This paper was presented at the Association for Education in Journalism and
Mass Communication in Toronto, Canada, August 2004.
        If you have questions about this paper, please contact the author
directly. If you have questions about the archives, email
[log in to unmask] For an explanation of the subject line, send email to
[log in to unmask] with just the four words, "get help info aejmc," in the
body (drop the "").
(Oct 2004)
Thank you.
Elliott Parker
************************************************************************

The role of scene in framing a story:
An analysis of scene's position, length and dominance in a story


By

Yun Jung Choi
Doctoral Student
Syracuse University
S. I. Newhouse School of Public Communications

[log in to unmask]
121 Lafayette Rd. # 521
Syracuse, NY 13205

* Student Paper






A paper submitted to the Communication Theory and Methodology Division of
the annual convention of the Association for Education in Journalism and
Mass Communication (AEJMC), Toronto, Canada, 2004.






The role of scene in framing a story:
An analysis of scene's position, length and dominance in a story



Abstract



This paper proposes a "scene" as a micro unit of analysis in broadcast news
by exploring the concept of a scene and describing the role of a scene in
broadcast news stories. A scene is defined as a sub-unit of a broadcast
news story that is composed of several shots with a unifying theme, a
character, or a place. This study explicates the concept of scene by
examining the relationship between scene position, scene length, and
dominance of scene frame in a story, and scene valence with the overall
story frame and valence. First, the study suggests that early, middle and
ending positions of scenes do not affect the overall evaluation of the
story. Frames and valence expressed in early scenes of a story do not
particularly become the overall frame and valence of the story. Second, the
study suggests that frames featured longer scenes do not particularly
influence the overall story frame. Third, the study reveals that the
dominance of scene frame and valence, defined as the proportions of a scene
frame or valence in a whole story, has impacts on the overall frame and
valence of a story. This means that frames and valence discussed in many
scenes in a story are likely to be the overall story frame and valence.








In studying newspaper articles, researchers adopt a paragraph, a sentence,
or a word as the unit of analysis because a whole newspaper article may
contain several sub-topics and both positive-negative valences. When
broadcast news stories are analyzed with a story unit, micro attributes of
news stories are often neglected. However, most broadcast news stories are
often analyzed as a whole story unit in communication research (Newhagen &
Reeves, 1992). Yet, broadcast news stories are also composed of complex
sub-topics and conflicting ideas. In reporting social issues, reporters try
to make sure that at least two sides of an issue are presented equally. If
a conservative view is presented, the liberal view likely will be
presented. Similarly, in covering a political campaign, reporters give
equal attention to both Democratic and Republican candidates running in the
race.
One reason broadcast news stories are studied with a whole story unit is
that broadcast news stories are difficult to break down into smaller units.
Broadcast news has a complex structure that is composed of both visual and
audio features. This paper proposes a "scene" as a micro unit of analysis
in broadcast news by exploring the concept of a scene and describing the
role of a scene in broadcast news stories. Scenes are sub-units of a
broadcast news story that are composed of several shots with a unifying
theme, a character, or a place. Several scenes representing different
subtopics compose a whole story.
Entman's (1993) four roles of framing – define problems, diagnose the
causes of the problem, make moral evaluations, and suggest remedies – are
used as the methodological framework for the study. Broadcast news stories
were content analyzed based on Entman's frames in both a story unit and a
scene unit to see how each scene's position, length, and dominance within a
story are related to framing of a whole broadcast news story. In addition,
how positive, negative and neutral valence of scenes are related to the
overall valence of a story is examined.
This study is important because this study provides an empirical example
for those who want to study micro attribute of broadcast news with smaller
units. This study suggests a more sophisticated methodology to examine
visual messages. The results of this study can be also applied to examining
other visual messages such as documentaries and news magazines.


Theory

First, the concept of scene should be explicated. Scene is a vague term.
This paper tries to contribute in defining and establishing the term in
communication research by examining the roles of scenes in framing a story.

