden und Analysen) and La Trobe, Australia
(Qualitative Solutions and Research) have helped to promote the use of qualitative
data across the social sciences.
The panorama with regard to text-processing software is changing quickly. Like
early word processors and database managers, the first generation of text processors
was designed to help us do what we already did. Today’s word processors let us do
things that printing presses could not do before computers, and today’s database
managers, with built-in statistical analysis capabilities, let us do things at our desks
that could only be done on a mainframe 15 years ago (and could not be done by
anyone 60 years ago).
The first generation of text analysis programs made light work of chores like
coding and finding the right quotes with which to illuminate a point in an article.
Text processors today still focus on coding and retrieving, but more and more we find
features that help with building conceptual models, linking concepts into networks,
and producing numerical text-by-variable matrices. Eventually, we expect text
processors to contain modules for doing word counts, concordances, KWIC (key-
word-in-context) studies, Boolean analysis, map analysis, and semantic network
analysis.
626 BERNARD / RYAN

Some General Observations

Text analysis as a research strategy permeates the social sciences, and the range of
methods for conducting text analysis is breathtaking. Investigators examine words,
sentences, paragraphs, pages, documents, ideas, meanings, paralinguistic features,
and even what is missing from the text. They interpret, mark, retrieve, and count. By
turns, they apply interpretive analysis and numerical analysis. They use text analysis
for exploratory and confirmatory purposes. Researchers identify themes, describe
them, compare them across cases and groups, and try to explain them. Text analysis
is used by avowed positivists and interpretivists alike.
Text analysts who are linguistically oriented rarely cite the work of their
sociologically and anthropologically oriented colleagues, and vice versa. Even within
the sociological tradition, the literatures on schema analysis, grounded theory, and
content analysis are essentially distinct, though they all involve identifying and
refining concepts (for example, categories, codes, themes), coding/marking concepts
in text, and linking concepts into theoretical models. One of our goals in this chapter
has been to make clear the importance of text analysis for all the social sciences.
We close this review by noting that some questions that might guide research on
methods of text analysis include the following:

1. What effect does the selection of textual units (thematic segments, sentences, or
words) have on the number, kind, and organization of themes that coders
identify?
2. To what extent does increasing or decreasing the number of coders affect the size
and composition of these themes?
3. Do different measures of intercoder agreement affect the content of themes?
4. How do different techniques for coding data (into thematic segments, sentences,
or words) affect descriptions of the range, central tendency, and distribution of
specific themes across informants?
5. Can different coding techniques produce contradictory findings? For example,
does an analysis of text segments make groups appear more or less similar than
does an analysis of words? Does an analysis of words suggest dimensions of
similarity or dissimilarity that are not captured by thematic or sentence
segments?
6. Do different methods for systematically treating text identify different sets of
prototypical quotes for illustrating major and minor themes?
7. How can we use qualitative and quantitative instruments to produce
complementary results? For example, can we identify informants who score high
and low on particular survey scale variables based only on their textual data?

Answers to these questions will help us understand better how to make the outcomes
and interpretations of qualitative data replicable.
 Text Analysis 627

NOTES

1. Holsti (1969) counted 2.5 content analysis studies per year, on average during the first
two decades of this century. By the 1920s, the number had risen to 13.3 per year; in the 1930s
it was 22.8, and in the 1940s it was 43.3. By the end of the 1950s, there were over 100 studies
a year using content analysis. We did a keyword survey in mid-1997 and retrieved more than
500 book titles, including more than 200 books written since 1990, on “qualitative data
analysis” or “text analysis.” The number of journal articles runs into the thousands, with entire
journals dedicated specifically to the analysis of qualitative data. These include: Qualitative
Sociology, Qualitative Inquiry, Journal of Contemporary Ethnography, Symbolic Interaction,
Qualitative Health Research, Quality and Quantity, Studies in Qualitative Methodology, and
International Journal of Qualitative Studies in Education. In anthropology, Cultural
Anthropology Methods Journal covers text analysis and other methods.
2. We do not review the fields of literary text analysis, biblical text analysis, authorship
studies, or hermeneutics. Text analysis in the social sciences owes much to these fields. On
hermeneutics, see Bleicher (1982), Bruns (1992), and Dilthey (1996). For a review of methods
in literary text analysis, see Segre and Kemeny (1988). For an introduction to biblical
hermeneutics, see Hayes and Holladay (1987). An important method, shared by all text
analysts, is the production of concordances and KWIC (key-words-in-context) lists.
Concordances are annotated lists, by page and/or line number, of the occurrence of every word,
phrase, or theme in a text or set of texts. KWIC lists are created by finding all the places in a
text where a particular word or phrase appears and printing it out in the context of some
number of words (say, 30) before and after it. For examples of major concordances, see Young
(1982), Kassis (1983), Spevack (1973), Prendergast (1971). See McKinnon (1993) and Burton
(1981a, 1981b, 1982) on the use of concordances in modern literary studies. In authorship
studies, differences in the use of words common to the writings of James Madison and
Alexander Hamilton led Mosteller and Wallace (1964) to conclude that Madison and not
Hamilton had written twelve of the Federalist Papers. See Yule (1944) and Martindale and
McKenzie (1995) for other examples of authorship studies.
3. Not all of Boas's students agreed that faithful reproduction of narrative was important.
Between 1903–1907, Clark Wissler collected 94 Blackfoot tales, all in English. His indigenous
collaborator, D. C. Duvall, provided the English translations, which Wissler edited and revised
for publication (Wissler and Duvall 1908). As it turned out, only a few of the tales were
translated from actual texts. Wissler explained:

In narration the Blackfoot often repeat sentences at irregular intervals, as if they wished
to prevent the listener from forgetting their import. Naturally such repetitions were
eliminated in the translations. A few narratives were recorded as texts. While texts will
be indispensable for linguistic research, the present condition of Blackfoot mythology
is such that its comparative study would not be materially facilitated by such records.
Each narrator has his own version, in the telling of which he is usually consistent; and
while the main features of the myths are the same for all, the minor differences are so
great that extreme accuracy of detail with one individual would avail little. (p. 5)
4. The Pear Story chronicles what happens to a bushel of pears after they have been picked
from a tree. The film is a series of scenes: A man picks the pears from a tree; another man and
a goat pass by; a boy on a bicycle stops and rides off with a basket of pears; when the boy turns
to look at a passing girl, his hat falls off, he runs into a rock and tips over; three other boys
help him up; on walking off, one of the three boys finds the hat; a whistle is heard; and the hat
628 BERNARD / RYAN

