History

Empirical inquiries into communication meanings began with theological studies in the late
1600s, as the Church viewed the printing of nonreligious material as a threat to its authority.
These inquiries have since expanded into numerous areas and form the backbone of
communication research. The history of content analysis involves several stages: quantitative
studies of the press, propaganda analysis during World War II, social scientific uses across
disciplines, computer text analysis and new media, and qualitative challenges. SOME
PRECURSORS Systematic text analysis existed before the term "content analysis" appeared in
English in 1941. Early traces go back to inquisitorial efforts by the Church in the 17th century,
driven by their fascination with the written word. The first documented quantitative analyses
occurred in 18th-century Sweden, centered around a controversy regarding potentially dangerous
ideas in a collection of hymns. This involved scholars listing religious symbols and debating
literal versus metaphorical interpretations. This process of method revision in response to
criticism generated ideas and methodological debates still relevant today. In 1903, Eugen Löbl
published a classification scheme for analyzing the "inner structure of content" of newspapers
according to social functions, contributing to the idea of newspaper science (Publizistik). Max
Weber proposed a large-scale content analysis of the press in 1910, but it did not proceed. Andrei
Markov published a statistical analysis of Pushkin's novel in 1913, but his influence on content
analysis literature was indirect.

QUANTITATIVE NEWSPAPER ANALYSIS The early 20th century saw a rise in mass-
produced newsprint in the United States, creating mass markets and interest in public opinion.
Journalism schools emerged, leading to demands for ethical standards and empirical inquiries
into newspapers. This context, along with a simplistic notion of scientific objectivity, led to
quantitative newspaper analysis. Early studies measured column inches newspapers devoted to
particular subject matters to reveal "the truth about newspapers". This approach attempted to
link profit motive to "cheap yellow journalism" or newspaper content to the growth of crime.
The respect for numbers meant quantified facts were considered irrefutable. This period saw the
development of valuable ideas, such as Tenney's 1912 proposal for a large-scale, continuous
survey of press content. Quantitative newspaper analysis culminated in Malcolm M. Willey's
1926 book The Country Newspaper, which analyzed circulation figures, subject matter changes,
and the social role of newspapers. This approach was later extended to radio, movies, and
television by measuring volumes of coverage in subject categories, and it continues today for
various printed matter.

EARLY CONTENT ANALYSIS The second phase (1930s and 1940s) was influenced by
several factors:

 Social and political problems following the 1929 economic crisis, with mass media
partially blamed.
 New electronic media (radio, TV) challenging newspapers; researchers recognized they
differed significantly from print.
 Political challenges to democracy, like the rise of fascism, linked to the new mass media
(e.g., radio).
  The emergence and public acceptance of behavioral and social sciences, along with their
methods. Sociologists' experience with survey research led to methodological
considerations for content analysis (Woodward 1934).

Interest in social stereotypes entered the analysis of communications. Researchers examined

how groups were represented (e.g., Negroes in the press, wars in textbooks, nationalism in
children's books). The concept of "attitude" from psychology added evaluative dimensions
(e.g., "pro-con," "favorable-unfavorable") to content analysis. This refined the assessment of
bias, with tools like Janis and Fadner's "coefficient of imbalance". Psychological experiments on
rumor transmission also influenced the study of newspaper content. The analysis of political
symbols became important, with Lasswell classifying symbols into categories like "self" and
"others". Factors influencing the transition from journalism-driven quantitative analysis to
content analysis included: involvement of eminent social scientists, the development of
theoretically motivated, specific concepts like stereotypes and propaganda devices, the use of
new statistical tools from other disciplines, and the inclusion of content analysis data in larger
research efforts. The first systematic presentation of these developments appeared in a 1948 text
by Berelson and Lazarsfeld, later published as Berelson's Content Analysis in Communications
Research (1952), which codified the field.