The Concept of Scene
If newspaper stories are composed of characters, words, sentences and
paragraphs, then broadcasting news stories are composed of frames, shots
and scenes. The most basic unit of a television visual analysis is a frame.
According to Iedema (2001), a frame is a salient or representative still of
a shot.
The next level is the shot. Hundreds of frames compose a shot, which is an
uncut camera action. In a shot, the camera movement is unedited. Shot
depicts a continuous view filmed by one camera without interruption
(Mascelli, 1965).
The next level is a scene, which is usually comprised of several shots. A
shot is incomplete and meaningless when considered out of context. It is
the scene that contains a complete unit of narration. According to Metz, "a
scene constructs a unit still experienced as being 'concrete': a place, a
moment in time, an action, compact and specific" (1974, pp 127-128).
Whitaker also states that a scene is "the smallest level of film form that
possesses that Aristotelian wholeness needed before one can dare to make
meaningful statements and criticisms" (Whitaker, 1974, p. 48). A voiceover
sentence of a reporter cannot be completed in a shot, unless the shot is
more than a few minutes long, which is unusual in a news story. Therefore,
scene is the smallest unit that contains a meaningful narration and
contains arguments and perspectives of a news story.
Scene has been defined differently according to different literature. Many
film and production literature defines the term scene differently. In film
studies, scene is simply defined as "a series of shots (or a shingle shot)
that take place in a single location and that deal with a single action"
(Monaco, 1977, p. 427).
Scene is sometimes referred as sequence or montage when defined as a broad
concept. Green (1969) defines a sequence as a series of shots depicting the
details of an event in their approximate chronological order of occurrence.
This is the most typical news film story where a progression of an event is
demonstrated. Many crime reports consist of matter-of-fact chronological
incidents.
According to Green, a montage is a series of shots not necessarily related
in time or space, but the totality of montage creates a specific effect. A
montage of trash cans, rats, leaking plumbing, falling plaster and a
homeless man may convey the reality of the urban ghetto, but not one of the
shots has any connection with any other, except that philosophical
connection created by their cinematic juxtaposition. Whitaker also describe
montage as "the arrangement in time of the film's linguistic elements so
that they interact to create a total message that is greater than, or
different from, the sum of messages considered separate." (Whitaker, 1970,
p. 128). Montage states a theme, demonstrates a condition, and establishes
a mood.
Green (1969) further proposes three characteristics of montage. First,
montage is characterized by timelessness. There are no time relations among
various shots that compose a montage. Second, a story does not progress in
montage. There is no beginning, middle, and end to the story, no
development of a line of action, climax and denouement. For example, a
juxtaposition of shots showing Sadam Hussein shooting a rifle in a balcony,
a military parade with tanks, thousands of soldiers marching in uniforms,
and a nuclear facility clearly creates a montage of "weapons of mass
destruction," yet there is no time relation among the shots and no story
development.
  Third, montage may be composed of monotonous camera shots without any
variation in their sizes and angles.  Montage may lack the pattern of shots
that is often found in sequence. A long sequence of monotonous close shots
can compose a montage. Such a calculated monotony would produce a specific
artistic purpose. Three characteristics, lack of time compression, lack of
progression, and monotony of shots make up unique features of montage.
In this study, scene is defined as a concept that both embody the concept
of sequence and montage. Scene is both a series of shots depicting a single
action taking place in a single space, and a montage which depicts a single
concept, a theme or an idea without the limitation of time and space.
There is no fixed format in a news story structure so that there are many
variations in scenes that construct a news story. The following attributes
of scenes are proposed in this study.
1. Each scene is focused on one action or a theme. In a news story about a
political candidate, if the candidate's stance on economy issues, foreign
policies, and environmental issues are discussed, each three sub-topics
constitute three separate scenes. In addition, the candidate's past voting
history can be discussed in one scene, and criticisms about the candidate
can be discussed in one scene. In this case, the whole story is comprised
of five scenes representing five sub-topics. Figure 1 illustrates the scene
structure.