is exchanged for three pears; the film returns to the picker who looks down at the baskets and
scratches his head. The three boys appear eating their pears, and the picker watches them walk
off into the distance.
5. For reviews of popular software for text analysis, see Weitzman and Miles (1995). This
volume is the current best source of information on the features available in the array of
software available for text analysis. For an early, but still useful discussion of software issues
in text analysis, see Tesch (1990). For discussions of epistemological issues involved in using
computers for the qualitative analysis of texts, see Kelle (1995, 1997), and Lee and Fielding
(1995, 1996). Review articles are also found in the Cultural Anthropology Methods Journal,
Qualitative Sociology, and Computers and the Humanities.
6. Content analysis has a long history in the social sciences. Good reviews are available in
Krippendorf (1980) and Weber (1990). Monitoring newspapers by counting the space given
to various themes was already in vogue in the late nineteenth century and early twentieth. By
1910, at the first meeting of the German Sociological Society, Max Weber proposed a major
effort at content analysis of the German press (Krippendorf 1980:13). Weber’s proposal was
not implemented, but content analysis became, and remains, an important method for tracking
the impact of news events across time and across print venues (see Danielson and Lasorsa
[1997] and Kleinnijenhuis et al. [1997]—and many recent issues of Journalism Quarterly—for
current examples).
Methods for content analysis were developed vigorously during World War II in studying
speeches by Germany's leaders for clues about their intentions in the conduct of the war. In
1955, the Social Science Research Council’s Committee on Linguistics and Psychology
sponsored a conference on content analysis, bringing together experts from across the social
sciences. Their contributions appear in a landmark volume, edited by de Sola Pool (1959).
Since then, extensive reviews have appeared every decade or so: Gerbner et al. (1969), Holsti
(1969), Krippendorf (1980), Weber (1990), and Roberts (1997).
Today, classical content analysis continues to be used across the social sciences in the study
of media (Brouwer et al. 1969; Spiggle 1986; Hirschman 1987; Gilly 1988; Craig 1992;
Cameron and Blount 1996; Fink and Gantz 1996; Kolbe and Albanese 1996), political rhetoric
(Kaid, Tedesco, and McKinnon 1996; Franzosi 1997), business (Spears et al. 1996), medicine
(Cardador et al. 1995; Potts et al. 1996; Sleath et al. 1997), psychiatry, clinical psychology,
and counseling (Rosenberg 1990; Smith et al. 1992; Gottschalk 1994, 1997; Handron and
Legget-Frazier 1994), and law (Imrich et al. 1995). Content analysis is represented in
anthropology in cross-cultural hypothesis testing (White and Burton 1988; Bradley et al. 1990;
Ember et al. 1992) and in quantitative, comparative studies of folklore (Colby et al. 1991;
Johnson and Price-Williams 1997).
7. The amount of that fraction depends on the number of coders and the precision of
measurement for each code. If two people code a theme present or absent, they could agree,
ceteris paribus, on any answer 25% of the time by chance. If a theme, like wealth, is measured
ordinally (low, medium, high); then the likelihood of chance agreement changes accordingly.
Fleiss (1971) and Light (1971) expand Kappa to handle multiple coders. Another measure of
intercoder agreement for two coders is Scott's pi (Scott 1955), which was generalized by
Krippendorf (1980:147–154) for multiple coders, metric data, and any sample size. Craig
(1981) generalized Scott's pi for subsets of coders. Craig’s measure is particularly useful when
one wants to use majority rule in making coding decisions like two out of three coders. See
Holsti (1969) for several other measures.
 Text Analysis 629

8. Monte Carlo studies might prove helpful. Various types of error could be introduced to see
how well a model performs under conditions of uncertainty. Ultimately, however, there may
be no single solution but one that is derived within the context of each research problem
(Holsti 1969:143).
9. The General Inquirer also uses the Lasswell Value Dictionary (Namenwirth and Weber
1987; Zuell et al. 1989). More dictionaries are being built for the system.
10. Developing rules for sorting words into dictionary categories requires an understanding
of the multiple meanings of words. Concordances (see Note 2) and KWIC (pronounced
“quick”), or key-word-in-context lists are used as aids in this process. KWIC lists are created
by finding all the places in a text where a particular word or phrase appears and printing it out
in the context of some number of words (say, 30) before and after it. Stone et al. (1966:158)
created KWIC lists on a corpus of text that had more than a half million words to help revise
the Harvard III Psychological Dictionary.
11. See Namenwirth and Weber (1987) and Zuell et al. (1989) for information on other
dictionaries. Laffal (1990, 1995) created a 43,000-word dictionary that categorized each word
into 1–5 of 168 potential concept categories to distinguish themes in different literary works.
12. Rosenberg et al. (1990) transcribed seventy-one speech samples from people who had been
diagnosed with psychological disorders (depression, paranoia, somatisization) or cancer. The
transcripts were hand scored by an expert and analyzed using various editions of the Harvard
Psychosociological Dictionary (Schnurr et al. 1986). The human coder did better than the
computer in diagnosing patients who had cancer, but the computer beat the human coder in
identifying the transcripts of people who had been diagnosed with the various psychological
disorders.
13. Here are the details of the logic of Schweizer’s analysis. Three possible hypotheses can
be derived from two binary variables: “if A then B,” “if B then A,” and “if A, then and only
then, B.” In the first hypothesis, A is a sufficient condition to B and B is necessary to A. This
hypothesis is falsified by all cases having A and not B. In the second hypothesis, B is a
sufficient condition to A and A is necessary to B. The second hypothesis is falsified by all
cases of B and not A. These two hypotheses are implications or conditional statements. The
third hypothesis (an equivalence or biconditional statement) is the strongest: whenever you see
A, you also see B and vice versa; the absence of A implies the absence of B and vice versa.
This hypothesis is falsified by all cases of A and not B, and all cases of B and not A.
Applied to the data from Chen Village, the strong hypothesis is falsified by many cases, but
the sufficient condition hypotheses (urban origin implies success; proletarian background
implies success; having external ties implies success) are true in 86% of the cases (this is an
average of the three sufficient condition hypotheses). The necessary condition hypotheses (suc-
cess implies urban origin; success implies proletarian background; success implies external
ties) is true in just 73% of cases (again, an average). (There are 7 disconfirming cases in 51
possible outcomes of the 12 sufficient condition possibilities—4 possible outcomes for each
of 3 independent variables and one dependent variable. There are 14 disconfirming cases in
51 possible outcomes of the 12 necessary condition possibilities.) To improve on this,
Schweizer tested multivariate hypotheses, using the logical operators OR and AND.
630 BERNARD / RYAN

We are sure that many Blackfoot today (not to mention linguists) would rather that Wissler had
followed Boas's instructions to gather the texts in the original language and to record the texts
as faithfully as possible. See Werner (1995) for a recent argument in favor of verbatim
transcription of text.
4. The Pear Story chronicles what happens to a bushel of pears after they have been picked
from a tree. The film is a series of scenes: A man picks the pears from a tree; another man and
a goat pass by; a boy on a bicycle stops and rides off with a basket of pears; when the boy turns
to look at a passing girl, his hat falls off, he runs into a rock and tips over; three other boys
help him up; on walking off, one of the three boys finds the hat; a whistle is heard; and the hat
is exchanged for three pears; the film returns to the picker who looks down at the baskets and
scratches his head. The three boys appear eating their pears, and the picker watches them walk
off into the distance.
5. For reviews of popular software for text analysis, see Weitzman and Miles (1995). This
volume is the current best source of information on the features available in the array of
software available for text analysis. For an early, but still useful discussion of software issues
in text analysis, see Tesch (1990). For discussions of epistemological issues involved in using
computers for the qualitative analysis of texts, see Kelle (1995, 1997), and Lee and Fielding
(1995, 1996). Review articles are also found in the Cultural Anthropology Methods Journal,
Qualitative Sociology, and Computers and the Humanities.
6. Content analysis has a long history in the social sciences. Good reviews are available in
Krippendorf (1980) and Weber (1990). Monitoring newspapers by counting the space given
to various themes was already in vogue in the late nineteenth century and early twentieth. By
1910, at the first meeting of the German Sociological Society, Max Weber proposed a major
effort at content analysis of the German press (Krippendorf 1980:13). Weber’s proposal was
not implemented, but content analysis became, and remains, an important method for tracking
the impact of news events across time and across print venues (see Danielson and Lasorsa
[1997] and Kleinnijenhuis et al. [1997]—and many recent issues of Journalism Quarterly—for
current examples).
Methods for content analysis were developed vigorously during World War II in studying
speeches by Germany's leaders for clues about their intentions in the conduct of the war. In
1955, the Social Science Research Council’s Committee on Linguistics and Psychology
sponsored a conference on content analysis, bringing together experts from across the social
sciences. Their contributions appear in a landmark volume, edited by de Sola Pool (1959).
Since then, extensive reviews have appeared every decade or so: Gerbner et al. (1969), Holsti
(1969), Krippendorf (1980), Weber (1990), and Roberts (1997).
Today, classical content analysis continues to be used across the social sciences in the study
of media (Brouwer et al. 1969; Spiggle 1986; Hirschman 1987; Gilly 1988; Craig 1992;
Cameron and Blount 1996; Fink and Gantz 1996; Kolbe and Albanese 1996), political rhetoric
(Kaid, Tedesco, and McKinnon 1996; Franzosi 1997), business (Spears et al. 1996), medicine
(Cardador et al. 1995; Potts et al. 1996; Sleath et al. 1997), psychiatry, clinical psychology,
and counseling (Rosenberg 1990; Smith et al. 1992; Gottschalk 1994, 1997; Handron and
Legget-Frazier 1994), and law (Imrich et al. 1995). Content analysis is represented in
anthropology in cross-cultural hypothesis testing (White and Burton 1988; Bradley et al. 1990;
Ember et al. 1992) and in quantitative, comparative studies of folklore (Colby et al. 1991;
Johnson and Price-Williams 1997).

