Lecture 1: Introduction
What is more popular in communication research? Corpus or Experiments since Content analysis cannot
answer cause and effect questions.
Corpus (Content) Analysis: Content analysis means counting what can be seen using counting and
categorizes for basic analyses and insights on speech, text, or screen. (Treadwell, 2003)
Content analysis is a research technique for objective, systematic, and quantitative description of
the manifest content of communication. (Berelson, 1952)
- Objective: Valid, reliable, and clearly defined unit of analysis. Another researcher can replicate.
- Systematic: Code/analyze all relevant aspects (rigid).
- Quantitative: Categorizing and counting instances.
- Manifest content of communication: Content that is only tangible, and observable.
Content analysis in 8 steps:
1. Formulate a research question/hypothesis.
2. Determine what kind of content you will analyze.
3. Create your corpus. Sample from your population.
4. Decide what you are going to code.
5. Establish how you are going to code the data. Categorizations.
6. Annotate the corpus. Define & Code.
7. Count occurrences.
8. Report results.
Questions to ask when doing a research:
What content will you analyze?
What will be your corpus?
What are you going to code and how will you code it?
Lecture 2: Corpus Analysis I
Rhetorical analysis (words, fragments) **
Narrative analysis (characters, storylines)
Discourse analysis (words, language, themes) **
Semiotic analysis (deeper layer, meanings)
,In Communication Research, generally Rhetorical and Discourse analysis are used among qualitative
analyses. So, it is possible to combine corpus research with these qualitative analyses to get a better
understanding.
Neuendorf’s definition of content analysis
“… a summarizing, quantitative analysis of messages that relies on the scientific method (including
attention to objectivity-intersubjectivity, a priori design, reliability, validity, generalizability, replicability,
and hypothesis testing) and is not limited as to the types of variables that may be measured or the
context in which the messages are created or presented.”
Neuendorf defines content analysis as “scientific”. What is meant by Scientific?
- Objective: avoid any kind of bias from the interpreter
- A Priori Design: it means before the fact – deciding on what you are going to code/do before
you collect data.
- Reliability: you want to have high level of agreement between multiple coders. inter-coder
reliability (among human coders)
- Validity: weather you are measuring what you say you are measuring
- Generalizability: to be able to say something more general about other cases, about the bigger
picture.
- Replicability: finding the same results if the study is conducted again
- Hypothesis testing: you generate a research question to hypothesis based on literature and you
collect data to test this hypothesis by using statistical tests – not so common with content
analysis.
Quantitative: conclusions based on numerical data.
Summarizing: seeking to make more general conclusion, a summary of the data rather than mentioning
details.
Not limited: applicable to many contexts
All message characteristics can be studied (both manifestant and latent – deeper levels - data)
Superficial - deeper layer(s) Berelson!
Text, speeches, websites, monologues
Types of corpus research:
Descriptive: describe what occurs and how often. Limits the conclusions because you do not test any
hypothesis.
Inferential: draw conclusions about sources or receivers of the message based on results that have not
been empirically supported.
, For example: text writers argue that exclamation marks in advertisements are persuasive
(because they occur often). So you make an influence on exclamation marks being persuasive,
however, It would require additional data to support that conclusion.
Psychometric: research on human output and link them to psychological characteristics
For example: people write shorter sentences when they are tense. Linking message
characteristics with type of mood.
Predictive: predict effects based on occurrences (combination of methods needed)
For example: analyze message characteristics of ads and combine with survey data about
readers’ appreciation to predict why one ad is more successful than others.
** Corpus research is often done in combination with other types of research: triangulation.
The ‘Units’ section
Units important for creating and analyzing corpus.
Unit: component serving as…
… a basis on which variables are measured –> unit of data collection. (In essence what do you collect)
… a basis to determine population and sampling –> unit of sampling. (more specific)
… a basis for reporting unit of analysis (what you wonder)
Example: adverts
Unit of data collection: print adverts
Unit of sampling: 5 most-read NL magazine from past 5 years.
Unit of analysis: adverts
Lecture 3: Corpus analysis II: Sampling
How do you create a corpus?
It is oftentimes impossible to study EVERYTHING, there is too much data available.
Therefore, researcher defines population(s). Once you know what your population and if it is
too big, you can study sample(s) from that population (sampling).
Population: “set of units to which the researcher wishes to generalize.”