PROPAGANDA ANALYSIS During World War II, content analysis faced a major challenge:
extracting information from propaganda. Before the war, the focus was on identifying
"propagandists" and their use of "tricks" (e.g., name-calling, glittering generalities), often leading
to a "witch-hunt" mentality. Fears stemmed from WWI propaganda use, inter-war demagogues,
and concerns about the effects of new media. By the 1940s, the focus shifted to military and
political intelligence, needing to understand and predict events rather than just identify
propagandists. Two centers emerged: Lasswell's group at the Library of Congress and Speier's
group at the New School. The Library of Congress group analyzed foreign media and worked on
basic methodological issues like sampling, measurement, and reliability. The FCC group
analyzed domestic enemy broadcasts to infer internal conditions and predict events. Alexander L.
George's book Propaganda Analysis (1959a) described and validated the methods used by the
FCC group. Assumptions that propagandists are rational and that meanings differ for different
people led analysts away from the idea of "content as shared" to understanding communicator
motivations and interests. The concept of "preparatory propaganda" was useful for inferring
intended actions. Analysts were able to predict campaigns and assess Nazi elites' perceptions. A
key British prediction was the date of German V weapons deployment, inferred from Goebbels'
speeches. Lessons learned from these applications include:

 Content is not inherent in communications; people read texts differently, and audience
interpretations may differ from authors or analysts. Temporal order, needs, expectations,
discourses, and social situations influence meaning.
 Content analysts must infer phenomena they cannot observe directly, which is often
the primary motivation for using the technique. This includes cases where sources hide
information, phenomena are inaccessible (e.g., attitudes, historical events), or difficult to
assess (e.g., audience learning). The questions asked are the analyst's questions and go
outside the text.
  To interpret texts, analysts need elaborate models of the systems in which
communications occur. Propaganda analysts succeeded when they viewed messages in
the context of users' lives, rather than as inherently meaningful, unit-by-unit entities.
 For specific political information, quantitative indicators can be insensitive and
shallow. Qualitative analyses can also be systematic, reliable, and valid. There was a
challenge to the reliance on simple counting, seen by some as an "immaturity of science"
confusing objectivity with quantification. Although challenged, quantification continued.

CONTENT ANALYSIS GENERALIZED After WWII, content analysis spread to numerous

disciplines. The "massiveness" of available communications continued to attract scholars.
Lasswell conducted a "world attention survey" analyzing political symbols in elite press to test a
"world revolution" hypothesis. Gerbner and colleagues developed "cultural indicators" by
analyzing TV fiction for violence profiles and portrayals of groups.

Psychologists used content analysis to:

 Infer motivational, mental, or personality characteristics from verbal records.

 Analyze verbal data from open-ended interviews, focus groups, and tests like the TAT as
a supplementary technique.
 Study communication processes where content is integral (e.g., Bales' interaction process
analysis of small groups).
 Generalize measures of meaning using tools like the semantic differential (Osgood).

Anthropologists applied techniques to myths, folktales, and riddles, contributing methods like
componential analysis. Ethnographers, after gathering field notes, heavily rely on methods
similar to content analysis. Historians embraced content analysis for analyzing historical
documents, especially where data is numerous and statistical accounts are helpful. Social
scientists used educational materials to study reading processes and societal trends. Literary
scholars applied techniques to identify authors of unsigned documents.The proliferation of
content analysis across disciplines led to both a loss of focus (everything seemed analyzable) and
a broadened scope embracing the essence of human behavior and mediated communication. A
1955 conference brought together scholars from various fields using content analysis. They
noted a convergence: a shift from analyzing "content" to inferring "antecedent conditions" and
a shift from measuring subject matter volumes to counting symbols, then relying on
contingencies (co-occurrences).

COMPUTER TEXT ANALYSIS Interest in computer applications for content analysis

emerged in the late 1950s, driven by the large volumes and repetitiveness of coding. Software
development for literal data processing stimulated new areas like information retrieval,
computational linguistics, and computational content analysis. Early computer-aided CA
included analyzing folktales and exploring systems for political documents. The General
Inquirer system (Stone et al. 1966) was groundbreaking, demonstrating applications in various
fields. Development was also stimulated by work in simulating human cognition and artificial
intelligence. A 1967 conference heavily discussed the use of computers in CA. Early systems
like General Inquirer had English-language biases, leading to more general versions like
TextPack. The use of dictionaries and thesauri became important. While enthusiasm for large
systems faded in the 1980s, text analysis software is now proliferating, fueled by the historically
unprecedented volumes of electronic and digital texts.