7. The amount of that fraction depends on the number of coders and the precision of
measurement for each code. If two people code a theme present or absent, they could agree,
 Text Analysis 631

ceteris paribus, on any answer 25% of the time by chance. If a theme, like wealth, is measured
ordinally (low, medium, high); then the likelihood of chance agreement changes accordingly.
Fleiss (1971) and Light (1971) expand Kappa to handle multiple coders. Another measure of
intercoder agreement for two coders is Scott's pi (Scott 1955), which was generalized by
Krippendorf (1980:147–154) for multiple coders, metric data, and any sample size. Craig
(1981) generalized Scott's pi for subsets of coders. Craig’s measure is particularly useful when
one wants to use majority rule in making coding decisions like two out of three coders. See
Holsti (1969) for several other measures.
8. Monte Carlo studies might prove helpful. Various types of error could be introduced to
see how well a model performs under conditions of uncertainty. Ultimately, however, there
may be no single solution but one that is derived within the context of each research problem
(Holsti 1969:143).
9. The General Inquirer also uses the Lasswell Value Dictionary (Namenwirth and Weber
1987; Zuell et al. 1989). More dictionaries are being built for the system.
10. Developing rules for sorting words into dictionary categories requires an understanding
of the multiple meanings of words. Concordances (see Note 2) and KWIC (pronounced
“quick”), or key-word-in-context lists are used as aids in this process. KWIC lists are created
by finding all the places in a text where a particular word or phrase appears and printing it out
in the context of some number of words (say, 30) before and after it. Stone et al. (1966:158)
created KWIC lists on a corpus of text that had more than a half million words to help revise
the Harvard III Psychological Dictionary.
11. See Namenwirth and Weber (1987) and Zuell et al. (1989) for information on other
dictionaries. Laffal (1990, 1995) created a 43,000-word dictionary that categorized each word
into 1–5 of 168 potential concept categories to distinguish themes in different literary works.
12. Rosenberg et al. (1990) transcribed 71 speech samples from people who had been
diagnosed with psychological disorders (depression, paranoia, somatisization) or cancer. The
transcripts were hand scored by an expert and analyzed using various editions of the Harvard
Psychosociological Dictionary (Schnurr et al. 1986). The human coder did better than the
computer in diagnosing patients who had cancer, but the computer beat the human coder in
identifying the transcripts of people who had been diagnosed with the various psychological
disorders.
13. Here are the details of the logic of Schweizer’s analysis. Three possible hypotheses can
be derived from two binary variables: “if A then B,” “if B then A,” and “if A, then and only
then, B.” In the first hypothesis, A is a sufficient condition to B and B is necessary to A. This
hypothesis is falsified by all cases having A and not B. In the second hypothesis, B is a
sufficient condition to A and A is necessary to B. The second hypothesis is falsified by all
cases of B and not A. These two hypotheses are implications or conditional statements. The
third hypothesis (an equivalence or biconditional statement) is the strongest: whenever you see
A, you also see B and vice versa; the absence of A implies the absence of B and vice versa.
This hypothesis is falsified by all cases of A and not B, and all cases of B and not A.
Applied to the data from Chen Village, the strong hypothesis is falsified by many cases, but
the sufficient condition hypotheses (urban origin implies success; proletarian background
implies success; having external ties implies success) are true in 86% of the cases (this is an
average of the three sufficient condition hypotheses). The necessary condition hypotheses (suc-
cess implies urban origin; success implies proletarian background; success implies external
ties) is true in just 73% of cases (again, an average). (There are 7 disconfirming cases in 51
possible outcomes of the 12 sufficient condition possibilities—4 possible outcomes for each
632 BERNARD / RYAN

of 3 independent variables and one dependent variable. There are 14 disconfirming cases in
51 possible outcomes of the 12 necessary condition possibilities.) To improve on this,
Schweizer tested multivariate hypotheses, using the logical operators OR and AND.

REFERENCES

Agar, Michael. 1979. Themes Revisited: Some Problems in Cognitive Anthropology.

Discourse Processes 2:11–31.
Agar, Michael. 1980. Getting Better Quality Stuff: Methodological Competition in an
Interdisciplinary Niche. Urban Life 9:34–50.
Agar, Michael. 1983. Microcomputers as Field Tools. Computers and the Humanities
17:19–26.
Agar, Michael. 1986. Speaking of Ethnography. Qualitative Research Methods Series, Vol.
2. Thousand Oaks, CA: Sage Publications.
Agar, Michael. 1996. The Professional Stranger, 2d ed. San Diego: Academic Press.
Araujo, Luis. 1995. Designing and Refining Hierarchical Coding Frames. In Computer-Aided
Qualitative Data Analysis: Theory, Methods and Practice. Udo Kelle, ed. Pp. 96–104.
London: Sage Publications.
Barnett, George, and James Danowski. 1992. The Structure of Communication: A Network
Analysis of the International Communication Association. Human Communication Resources
19:164–285.
Bartlett, F. C. 1967 [1932]. Remembering: A Study in Experimental and Social Psychology.
Cambridge: Cambridge University Press.
Becker, Howard, B. Geer, E. C. Hughes, and A. L. Strauss. 1961. Boys in White: Student
Culture in Medical School. Chicago: University of Chicago Press.
Berelson, Bernard. 1952. Content Analysis in Communication Research. Glencoe, IL: Free
Press.
Bernard, H. Russell. 1991. About Text Management and Computers. CAM Newsletter
3(1):1–4, 7, 12.
Bernard, H. Russell. 1994. Research Methods in Anthropology: Qualitative and Quantitative
Approaches, 2d ed. Walnut Creek, CA: AltaMira Press.
Bernard, H. Russell. 1997. Language Preservation and Publishing. In Indigenous Literacy in
the Americas. Nancy H. Hornberger, ed. Pp. 139–156. Berlin: Mouton de Gruyter.
Bernard, H. Russell, and Jesús Salinas Pedraza. 1976. Otomí Parables, Folktales, and Jokes.
International Journal of American Linguistics, Native American Texts Series 1(2). Chicago:
University of Chicago Press.
Bernard, H. Russell, and Jesús Salinas Pedraza. 1989. Native Ethnography. A Mexican Indian
Describes His Culture. Thousand Oaks, CA: Sage Publications.
Besnier, Niko 1995. Literacy, Emotion, and Authority. Reading and Writing on a Polynesian
Atoll. Cambridge: Cambridge University Press.
Bizarro Ujpan, Ignacio, and James Sexton. 1981. Son of Tecun Uman: A Maya Indian Tells
His Life Story. Tucson: University of Arizona Press.
Bizarro Ujpan, Ignacio, and James Sexton. 1985. Campesino: The Diary of a Guatemalan
 Text Analysis 633

Indian. Tucson: University of Arizona Press.

