Content Analysis of Media Frames: Toward improving reliability and validity
Matthes & Kohring 2008
Overview
Reviews 5 common content analysis methods and explores their
limitations. Presents an alternative procedure to improve reliablilty and
validiy by studying the grouping of frame elements. Finally applying this
new proposed method to analyse the framing of biotechnology in the New
York Times.
5 Existing Approaches
- Hermeneutic Approach
o Identify frames by interpreting media which links to broader
cultural elements.
o Limitations: Arbitrary/non-explicit frame selection or
extraction, we tend to find the frames we are looking for.
- Linguistic Approach
o Paragraph is unit of analysis, not whole text
o Analyses selection, placement and structure of certain words
and sees words as the building blocks for frames.
o Limitations: very complex and time consuming method
generally unsuitable for large samples of text. The overall
picture can get lost in minute details.
- Manual Holistic Approach
o Frames are generated by a qualitative analysis and then coded
as holistic variables
o Limitations: Once again, frames can appear arbitrarily selected
by the researcher. Once frames are discovered, new ones are
unlikely to arise in the research
- Computer Assisted Approach
o Frame-mapping using word ‘cluster algorithms’. Frames are
computed by the arlgorithm rather than identified by the
researcher.
o Limitations: Only works for digital texts and loses the ‘human
touch’, e.g. the nuance and subtleties which a person can pick
up on and a computer can not.
- Deductive Approach
o Derives frames from the literatre which are then coded. We
postulate generic frames and use factor analysis to confirm
o Limitations: Limited to already existing frames
Proposed Procedure: Frames as clusters of page elements
Basic idea:
- Frame is a certain pattern in a text made up of several elements.
- These elements are previously defined components or devices of
frames.
, - These frames should be split into their separate elements and
coded.
- Then, a cluster analysis of those elements should reveal the frame .
- When some elements group together systematically they form a
pattern which can be identified across several texts. That’s the
frame.
A frame consists of several frame elements and each frame element
consists of several content analytical variables. We group everything using
hierarchical cluster analysis. We use Entman 93 framing functions to
operationalize frame elements, but we can use any operational frame
definition as long as it denotes frame elements.
- Advantages
o Frames are neither identified beforehand nor directly coded
with a single variable.
o Coders do not know which frame they are working on, because
the frames are not coded as single units. This reduces the
impact of coder expectations.
o Newly emerging frames can be easily detected
- Crucial Difference between this method and the others
o Frames are empirically determined rather than subjectively
defined.
Applying the Cluster Procedure: The Framing of Biotechnology in the New
York Times
Methodology
- Comparative analysis of 2 periods of NYT coverage of biotechnology
o 1992-1996 and 1997-2001
o Dolly the sheep was cloned in February 1997, which could
mark a shift in the debate, or have some kind of impact.
- 2 Research Questions
o Can meaningful frames be empirically determined by a cluster
analysis of frame elements?
o Assuming a difference in coverage ‘post-dolly’, how can this
method extract frames from different time periods.
- Keywords (searched in LexisNexis)
o Biotech
o Genetic
o Genome
o DNA
- Randomly chose 100 articles from each of the 10 years studied,
resulting in a sample of 1000 randomly selected articles.
- 2 coders
- Article is unit of Analysis, not paragraph
- For element breakdown see table 1 in the paper.
- A hierarchical cluster analysis (Ward Method) was used
Results
Matthes & Kohring 2008
Overview
Reviews 5 common content analysis methods and explores their
limitations. Presents an alternative procedure to improve reliablilty and
validiy by studying the grouping of frame elements. Finally applying this
new proposed method to analyse the framing of biotechnology in the New
York Times.
5 Existing Approaches
- Hermeneutic Approach
o Identify frames by interpreting media which links to broader
cultural elements.
o Limitations: Arbitrary/non-explicit frame selection or
extraction, we tend to find the frames we are looking for.
- Linguistic Approach
o Paragraph is unit of analysis, not whole text
o Analyses selection, placement and structure of certain words
and sees words as the building blocks for frames.
o Limitations: very complex and time consuming method
generally unsuitable for large samples of text. The overall
picture can get lost in minute details.
- Manual Holistic Approach
o Frames are generated by a qualitative analysis and then coded
as holistic variables
o Limitations: Once again, frames can appear arbitrarily selected
by the researcher. Once frames are discovered, new ones are
unlikely to arise in the research
- Computer Assisted Approach
o Frame-mapping using word ‘cluster algorithms’. Frames are
computed by the arlgorithm rather than identified by the
researcher.
o Limitations: Only works for digital texts and loses the ‘human
touch’, e.g. the nuance and subtleties which a person can pick
up on and a computer can not.
- Deductive Approach
o Derives frames from the literatre which are then coded. We
postulate generic frames and use factor analysis to confirm
o Limitations: Limited to already existing frames
Proposed Procedure: Frames as clusters of page elements
Basic idea:
- Frame is a certain pattern in a text made up of several elements.
- These elements are previously defined components or devices of
frames.
, - These frames should be split into their separate elements and
coded.
- Then, a cluster analysis of those elements should reveal the frame .
- When some elements group together systematically they form a
pattern which can be identified across several texts. That’s the
frame.
A frame consists of several frame elements and each frame element
consists of several content analytical variables. We group everything using
hierarchical cluster analysis. We use Entman 93 framing functions to
operationalize frame elements, but we can use any operational frame
definition as long as it denotes frame elements.
- Advantages
o Frames are neither identified beforehand nor directly coded
with a single variable.
o Coders do not know which frame they are working on, because
the frames are not coded as single units. This reduces the
impact of coder expectations.
o Newly emerging frames can be easily detected
- Crucial Difference between this method and the others
o Frames are empirically determined rather than subjectively
defined.
Applying the Cluster Procedure: The Framing of Biotechnology in the New
York Times
Methodology
- Comparative analysis of 2 periods of NYT coverage of biotechnology
o 1992-1996 and 1997-2001
o Dolly the sheep was cloned in February 1997, which could
mark a shift in the debate, or have some kind of impact.
- 2 Research Questions
o Can meaningful frames be empirically determined by a cluster
analysis of frame elements?
o Assuming a difference in coverage ‘post-dolly’, how can this
method extract frames from different time periods.
- Keywords (searched in LexisNexis)
o Biotech
o Genetic
o Genome
o DNA
- Randomly chose 100 articles from each of the 10 years studied,
resulting in a sample of 1000 randomly selected articles.
- 2 coders
- Article is unit of Analysis, not paragraph
- For element breakdown see table 1 in the paper.
- A hierarchical cluster analysis (Ward Method) was used
Results