Comparisons between computer-based and human-based content analyses showed mixed results,
with some finding low agreement and others satisfactory correlations. A conclusion was that
computers are best used as aids, not replacements for human reading, transcribing, and
translating capabilities. Different opinions exist on the future of computer-based CA. Word
processing software features like spell-check and search/replace also allow for basic, albeit
laborious, word counts and KWIC analyses. The interactive nature of word processing highlights
the potential symbiosis between human understanding and computer scanning. In such
collaborations, human coders may act as translators into categories for the computer to process.
This led to software for computer-aided qualitative text analysis (e.g., NVivo, ATLAS.ti). The
most significant stimulus for computational content analysis is the increasing availability of
text in digital form, bypassing the costly transcription process. Digital text generation (word
processing, email, Internet) has created vast databases of raw textual data. This explosive growth
and corresponding demand for tools like search engines and text managers will benefit
computer-aided CA. The current culture of computation is seen as moving content analysis
toward a promising future.

QUALITATIVE APPROACHES In response to older quantitative methods or shallow results,

various approaches emerged calling themselves qualitative. The distinction between quantitative
and qualitative content analysis is questioned, as all reading is ultimately qualitative, even
when converted to numbers. Qualitative approaches offer alternative ways to systematically
explore texts. Examples include:

 Discourse analysis: Focuses on text above the sentence level and how phenomena are
represented. Studies include racism in the press, ideology in TV, and portrayals of social
groups/events.
 Social constructivist analyses: Focus on discourse to understand how reality is
constituted in human interactions and language, rather than just criticizing
representations. Examples include the construction of emotions, facts, self, and sexuality.
 Rhetorical analysis: Focuses on how messages are delivered and their intended or actual
effects, identifying structural elements, styles, and speech acts.
 Ethnographic content analysis: Does not avoid quantification but encourages analytical
accounts to emerge from reading texts, focusing on situations, settings, styles, images,
meanings, and nuances recognizable by human actors.
 Conversation analysis: Analyzes transcripts of recorded verbal interactions in natural
settings to understand conversational moves and collaborative construction.

These qualitative approaches have roots in literary theory, social sciences (symbolic
interactionism, ethnomethodology), and critical scholarship (Marxist, cultural studies, feminist
theory). They are sometimes labeled interpretive and share characteristics:

 Require close reading of relatively small amounts of text.

 Involve rearticulation (interpretation) of texts into new, analytical narratives accepted
within specific scholarly communities.
  Analysts acknowledge working within hermeneutic circles, where their own
understandings participate.

Content analysis has evolved into a repertoire of methods yielding inferences from various data
forms. It has migrated across fields and clarified methodological issues. After a stagnation period
in the 1970s, it is now growing exponentially, largely due to computer use. Internet search results
indicate its astonishing growth compared to other research techniques.

Conceptual Foundation

Content analysis has a distinct approach rooted in how "content" is conceived.

DEFINITION Content analysis is defined as a research technique for making replicable and
valid inferences from texts (or other meaningful matter) to the contexts of their use.

 As a technique, it involves specialized, learnable procedures not reliant on personal

authority.
 As a research technique, it provides new insights, increases understanding, or informs
actions.
 It is a scientific tool.

Research techniques must be reliable (resulting in replicable findings), meaning different

researchers applying the same technique to the same data should get the same results. They must
also yield valid results, meaning the research effort is open to scrutiny, and claims can be upheld
by independent evidence. These are crucial requirements for content analysis. The term "text" in
the definition includes various meaningful matter beyond written material, such as art, images,
sounds, signs, symbols, and numerical records, provided they speak to someone about
phenomena outside what is directly observable. The crucial distinction is that a text means
something to someone, is produced for others' meanings, and these meanings are essential. Text
serves as a convenient metaphor. Three kinds of definitions exist in the literature:

1. Content is inherent in the text.

2. Content is a property of the source.
3. Content emerges in the process of a researcher analyzing a text relative to a context. Each
type leads to different conceptualizations and procedures.