Bleicher, J. 1982. The Hermeneutic Imagination: Outline of a Positive Critique of Scientism
and Sociology. London: Routledge and Kegan Paul.
Blowsnake, Sam, and Paul Radin (ed. and tr.). 1983. Crashing Thunder: Autobiography of a
Winnebago Indian. Lincoln: University of Nebraska Press. Orig. pub. 1920 by the University
of California Press and again in 1926 by Appleton.
Boas, Franz. 1910. Kwakiutl Tales. New York: Columbia University Press.
Boas, Franz. 1917. Folk-Tales of Salishan and Sahaptin Tribes. Lancaster, PA: American
Folk-Lore Society.
Boas, Franz. 1928a. Kerasan Texts. New York: The American Ethnological Society.
Boas, Franz. 1928b. Bella Bella Texts. New York: Columbia University Press.
Boas, Franz. 1935–43. Kwakiutl Tales, New Series. New York: Columbia University Press.
Boas, Franz, and George Hunt. 1902–1905. Kwakiutl Texts. Leiden: E. J. Brill.
Boas, Franz, and George Hunt. 1906. Kwakiutl Texts, Second Series. Memoirs of the
American Museum of Natural History, 14. New York: G. E. Stechert.
Bogdan, Robert, and Sari Knopp Biklen. 1992. Qualitative Research for Education: An Intro-
duction to Theory and Methods, 2d ed. Boston: Allyn and Bacon.
Borgatti, Steven. 1992. ANTHROPAC 4.0. Columbia, SC: Analytic Technologies
(www.analytictech.com).
Bradley, Candice, C. C. Moore, M. L. Burton, and D. R. White. 1990. A Cross-Cultural
Historical Analysis of Subsistence Change. American Anthropologist 92:447–457.
Brouwer, M., C. C. Clark, G. Gerbner, and K. Krippendorff. 1969. The Television World of
Violence. In Mass Media and Violence: A Report to the National Commission on the Causes
and Prevention of Violence. R. K. Baker and S. J. Ball, eds. Pp. 311–339, 519–591.
Washington, DC: Government Printing Office.
Bruns, Gerald. 1992. Hermeneutics Ancient and Modern. New Haven: Yale University Press.
Burton, Dolores M. 1981a. Automated Concordances and Word Indexes: The Early Sixties and
the Early Centers. Computers and the Humanities 15:83–100.
Burton, Dolores M. 1981b. Automated Concordances and Word Indexes: The Process, the
Programs, and the Products. Computers and the Humanities 15:139–154.
Burton, Dolores M. 1982. Automated Concordances and Word-Indexes: Machine Decisions
and Editorial Revisions. Computers and the Humanities 16:195–218.
Cameron, G. T., and D. Blount. 1996. VNRs and Air Checks: A Content Analysis of the Use
of Video News Releases in Television Newscasts. Journalism and Mass Communication
Quarterly 73(4):890–904.
Cardador, M. T., A. R. Hazan, and S. A. Glantz. 1995. Tobacco Industry Smokers’ Rights
Publications—A Content Analysis. American Journal of Public Health 85(9): 1212–1217.
Carey, James W., Mark Morgan, and Margaret J. Oxtoby. 1996. Intercoder Agreement in
Analysis of Responses to Open-Ended Interview Questions: Examples from Tuberculosis
Research. Cultural Anthropology Methods Journal 8(3):1–5.
Carley, Kathleen. 1993. Coding Choices for Textual Analysis: A Comparison of Content
Analysis and Map Analysis. In Sociological Methodology. P. Marsden, ed. Pp. 75–126.
634 BERNARD / RYAN

Oxford: Blackwell.
Carley, Kathleen. 1997. Network Text Analysis: The Network Position of Concepts. In Text
Analysis for the Social Sciences: Methods for Drawing Statistical Inferences from Texts and
Transcripts. Carl W. Roberts, ed. Pp. 79–100. Mahwah, NJ: Lawrence Erlbaum.
Carley, Kathleen, and D. S. Kaufer. 1993. Semantic Connectivity: An Approach for Analyzing
Semantic Networks. Communication Theory 3:182–213.
Carley, Kathleen, and Michael Palmquist. 1992. Extracting, Representing, and Analyzing
Mental Models. Social Forces 70:601–636.
Casson, Ronald. 1983. Schemata in Cultural Anthropology. Annual Review of Anthropology
12:429–462.
Chafe, W. 1980. The Deployment of Consciousness in the Production of Narrative. In The Pear
Stories. Cognitive, Cultural, and Linguistic Aspects of Narrative Production. W. Chafe, ed.
Pp. 9–50. Norwood, NJ: Ablex.
Chan, Anita, R. Madsen, and J. Unger. 1984. Chen Village. The Recent History of a Peasant
Community in Mao’s China. Berkeley: University of California Press.
Charmaz, Kathy. 1990. “ Discovering” Chronic Illness: Using Grounded Theory. Social Science
and Medicine 30:1161–1172.
Cohen, Alex. 1990. A Cross-Cultural Study of the Effects of Environmental Unpredictability
on Aggression in Folktales. American Anthropologist 92:474–481.
Cohen, Jacob. 1960. A Coefficient of Agreement for Nominal Scales. Educational and
Psychological Measurement 20:37–48.
Colby, B. N. 1966. The Analysis of Culture Content and the Patterning of Narrative Concern
in Texts. American Anthropologist 68:374–388.
Colby, B. N., S. Kennedy, and L. Milanesi. 1991. Content Analysis, Cultural Grammars, and
Computers. Qualitative Sociology 14:373–384.
Cowan, G., and M. O’Brien. 1990. Gender and Survival Versus Death in Slasher Films—A
Content Analysis. Sex Roles 23(3/4):187–196.
Crabtree, Benjamin F., and William L. Miller. 1992. A Template Approach to Text Analysis:
Developing and Using Codebooks. In Doing Qualitative Research. William L. Miller and
Benjamin F. Crabtree, eds. Pp. 93–109. Thousand Oaks, CA: Sage Publications.
Craig, Robert T. 1981. Generalization of Scott’s Index of Intercoder Agreement. Public
Opinion Quarterly 45:260–264.
Craig, R. Stephen. 1992. The Effect of Television Day Part on Gender Portrayals in Television
Commercials: A Content Analysis. Sex Roles 26(5/6):197–211.
Cressey, Donald R. 1973 (orig 1953). Other People’s Money: A Study in the Social Psychology
of Embezzlement. Montclair, NJ: Patterson Smith.
D’Andrade, Roy. 1984. Cultural Meaning Systems. In Culture Theory: Essays on Mind, Self,
and Emotion. R. A. Shweder and R. A. Levine, eds. Pp. 88–119. Cambridge: Cambridge
University Press.
D’Andrade, Roy. 1987. A Folk Model of the Mind. In Cultural Models in Language and
Thought. N. Quinn and D. Holland, eds. Pp. 112–148. Cambridge: Cambridge University
Press.
D’Andrade, Roy. 1991. The Identification of Schemas in Naturalistic Data. In Person Schemas
 Text Analysis 635

and Maladaptive Interpersonal Patterns. Mardi J. Horowitz, ed. Pp. 279–301. Chicago:
University of Chicago Press.
D’Andrade, Roy. 1995. The Development of Cognitive Anthropology. Cambridge: Cambridge
University Press.
Danielson, Wayne A., and D. L. Lasorsa. 1997. Perceptions of Social Change: 100 Years of
Front-Page Content in The New York Times and The Los Angeles Times. In Text Analysis for
the Social Sciences: Methods for Drawing Statistical Inferences from Texts and Transcripts.
Carl W. Roberts, ed. Pp. 103–116. Mahwah, NJ: Lawrence Erlbaum.
Danowski, James. 1982. A Network-Based Content Analysis Methodology for Computer-
Mediated Communication: An Illustration with a Computer Bulletin Board. In Commu-nica-
tion Yearbook. R. Bostrom, ed. Pp. 904–925. New Brunswick, NJ: Transaction Books.
Danowski, James. 1993. Network Analysis of Message Content. In Progress in Commu-
nication Science, XII. W. D. Richards and G. A. Barnett, eds. Pp. 197–221. Norwood, NJ:
Ablex Publishing.
de Sola Pool, Ithiel, ed. 1959. Trends in Content Analysis. Urbana: University of Illinois Press.
Deese, James. 1969. Conceptual Categories in the Study of Content. In The Analysis of
Communication Content: Development in Scientific Theories and Computer Techniques.
George Gerbner, Ole R. Holsti, Klaus Krippendorff, William J. Paisley, and Philip J. Stone,
eds. Pp. 39–56. New York: Wiley.
Denzin, N. K. 1978. The Research Act. A Theoretical Introduction to Sociological Methods,
2d ed. New York: McGraw-Hill.
Dey, I. 1993. Qualitative Data Analysis: A User-Friendly Guide for Social Scientists. London:
Routledge and Kegan Paul.
Dilthey, Wilhelm. 1996. Hermeneutics and the Study of History. Princeton: Princeton
University Press.
Doerfel, Marya L., and George A. Barnett. 1996. The Use of Catpac for Text Analysis.
Cultural Anthropology Methods Journal 8(2):4–7.
Downing, Pamela. 1980. Factors Influencing Lexical Choice in Narrative. In The Pear Stories.
Cognitive, Cultural, and Linguistic Aspects of Narrative Production. W. Chafe, ed. Pp.
89–126. Norwood, NJ: Ablex Publishing.
Drass, K. 1980. The Analysis of Qualitative Data: A Computer Program. Urban Life
9:332–353.
Edmonson, Munro S. 1982. The Ancient Future of the Itza: The Book of Chilam Balam of
Tizimin. Austin: University of Texas Press.
Edmonson, Munro S., tr. and ed. 1986. Heaven Born Mérida and Its Destiny: The Book of
Chilam Balam of Chumayel. Austin: University of Texas Press.
El Guindi, Fadwa, and Abel Hernández Jiménez. 1986. The Myth of Ritual. A Native’s
Ethnography of Zapotec Life-Crisis Rituals. Tucson: University of Arizona Press.
Ember, Carol R., and Melvin Ember. 1992. Resource Unpredictability, Mistrust, and War: A
Cross-Cultural Study. Journal of Conflict Resolution 36:242–262.
Ember, Carol R., Melvin Ember, and B. M. Russett. 1992. Peace Between Participatory
Polities: A Cross-Cultural Test of the ‘Democracies Rarely Fight Each Other’ Hypothesis.
World Politics 44:573–599.
636 BERNARD / RYAN