What is more popular in communication research? Corpus or Experiments since Content analysis cannot
answer cause and effect questions.
Corpus (Content) Analysis: Content analysis means counting what can be seen using counting and
categorizes for basic analyses and insights on speech, text, or screen. (Treadwell, 2003)
Content analysis is a research technique for objective, systematic, and quantitative description of
the manifest content of communication. (Berelson, 1952)
- Objective: Valid, reliable, and clearly defined unit of analysis. Another researcher can replicate.
- Systematic: Code/analyze all relevant aspects (rigid).
- Quantitative: Categorizing and counting instances.
- Manifest content of communication: Content that is only tangible, and observable.
Content analysis in 8 steps:
1. Formulate a research question/hypothesis.
2. Determine what kind of content you will analyze.
3. Create your corpus. Sample from your population.
4. Decide what you are going to code.
5. Establish how you are going to code the data. Categorizations.
6. Annotate the corpus. Define & Code.
7. Count occurrences.
8. Report results.
Questions to ask when doing a research:
What content will you analyze?
What will be your corpus?
What are you going to code and how will you code it?
Lecture 2: Corpus Analysis I
Rhetorical analysis (words, fragments) **
Narrative analysis (characters, storylines)
Discourse analysis (words, language, themes) **
Semiotic analysis (deeper layer, meanings)
,In Communication Research, generally Rhetorical and Discourse analysis are used among qualitative
analyses. So, it is possible to combine corpus research with these qualitative analyses to get a better
understanding.
Neuendorf’s definition of content analysis
“… a summarizing, quantitative analysis of messages that relies on the scientific method (including
attention to objectivity-intersubjectivity, a priori design, reliability, validity, generalizability, replicability,
and hypothesis testing) and is not limited as to the types of variables that may be measured or the
context in which the messages are created or presented.”
Neuendorf defines content analysis as “scientific”. What is meant by Scientific?
- Objective: avoid any kind of bias from the interpreter
- A Priori Design: it means before the fact – deciding on what you are going to code/do before
you collect data.
- Reliability: you want to have high level of agreement between multiple coders. inter-coder
reliability (among human coders)
- Validity: weather you are measuring what you say you are measuring
- Generalizability: to be able to say something more general about other cases, about the bigger
picture.
- Replicability: finding the same results if the study is conducted again
- Hypothesis testing: you generate a research question to hypothesis based on literature and you
collect data to test this hypothesis by using statistical tests – not so common with content
analysis.
Quantitative: conclusions based on numerical data.
Summarizing: seeking to make more general conclusion, a summary of the data rather than mentioning
details.
Not limited: applicable to many contexts
All message characteristics can be studied (both manifestant and latent – deeper levels - data)
Superficial - deeper layer(s) Berelson!
Text, speeches, websites, monologues
Types of corpus research:
Descriptive: describe what occurs and how often. Limits the conclusions because you do not test any
hypothesis.
Inferential: draw conclusions about sources or receivers of the message based on results that have not
been empirically supported.
, For example: text writers argue that exclamation marks in advertisements are persuasive
(because they occur often). So you make an influence on exclamation marks being persuasive,
however, It would require additional data to support that conclusion.
Psychometric: research on human output and link them to psychological characteristics
For example: people write shorter sentences when they are tense. Linking message
characteristics with type of mood.
Predictive: predict effects based on occurrences (combination of methods needed)
For example: analyze message characteristics of ads and combine with survey data about
readers’ appreciation to predict why one ad is more successful than others.
** Corpus research is often done in combination with other types of research: triangulation.
The ‘Units’ section
Units important for creating and analyzing corpus.
Unit: component serving as…
… a basis on which variables are measured –> unit of data collection. (In essence what do you collect)
… a basis to determine population and sampling –> unit of sampling. (more specific)
… a basis for reporting unit of analysis (what you wonder)
Example: adverts
Unit of data collection: print adverts
Unit of sampling: 5 most-read NL magazine from past 5 years.
Unit of analysis: adverts
Lecture 3: Corpus analysis II: Sampling
How do you create a corpus?
It is oftentimes impossible to study EVERYTHING, there is too much data available.
Therefore, researcher defines population(s). Once you know what your population and if it is
too big, you can study sample(s) from that population (sampling).
Population: “set of units to which the researcher wishes to generalize.”