Berelson's 1952 definition ("objective, systematic and quantitative description of the manifest
content of communication") is an example of the first kind. Our definition subsumes "objective"
and "systematic" under replicability and validity. Systematicity counters selective reading, and
validity requires external criteria. However, "objectivity" is neither measurable nor testable. Our
definition omits "quantitative," acknowledging the success of qualitative methods. Computers
processing words, though quantitative in outcome, rely on fundamentally qualitative algorithms.
It omits "manifest" because this excludes "reading between the lines," which experts do. The
main objection to Berelson's definition is the phrase "description of the manifest content," which
implies content is inside the message. This relies on a container metaphor for meaning,
suggesting messages hold single meanings, which is insufficient. Definitions of the second kind
tie content analysis to inferences about the source. Holsti links describing communication
characteristics ("what, how, to whom") to inferring antecedents ("who, why") and consequences.
However, this may miss communicator intents and fails to acknowledge the analyst's conceptual
contributions. Ethnographic content analysis, acknowledging the analyst's role and linking
analysis to the studied communicators, leans towards the third kind.

EPISTEMOLOGICAL ELABORATIONS Our definition is of the third kind, focusing on the

process and the analyst's contributions. Texts are meaningful, unlike physical events, and
researchers engage with them because they are significant to others, not just the analyst. Analysts
must acknowledge that texts are produced and read by others and are expected to be significant
to them. Content analysts cannot focus only on the physicality of text but must examine how
individuals use texts. The popular measurement model is misleading because it implies inherent
measurable qualities without interpretation by competent users. Six features of texts are relevant
to our definition:

1. Texts have no objective, reader-independent qualities. Seeing something as a text

implies an invitation to read it; texts arise from engagement. Meanings are always
brought to a text by someone.
2. Texts do not have single meanings to be found. They can be read from numerous
perspectives, and different, valid accounts are possible. The belief in a single content is
naive and tied to the container metaphor.
3. The meanings invoked by texts need not be shared. Intersubjective agreement is rare.
Content analysis would be pointless if analysts couldn't read texts differently from others.
The issue arises when expert interpretations ignore users or lack validation criteria.
4. Meanings (contents) speak to something other than the given texts. Communications
inform, evoke feelings, or cause behavioral changes, linking the reading of texts to
phenomena outside the text itself (mental constructions, past/future experiences, hidden
causes). Analysts must look outside physicality to how others use texts and what they tell
them. This highlights a limitation of computer text analysis, which is confined by
programmer conceptions and lacks human understanding of context.
5. Texts have meanings relative to particular contexts, discourses, or purposes.
Analysts choose or construct a context where texts make sense and can answer research
questions. This context is the analyst's hypothesis for how texts came to be and what they
mean. Knowledge of the context includes networks of correlations (text to answers) and
contributing conditions affecting those correlations. Analysts must make their chosen
contexts explicit so their analyses are clear and results validatable.
6. The nature of text demands that content analysts draw specific inferences from a
body of texts to their chosen context. Texts inform someone, narrowing interpretations.
Systematic reading narrows inferences about unobserved phenomena. Drawing
inferences is the centerpiece of this technique.

The inferences in content analysis are abductive, moving across logically distinct domains, from
texts to answers about inaccessible phenomena. Unlike deductive or inductive inferences,
abduction moves from particulars of one kind (texts) to particulars of another kind (unobserved
phenomena). Examples include inferring an author's age from vocabulary or political affiliation
from TV shows. This process is akin to Sherlock Holmes' logic. In the framework, the move
from Texts (Data) to the Answer (Conclusion) is warranted by the Analytical Construct
(Warrant) and backed by the Analyst's knowledge of the Context (Backing).

VALIDATING EVIDENCE Any content analysis should be validatable in principle.

Although difficult in practice when direct observation is absent (e.g., wartime analysis, historical
events), this principle prevents researchers from pursuing questions that lack empirical validation
or rely solely on the researcher's authority. Validation involves showing that the conclusion is
not just an abstraction from the text but corresponds to some independently observable reality.
Ex post facto validation can increase confidence in future analyses using similar categories and
constructs.