Fan, David P., and Carol L. Shaffer. 1990. Use of Open-Ended Essays and Computer Content
Analysis to Survey College Students’ Knowledge of AIDS. College Health 38 (March):
221–229.
Fielding, Nigel G., and Raymond M. Lee. 1992. Using Computers in Qualitative Research,
rev. ed. London: Sage Publications.
Fink, E. J., and W. Gantz. 1996. A Content Analysis of Three Mass Communication Research
Traditions—Social Science, Interpretive Studies, and Critical Analysis. Journalism and Mass
Communication Quarterly 73:114–134.
Fleiss, Joseph L. 1971. Measuring Nominal Scale Agreement among Many Raters.
Psychological Bulletin 76:378–382.
Franzosi, Roberto. 1997. Labor Unrest in the Italian Service Sector: An Application of
Semantic Grammars. In Text Analysis for the Social Sciences: Methods for Drawing Statistical
Inferences from Texts and Transcripts. Carl W. Roberts, ed. Pp. 131–146. Mahwah, NJ:
Lawrence Erlbaum.
Furbee, Louanna 1996. The Religion of Politics in Chiapas: Founding a Cult of Com-
municating Saints. Presented at the 96th Meeting of the American Anthropological
Association, San Francisco.
Furbee-Losee, Louanna, ed. 1976. Mayan Texts I. International Journal of American
Linguistics. Native American Texts Series, 1(1). Chicago: University of Chicago Press.
Gerbner, G., O. R. Holsti, K. Krippendorf, W. J. Paisey, and P. J. Stone, eds. 1969. The
Analysis of Communication Content: Developments in Scientific Theories and Computer
Techniques. New York: Wiley.
Gilly, Mary C. 1988. Sex Roles in Advertising: A Comparison of Television Advertisements
in Australia, Mexico, and the United States. Journal of Marketing 52(April):75–85.
Glaser, Barney G., and Anselm Strauss. 1967. The Discovery of Grounded Theory: Strategies
for Qualitative Research. New York: Aldine.
Goddard, Pliny E. 1911. Jicarilla Apache Texts. Anthropological Papers of the American
Museum of Natural History 8. New York: The Trustees.
González Ventura, Josefa. 1997. Experiences in the Development of a Writing System for Ñuu
Savi. In Indigenous Literacy in the Americas. Nancy H. Hornberger, ed. Pp. 157–171. Berlin:
Mouton de Gruyter.
Gottschalk, Louis A. 1994. The Development, Validation, and Applications of a Com-
puterized Measurement of Cognitive Impairment from the Content Analysis of Verbal
Behavior. Journal of Clinical Psychology 50(3):349–361.
Gottschalk, Louis A. 1997. The Unobtrusive Measurement of Psychological States and Traits.
In Text Analysis for the Social Sciences: Methods for Drawing Statistical Inferences from
Texts and Transcripts. Carl W. Roberts. ed. Pp. 117–129. Mahway, NJ: Lawrence Erlbaum
Associates.
Gottschalk, Louis A., and R. J. Bechtel. 1993. Psychologic and Neuropsychiatric
Assessment Applying the Gottschalk-Gleser Content Analysis Method to Verbal Sample
Analysis Using the Gottschalk-Bechtel Computer Scoring System. Palo Alto, CA: Mind
Garden.
Hage, Per, and Frank Harary 1983. Structural Models in Anthropology. New York: Cambridge
University Press.
 Text Analysis 637

Handron, D.S., and N. K. Leggett-Frazier. 1994. Utilizing Content Analysis of Counseling

Sessions to Identify Psychosocial Stressors among Patients with Type II Diabetes. The Diabetes
Educator 20(6):515–520.
Hanks, William F. 1988. Grammar, Style, and Meaning in a Maya Manuscript. Review of
Heaven Born Mérida and Its Destiny: The Book of Chilam Balam of Chumayel. Inter-national
Journal of American Linguistics 54:331–369.
Hanks, William F. 1989. Texts and Textuality. Annual Review of Anthropology 18: 95–127.
Hayes, John, and Carl Holladay. 1987. Biblical Exegesis, rev. ed. Atlanta: John Knox Press.
Herzfeld, Michael. 1977. Ritual and Textual Structures: The Advent of Spring in Rural
Greece. In Text and Context. Ravindra K. Jain, ed. Pp. 29–45. Philadelphia: Institute for the
Study of Human Issues.
Hirschman, E. C. 1987. People as Products: Analysis of a Complex Marketing Exchange.
Journal of Marketing 51:98–108.
Hitchcock, J., and H. S. Wilson. 1992. Personal Risking: Lesbian Self-Disclosure to Health
Professionals. Nursing Research 41:178–183.
Holland, Dorothy. 1985. From Situation to Impression: How Americans Get to Know Them-
selves and One Another. In Directions in Cognitive Anthropology. J.W.D. Dougherty, ed. Pp.
389–412. Urbana: University of Illinois Press.
Holsti, Ole R. 1966. External Conflict and Internal Consensus: The Sino-Soviet Case. In The
General Inquirer: A Computer Approach to Content Analysis. P. J. Stone, D. C. Dunphy, M.
S. Smith, and D. M. Ogilvie, eds. Pp. 343–358. Cambridge: M.I.T. Press.
Holsti, Ole R. 1969. Content Analysis for the Social Sciences and Humanities. Reading, MA:
Addison-Wesley.
Hunt, J. G., and A. Ropo. 1995. Multi-Level Leadership—Grounded Theory and Mainstream
Theory Applied to the Case of General Motors. Leadership Quarterly 6:379-412.
Hutchins, Edwin. 1980. Culture and Inference. Cambridge: Harvard University Press.
Hymes, Dell. 1976. Louis Simpson’s ‘The Deserted Boy.’ Poetics 5:119–155.
Hymes, Dell. 1977. Discovering Oral Performance and Measured Verse in American Indian
Narrative. New Literary History 8:431–457.
Hymes, Dell. 1980a. Verse Analysis of a Wasco Text: Hiram Smith’s “At’unaqa.” Inter-
national Journal of American Linguistics 46:65–77.
Hymes, Dell 1980b. Particle, Pause, and Pattern in American Indian Narrative Verse.
American Indian Culture and Research Journal 4(4):7–51).
Hymes, Dell. 1981. In Vain I Tried to Tell You: Essays in Native American Ethnopoetics.
Philadelphia: University of Pennsylvania Press.
Hymes, Virginia. 1987. Warm Springs Sahaptin Narrative Analysis. In Native American
Discourse: Poetics and Rhetoric. Joel Sherzer and Anthony Woodbury, eds. Pp. 62–102.
Cambridge: Cambridge University Press.
Imrich, D. J., C. Mullin, and D. Linz. 1995. Measuring the Extent of Prejudicial Pretrial
Publicity in Major American Newspapers—A Content Analysis. Journal of Commu-nication
45(3):94–117.
Irurita, V. F. 1996. Hidden Dimensions Revealed—Progressive Grounded Theory Study of
638 BERNARD / RYAN

Quality Care in the Hospital. Qualitative Health Research 6:331–349.

Jang, H-Y., and G. Barnett. 1994. Cultural Differences in Organizational Communication: A
Semantic Network Analysis. Bulletin de Méthodologie Sociologique 44 (Septem-ber):31–59.
Jehn, Karen A., and Lorna Doucet. 1996. Developing Categories from Interview Data: Text
Analysis and Multidimensional Scaling. Part 1. Cultural Anthropology Methods Journal
8(2):15–16.
Jehn, Karen A., and Lorna Doucet. 1997. Developing Categories for Interview Data:
Consequences of Different Coding and Analysis Strategies in Understanding Text. Part 2.
Cultural Anthropology Methods Journal 9(1):1–7.
Jehn, Karen A., and O. Werner. 1993. Hapax Legomenon II: Theory, A Thesaurus, and Word
Frequency. Cultural Anthropology Methods Journal 5(1):8–10.
Johnson, Allen, and Douglass Price-Williams. 1997. Oedipus Ubiquitous: The Family Com-
plex in World Folk Literature. Stanford: Stanford University Press.
Johnson-Laird, P. N. 1983. Mental Models: Towards a Cognitive Science of Language,
Inference, and Consciousness. Cambridge: Harvard University Press.
Johnson-Laird, P. N. 1989. Mental Models. In Foundations of Cognitive Science. Michael I.
Posner, ed. Pp. 469–500. Cambridge: M.I.T. Press.
Kaid, Lynda Lee, John C. Tedesco, and Lori Melton McKinnon. 1996. Presidential Ads as
Nightly News: A Content Analysis of 1988 and 1992 Televised Adwatches. Journal of
Broadcasting Electronic Media 40:297–308.
Kassis, Hanna. 1983. A Concordance of the Qur’an. Berkeley: University of California Press.
Kearney M. H., S. Murphy, and M. Rosenbaum. 1994. Mothering on Crack Cocaine—A
Grounded Theory Analysis. Social Science and Medicine 38:351–361.
Kearney, M. H., S. Murphy, K. Irwin, and M. Rosenbaum. 1995. Salvaging Self—A Grounded
Theory of Pregnancy on Crack Cocaine. Nursing Research 44(4):208–213.
Kelle, Udo. 1995. An Overview of Computer-Aided Methods in Qualitative Research. In
Computer-Aided Qualitative Data Analysis: Theory, Methods and Practice. Udo Kelle, ed. Pp.
1–18. London: Sage Publications.
Kelle, Udo. 1997. Theory Building in Qualitative Research and Computer Programs for the
Man agemen t of Textual Data. Sociological Research Online 2 (2):
http://www.socresonline.org.uk/socresonline/2/1/1.html.
Kelly, E. F., and P. J. Stone. 1975. Computer Recognition of English Word Senses.
Amsterdam: North-Holland.
Kempton, Willett. 1987. Two Theories of Home Heat Control. In Cultural Models in
Language and Thought. D. Holland and N. Quinn, eds. Pp. 222–242. Cambridge: Cambridge
University Press.
Kleinnijenhuis, Jan, J. A. de Ridder, and E. M. Rietberg. 1997. Reasoning in Economic
Discourse: An Application of the Network Approach to the Dutch Press. In Text Analysis
for the Social Sciences: Methods for Drawing Statistical Inferences from Texts and
Transcripts. Carl W. Roberts, ed. Pp. 191–208. Mahwah, NJ: Lawrence Erlbaum.
Kolbe, R. H., and J. P. Albanese. 1996. Man to Man: A Content Analysis of Sole-Male Images
in Male-Audience Magazines. Journal of Advertising 25(4):1–20.
Kortendick, Oliver. 1996. Drei Schwestern un ihre Kinder. Rekonstruktion von
 Text Analysis 639

Familiengeschichte und Identitätstransmission bei Indischen Nederlanders mit Hilfe

computerunterstützer Inhaltsanlyze. Canterbury, England: University of Kent, Center for
Anthropology and Computing.
Krippendorf, Klaus. 1980. Content Analysis: An Introduction to Its Methodology. Beverly
Hills, CA: Sage Publications.
Kroeber, Alfred. 1907. Gros Ventre Myths and Tales. Anthropological Papers of the American
Museum of Natural History 1(Part 3). New York: The Trustees.
Laffal, Julius. 1990. A Concept Dictionary of English, with Computer Programs for Content
Analysis. Essex, CT: Gallery Press.
Laffal, Julius. 1995. A Concept Analysis of Jonathan Swift’s A Tale of a Tub and Gulliver’s
Travels. Computers and the Humanities 29:339–361.

Lakoff, George, and Mark Johnson. 1980. Metaphors We Live By. Chicago: University of
Chicago Press.
Lee, Raymond M., and Nigel G. Fielding. 1995. User’s Experiences of Qualitative Data
Analysis Software. In Computer-Aided Qualitative Data Analysis: Theory, Methods and
Practice. Udo Kelle, ed. Pp. 29–40. London: Sage Publications.
Lee, Raymond M., and Nigel G. Fielding. 1996. Qualitative Data Analysis: Representations
of a Technology—A Comment on Coffey, Holbrook and Atkinson. Sociological Research
Online 1(4): <http://www.socresonline.org.uk/socresonline/1/4/lf.html>.
Light, Richard J. 1971. Measures of Response Agreement for Qualitative Data: Some
Generalizations and Alternatives. Psychological Bulletin 76(5):365–377.
Lincoln, Yvonna S., and Egon G. Guba. 1985. Naturalistic Inquiry. Beverly Hills, CA: Sage
Publications.
Lindesmith, A. R. 1968 [1947]. Addiction and Opiates. Chicago: Aldine.
Locke, K. 1996. Rewriting the Discovery of Grounded Theory—After 25 Years. Journal of
Management Inquiry 5(3):239–245.
Lofland, John, and Lyn H. Lofland. 1995. Analyzing Social Settings, 3rd ed. Belmont, CA:
Wadsworth.
Lonkila, Markku. 1995. Grounded Theory as an Emerging Paradigm for Computer-Assisted
Qualitative Data Analysis. In Computer-Aided Qualitative Data Analysis. Udo Kelle, ed. Pp.
41–51. Thousand Oaks, CA: Sage Publications.
Lowie, Robert H. 1930. A Crow Text, with Grammatical Notes. Berkeley: University of
California Press.
Lurie, Nancy. 1961. Mountain Lone Wolf Woman, Sister of Crashing Thunder: The
Autobiography of a Winnebago Woman. Ann Arbor: University of Michigan Press.
Mandler, Jean. 1984. Stories, Scripts, and Scenes: Aspects of Schema Theory. Hillsdale, NJ:
Lawrence Erlbaum.
Manning, Peter K. 1982. Analytic Induction. In Handbook of Social Science Methods, Vol. 2,
Qualitative Methods. R. Smith and P. K. Manning, eds. Pp. 273–302. New York: Harper.
Martindale, Colin, and Dean McKenzie. 1995. On the Utility of Content Analysis in Author
Attribution: The Federalist. Computers and the Humanities 29:259–270.
Mathews, Holly. 1992. The Directive Force of Morality Tales in a Mexican Community. In
640 BERNARD / RYAN