Framework

The proposed framework for content analysis emphasizes drawing inferences and clarifies the
analyst's role. It serves prescriptive (guiding research), analytical (examining published work),
and methodological (evaluating ongoing work) purposes.The framework consists of key
components:

 Body of Text (Data): The available material, taken as givens. These are texts intended
for others' interpretation.
 Research Question: The target of the analyst's inferences, delineating possible answers
about currently inaccessible phenomena. Answers should be validatable in principle.
 Context: The conceptual environment constructed by the analyst, within which texts
make sense and relate to the research questions. It embraces the analyst's relevant
knowledge.
 Analytical Construct: Operationalizes the analyst's knowledge of the context,
particularly the network of correlations between texts and answers, and contributing
conditions. These function like rules of inference from texts to answers. Analytical
constructs must model the chosen contexts.
 Inferences: The move from texts to the probable answer to the research question. These
are typically abductive.
 Validating Evidence: External evidence that can, in principle, confirm or invalidate the
inferences made.

Analysts must acknowledge that their reading of texts differs from others' and is not the only
legitimate one. A key strategy to avoid bias from sources' potential manipulation is to focus on
textual features or categories of which sources are unconscious or cannot control. Research
questions are crucial for efficiency and empirical grounding. The context is the analyst's
construction, embodying the knowledge used to interpret texts. Analytical constructs are the
computable representation of this knowledge. Inferences are the core accomplishment.
Validation, even if only in principle, is necessary to ground the analysis in reality and allow for
evaluation.

Examples

The text provides examples illustrating the definition and framework:

  Wartime analysts inferring popular support or planned military actions from enemy
domestic broadcasts, understanding the communication network as the context.
 Historians reconstructing past events by inferring from surviving documents, placing
them within historiographical contexts.
 Psychological researchers inferring psychological variables from personal documents,
diaries, or letters, using theories of correlation between language and psychological traits
as context.
 Analysis of open-ended interview and focus group transcripts to explore participant
conceptions, with the researcher's theory or an adapted theory providing the context for
coding and analysis.
 Mass communication researchers analyzing mass media messages to infer
communicator conceptions, media biases, audience perceptions, or societal trends,
requiring a framework and context to handle large volumes and diverse questions. The
issue of purely descriptive analyses failing to make their underlying norms and contexts
explicit is discussed.
 Commercial uses, such as using word-association databases as context to infer
associations for products, or analyzing media coverage of publicity to infer its
effectiveness and journalists' attitudes.

These examples highlight that content analysis is used to infer phenomena not directly
observable and that the context constructed by the analyst is crucial for making sense of the
data and validating findings.

Contrasts and Comparisons

Content analysis is contrasted with other social research techniques based on four features:

 Unobtrusive technique: Content analysis is nonreactive, unlike experiments, interviews,

surveys, and projective tests, which can be contaminated by subjects' awareness of being
observed or task artificiality. This avoids biases that can occur when sources might
manipulate data if they knew the analysis methods.
 Handles unstructured matter as data: Unlike structured methods (surveys, structured
interviews) that impose predefined choices and ignore individual voices, content analysis
copes with diverse, unstructured texts, thereby preserving the conceptions of the data's
sources.
 Context sensitive: Content analysis acknowledges that data are texts read by others and
make sense within contexts. Context-sensitive methods yield inferences more relevant to
the users of the analyzed texts. Context-insensitive methods like experiments or surveys
disembody data from their original context.
 Can cope with large volumes of data: While qualitative methods often rely on small
samples, content analysis, especially with explicit procedures and computer aid, can
process vast amounts of text, far exceeding what individuals can handle reliably. This
explicit systematicity allows for the use of multiple coders and computer software. The
exponential growth of digital text availability has significantly increased the volume of
data accessible for content analysis. This brings CA closer to large population studies but
without the obtrusive and decontextualizing qualities of methods like surveys. The
challenge is shifting from data access to the need for strong theory, methodology, and
software.