Human Motives and Cultural Models. Roy D’Andrade and Claudia Stauss, eds. Pp. 127–162.
New York: Cambridge University Press.
McCleary, R. 1978. Dangerous Men: The Sociology of Parole. Beverly Hills, CA: Sage
Publications.
McKinnon, Alastair. 1993. The Multi-Dimensional Concordance: A New Tool for Literary
Research. Computers and the Humanities 27:165–183.
McTavish, Donald G., and E. B. Pirro. 1990. Contextual Content Analysis. Quality and
Quantity 24:44–63.
Miles, Matthew, and A. Michael Huberman. 1994. Qualitative Data Analysis, 2d ed.
Thousand Oaks, CA: Sage Publications.
Morse, J. M. 1994. Designing Funded Qualitative Research. In Handbook of Qualitative
Research. Norman K. Denzin and Y. S. Lincoln, eds. Pp. 220–235. Thousand Oaks, CA: Sage
Publications.
Mosteller, F., and D. L. Wallace. 1964. Inference and Disputed Authorship: The Federalist
Papers. Reading, MA: Addison-Wesley.
Muhr, Thomas. 1991. Atlas/ti—A Protoype for the Support of Text Interpretation. Qualitative
Sociology 14(4):349–371.
Namenwirth, J., and R. Weber. 1987. Dynamics of Culture. Winchester, MA: Allen and
Unwin.
Ogilvie, D. M., P. J. Stone, and E. S. Schneidman. 1966. Some Characteristics of Genuine
Versus Simulated Suicide Notes. In The General Inquirer: A Computer Approach to Content
Analysis. P. J. Stone, D. C. Dunphy, M. S. Smith, and D. M. Ogilvie, eds. Pp. 527–535.
Cambridge: M.I.T. Press.
Osgood, Charles. 1959. The Representational Model and Relevant Research Methods. In
Trends in Content Analysis. Ithiels de Sola Pool, ed. Pp. 33–88. Urbana: University of Illinois
Press.
Palmquist, Michael, Kathleen M. Carley, and Thomas Dale. 1997. Two Applications of
Automated Text Analysis: Analyzing Literary and Non-Literary Texts. In A Theoretical Map
for Selecting among Text Analysis Methods. C. W. Roberts, ed. Chap. 10. Mahwah, NJ:
Lawrence Erlbaum.
Patton, Michael Q. 1990. Qualitative Evaluation and Research Methods, 2d ed. Thousand
Oaks, CA: Sage Publications.
Peters, Vincent, and F. West. 1990. Qualitative Analysis in Practice. Nijmegen, Netherlands:
University of Nijmegen.
Piaget, Jean. 1970. Piaget’s Theory. In Carmichael’s Manual of Child Psychology, 3rd ed.,
Vol. 1. P. Mussen, ed. Pp. 703–732. New York: Wiley.
Postel-Coster, Els. 1977. The Indonesian Novel as a Source of Anthropological Data. In Text
and Context. Ravindra K. Jain, ed. Pp. 135–150. Philadelphia: Institute for the Study of
Human Issues.
Potts R., D. Runyan, A. Zerger, and K. Marchetti. 1996. A Content Analysis of Safety
Behaviors of Television Characters—Implications for Children’s Safety and Injury. Journal
of Pediatric Psychology 21(4):517–528.
Prendergast, Guy L. 1971. A Complete Concordance to the Illiad of Homer. Hildesheim,
 Text Analysis 641

Germany: G. Olms.
Price, Laurie. 1987. Ecuadorian Illness Stories. In Cultural Models in Language and
Thought. D. Holland and N. Quinn, eds. Pp. 313–342. Cambridge: Cambridge University
Press.
Quinn, Naomi. 1982. ‘Commitment’ in American Marriage: A Cultural Analysis. American
Ethnologist 9:755–798.
Quinn, Naomi. 1987. Convergent Evidence for a Cultural Model of American Marriage. In
Cultural Models in Language and Thought. N. Quinn and D. Holland, eds. Pp. 173–192.
Cambridge: Cambridge University Press.
Quinn, Naomi. 1992. The Motivational Force of Self-Understanding: Evidence from Wives’
Inner Conflicts. In Human Motives and Cultural Models. Roy D’Andrade and Claudia Stauss,
eds. Pp. 90–126. New York: Cambridge University Press.
Quinn, Naomi. 1996. Culture and Contradiction: The Case of Americans Reasoning about
Marriage. Ethos 24:391–425.
Quinn, Naomi. 1997. Research on the Psychodynamics of Shared Task Solutions. In A
Cognitive Theory of Cultural Meaning. Claudia Strauss and Naomi Quinn, eds. Pp. 189–209.
Cambridge: Cambridge University Press.
Ragin, Charles C. 1987. The Comparative Method: Moving Beyond Qualitative and
Quantitative Strategies. Berkeley: University of California Press.
Ragin, Charles C. 1994. Introduction to Qualitative Comparative Analysis. In The Com-
parative Political Economy of the Welfare State. T. Janowski and A. M. Hicks, eds. Pp.
299–317. Cambridge: Cambridge University Press.
Richards, Tom, and Lyn Richards. 1991. The NUDIST Qualitative Data Analysis System.
Qualitative Sociology 14:307–325.
Rice, G. Elizabeth. 1980. On Cultural Schemata. American Ethnologist 7:152–171.
Richards, Thomas J. Using Computers in Qualitative Research. In Handbook of Qualitative
Research. Norman K. Denzin and Yvonna S. Lincoln, eds. Pp. 445–462. Thousand Oaks, CA:
Sage Publications.
Roberts, Carl W. 1997. A Theoretical Map for Selecting among Text Analysis Methods. In
Text Analysis for the Social Sciences: Methods for Drawing Statistical Inferences from Texts
and Transcripts. Carl W. Roberts, ed. Pp. 275–283. Mahwah, NJ: Lawrence Erlbaum.
Rohner, Ronald. 1966. Franz Boas, Ethnographer of the Northwest Coast. In Pioneers of
American Anthropology. June Helm, ed. Pp. 149–212. Seattle: University of Washington
Press.
Romme, A.G.L. 1995. Boolean Comparative Analysis of Qualitative Data: A Methodological
Note. Quality & Quantity 29(3):317–329.
Rosenberg, Stanley D., P. P. Schnurr, and T. E. Oxman 1990. Content Analysis: A Comparison
of Manual and Computerized Systems. Journal of Personality Assessment 54(1 and
2):298–310.
Rumelhart, David E. 1980. Schemata: The Building Blocks of Cognition. In Theoretical Issues
in Reading Comprehension: Perspectives from Cognitive Psychology, Linguistics, Artificial
Intelligence, and Education. R. J. Spiro, B. C. Bruce, and W. B. Brewer, eds. Pp. 33–58.
Hillsdale, NJ: Lawrence Erlbaum.
Ryan, Gery, and Thomas Weisner. 1996. Analyzing Words in Brief Descriptions: Fathers
642 BERNARD / RYAN

and Mothers Describe Their Children. Cultural Anthropology Methods Journal 8(3): 13–16.
Salinas Pedraza, Jesús, and H. Russell Bernard. 1978. Rc Hnychnyu. The Otomí. Albuquerque:
University of New Mexico Press.
Salinas Pedraza, Jesús. 1997. Saving and Strengthening Indigenous Mexican Languages: The
CELIAC Experience. In Indigenous Literacy in the Americas. Nancy H. Hornberger, ed. Pp.
171–188. Berlin: Mouton de Gruyter.
Sandelowski, Margarete. 1995a. Qualitative Analysis: What It Is and How to Begin. Research
in Nursing and Health 18:371–375.
Sandelowski, Margarete 1995b. Sample Size in Qualitative Research. Research in Nursing and
Health 18:179–183.
Sapir, Edward, and Jeremiah Curtain. 1974 [1909]. Wishram Texts. New York: AMS Press.

Sapir, Edward, and Roland B. Dixon. 1910. Yana Texts. University of California
Publications in American Archeology and Ethnology 9(1). Berkeley: University of California
Press.
Sapir, Edward, and Morris Swadesh. 1978 [1939]. Nootka Texts: Tales and Ethnological
Narratives. New York: AMS Press.
Schank, Robert C., and Robert P. Abelson, 1977. Scripts, Plans, Goals, and Understanding:
An Enquiry into Human Knowledge Structures. Hillsdale, NJ: Lawrence Erlbaum.
Schnegg, Michael, and H. Russell Bernard. 1996. Words as Actors: A Method for Doing
Semantic Network Analysis. Cultural Anthropology Methods Journal 8(2):7–10.
Schnurr, P. P., S. D. Rosenberg, T. E. Oxman, and G. Tucker. 1986. A Methodological Note
on Content Analysis: Estimates of Reliability. Journal of Personality Assessment 50: 601–609.
Schweizer, Thomas. 1991. The Power Struggle in a Chinese Community, 1950–1980: A Social
Network Analysis of the Duality of Actors and Events. Journal of Quantitative Anthropology
3:19–44.
Schweizer, Thomas. 1996. Actor and Event Orderings Across Time: Lattice Representation and
Boolean Analysis of the Political Disputes in Chen Village, China. Social Networks
18:247–266.
Scott, William A. 1955. Reliability of Content Analysis. The Case of Nomimal Scale Coding.
Public Opinion Quarterly 19:321–325.
Segre, Cesare, and T. Kemeny. 1988. Introduction to the Analysis of the Literary Text. John
Meddemmen, tr. Bloomington: Indiana University Press.
Seidel, John, and Udo Kelle. 1995. Different Functions of Coding in the Analysis of Textual
Data. In Computer-Aided Qualitative Data Analysis: Theory, Methods and Practice. Udo
Kelle, ed. Pp. 52–61. London: Sage Publications.
Shapiro, Gilbert. 1997. The Future of Coders: Human Judgments in a World of Sophisticated
Software. In Text Analysis for the Social Sciences: Methods for Drawing Statistical Inferences
from Texts and Transcripts. Carl W. Roberts, ed. Pp. 225–238. Mahwah, NJ: Lawrence
Erlbaum.
Sherzer, Joel. 1990. Verbal Art in San Blas. Cambridge: Cambridge University Press.
Sherzer, Joel. 1994. The Kuna and Columbus: Encounters and Confrontations of Discourse.
American Anthropologist 96:902–925.
 Text Analysis 643

Sherzer, Joel, and Anthony Woodbury. 1987. Introduction. In Native American Discourse:
Poetics and Rhetoric. Joel Sherzer and Anthony Woodbury, eds. Pp. 1–16. Cambridge:
Cambridge University Press.
Silverberg, S. B., S. C. Betts, A. J. Huebner, and S. Cotarobles. 1996. Implicit Beliefs about
Change: A Theory-Grounded Measure Applied to Community Organizations Serving
Children, Youth, and Families. Journal of Sociology and Social Welfare 23(4):57–76.
Sleath, B., B. Svarstad, and D. Roter. 1997. Physician Versus Patient Initiation of Psychotropic
Prescribing in Primary Care Settings: A Content Analysis of Audiotapes. Social Science and
Medicine 44(4):541–548.
Smith, Charles P., S. C. Feld, and C. E. Franz. 1992. Methodological Considerations: Steps
in Research Employing Content Analysis Systems. In Motivation and Personality: Handbook
of Thematic Content Analysis. C. Smith, J. N. Atkinson, D. C. McClelland, and J. Veroff, eds.
Pp. 515–536. Cambridge: Cambridge University Press.
Sohier, R. 1993. Filial Reconstruction: A Theory on Development Through Adversity.
Qualitative Health Research 3(4):465–492.
Spears N. E, J. C. Mowen, and G. Chakraborty. 1996. Symbolic Role of Animals in Print
Advertising—Content Analysis and Conceptual Development. Journal of Business Research
37(2):87–95.
Spevack, Marvin. 1973. The Harvard Concordance to Shakespeare. Cambridge: Belknap
Press.
Spiggle, S. 1986. Measuring Social Values: A Content Analysis of Sunday Comics and
Underground Comix. Journal of Consumer Research 13:100–113.
Spradley, James. 1979. The Ethnographic Interview. New York: Holt, Rinehart and
Winston.
Stone, P. J., D. C. Dunphy, M. S. Smith, and D. M. Ogilvie, eds. 1966. The General Inquirer:
A Computer Approach to Content Analysis. Cambridge: M.I.T. Press.
Strauss, Anselm. 1987. Qualitative Analysis for Social Scientists. Cambridge: Cambridge
University Press.
Strauss, Anselm, and J. Corbin. 1990. Basics of Qualitative Research. Grounded Theory
Procedures and Techniques. Thousand Oaks, CA: Sage Publications.
Strauss, Claudia. 1992. What Makes Tony Run? Schemas as Motive Reconsidered. In Human
Motives and Cultural Models. Roy D’Andrade and Claudia Strauss, eds. Pp. 191–224.
Cambridge: Cambridge University Press.
Strauss, Claudia. 1997. Research on Cultural Discontinuities. In A Cognitive Theory of
Cultural Meaning. Claudia Strauss and Naomi Quinn, eds. Pp. 210–251. Cambridge:
Cambridge University Press.
Swanton, John Reed. 1909. Tlingit Myths and Texts. Washington, DC: Government Printing
Office.
Taylor, Steven J., and Robert Bogdan. 1984. Introduction to Qualitative Research Methods,
2d ed. New York: Wiley.
Tedlock, Dennis. 1972. Finding the Center: Narrative Poetry of the Zuni Indians. New York:
Dial.
Tedlock, Dennis. 1985. Popol Vuh: The Mayan Book of the Dawn of Life and the Glories of
644 BERNARD / RYAN

Gods and Kings. New York: Simon and Schuster.

Tedlock, Dennis. 1987. Hearing a Voice in an Ancient Text: Quiché Maya Poetics in
Performance. In Native American Discourse: Poetics and Rhetoric. Joel Sherzer and Anthony
Woodbury, eds. Pp. 140–175. Cambridge: Cambridge University Press.
Tesch, R. 1990. Qualitative Research: Analysis Types and Software Tools. New York: Falmer
Press.
Trost, J. E. 1986. Statistically Nonrepresentative Stratified Sampling: A Sampling Technique
for Qualitative Studies. Qualitative Sociology 9:54–57.
Viney, L. L. 1983. The Assessment of Psychological States Through Content Analysis of
Verbal Communications. Psychological Bulletin 94:542–563.
Waitzkin, Howard, and Theron Britt. 1993. Processing Narratives of Self-Destructive Behavior
in Routine Medical Encounters: Health Promotion, Disease Prevention, and the Discourse of
Health Care. Social Science and Medicine 36(9):1121–1136.
Weber, Robert. 1990. Basic Content Analysis, 2d ed. Quantitative Applications in the Social
Sciences, Vol. 48. Thousand Oaks, CA: Sage Publications.
Weitzman, E., and M. Miles. 1995. Computer Programs for Qualitative Data Analysis.
Thousand Oaks, CA: Sage Publications.
Werner, Oswald. 1995. The Case for Verbatim Cases. Cultural Anthropology Methods
Journal 7(1):6–8.
White, Douglas R., and M. L. Burton. 1988. Causes of Polygyny: Ecology, Economy, Kinship,
and Warfare. American Anthropologist 90:871–887.
Willms, D. G., J. A. Best, D. W. Taylor, J. R. Gilbert, D.M.C. Wilson, E. A. Lindsay, and J.
Singer. 1990. A Systematic Approach for Using Qualitative Methods in Primary Prevention
Research. Medical Anthropology Quarterly 4:391–409.
Wilson, Holly Skodol, and Sally Ambler Hutchinson. 1996. Methodologic Mistakes in
Grounded Theory. Nursing Research 45(2):122–124.
Wissler, C., and D. C. Duvall. 1908. Mythology of the Blackfoot Indians. Anthropological
Papers of the American Museum of Natural History 2(Pt. 1). New York: Trustees, AMNH.
Wodak, Ruth. 1992. Strategies in Text Production and Text Comprehension: A New
Perspective. In Cooperating with Written Texts. Dieter Stein, ed. Pp. 493–528. New York:
Mouton de Gruyter.
Wolf, Eric. 1964. Anthropology. Englewood Cliffs, NJ: Prentice-Hall.
Woodbury, Anthony C. 1987. Rhetorical Structure in a Central Alaskan Yupik Eskimo
Traditional Narrative. In Native American Discourse: Poetics and Rhetoric. Joel Sherzer and
Anthony Woodbury, eds. Pp. 176–239. Cambridge: Cambridge University Press.
Wright, K. B. 1997. Shared Ideology in Alcoholics Anonymous: A Grounded Theory Approach.
Journal of Health Communication 2(2):83–99.
Young, Robert 1982. Analytical Concordance to the Bible. Nashville: T. Nelson.
Yule, G. U. 1968 [1944]. The Statistical Study of Literary Vocabulary. Hamden, CT: Archon
Books.
Znaniecki, Florian. 1934. Analytic Induction in Sociology. In The Method of Sociology. New
York: Farrar and Rinehart.
 Text Analysis 645

Zuell, C., R. P. Weber, and P. Mohler, eds. 1989. Computer-Assisted Text Analysis for the
Social Sciences: The General Inquirer III. Mannheim, Germany: Center for Surveys, Methods,
and Analysis (ZUMA).