[Insert Figure 1 about here]

2. Scenes with a same theme or idea may appear more than once in a story.
In a story where two conflicting views are discussed, two opposing views
can be shown in an alternating sequence. In a story about an abortion
issue, a scene with pro-life supporters demonstrating in front of an
abortion clinic is usually followed by a scene where pro-choice supporters
are demonstrating in front of the Congress. The next scene may be an
interview with a pro-life advocate, and the fourth scene may be an
interview scene with a pro-choice advocate. Figure 2 illustrates the scene
structure within a story.

[Insert Figure 2 about here]

3. There is no limit on the number of scenes in a news story. Some news
stories may be comprised of two scenes whereas some news stories are made
up with more than eight scenes.
4. Not every scene is equal in length. Some scenes last longer than others
within a story because some subtopics are treated more importantly than
others within a story. In a political campaign story where candidates are
introduced, the leading candidate is usually featured in the longer scenes.
5. Scene has a visual continuity. A scene is composed of one theme, which
is often expressed in continuity in visuals. Visuals should flow smoothly
within a scene without interruption or disturbance.

Broadcast News Structure
Literature on news structure also demonstrates that a broadcast news story
is not constructed with one unifying theme, but rather with several scenes
with indiscrete ideas and themes that compose a dramatic unity. Newspaper
reporters write most of their stories in the "inverted pyramid" style,
putting the most important information at the beginning of the story while
the least important toward the end (Papper, 1995; Stephens, 1993). However,
the inverted pyramid style has no place in broadcast journalism. Television
reporters construct their stories so that they have a beginning, middle and
ending, more like a novel.
There is no fixed story format for broadcast news stories. It is generally
believed that the lead is the most important sentence in broadcast news,
because it draws viewers' attention, sets the tone for the story and often
includes many of the basic facts of the story. (Keller & Hawkins, 2002;
Stephens, 1993). The second most important sentence in broadcast news is
the last sentence. The last sentence should have a clear, strong closing
(Papper, 1995). Similarly, Green (1969) suggests a broadcasting story
structure that proceeds from climax to causes to effects. He states that in
broadcast news stories, the climax is placed near the beginning with causes
corresponding to the rising of action of a play following. The last part of
the story corresponds to the denouement of the play where effects and
consequences are stated.
Keller and Hawkins (2002) propose three formats for typical story patterns
– chronological order, the main point, and the present, past and future
order. Some stories are told best in a chronological order, the order in
which they actually happened, while some stories are best written from the
main point of the story. Sometimes news stories are often told in the
present, past and future order format, which surveys the present
situations, reviews the past, and looks into the futures. A typical fire
story would be an example. Fire fighters are still continuing their battle
with the flames. The fire broke out last night. Officials say they hope to
have the blaze under control by early afternoon.
Coates (1994) also suggests three patterns of basic television news
coverage in terms of combining sound bites and reporter segments. The first
pattern is the reporter/sound bite/reporter sequence, in which reporter
voice over sentence telling what it is all about, a bite from a key figure
in the story, and then the reporter's voice over. This pattern is often
found in straightforward, uncomplicated stories, such as a news conference,
a speech, a fatal accident. The second pattern is the reporter/ sound bite/
reporter/ sound bite/ reporter sequence, which is often used in presenting
two sides of a controversial issue or two aspects of an event. The first
reporter and the sound bite represent one side of view, the second reporter
and the sound bite represents the second view points, and then the last
report summarizes the event. The third pattern is the reporter/ sound bite/
sound bite/ reporter sequence, often used when getting two comments about
something that does not involve two sides of a story. For example, in a
story about a charity auction, reporters can interview several people about
the events.
Although literature on broadcasting journalism structure does not use the
term scene, it acknowledges that a broadcast news story is a carefully
planned composition of several scenes or sub-segments.

Framing
In order to understand the attribute of a scene, framing is adopted as the
theoretical framework for this study. The general idea of framing theory is
that reporters underscore some aspects of reality and exclude other aspects
of reality in the news production process. Gitlin (1980) defined media
frame as "persistent patterns of cognition, interpretation and
presentation, of selection, emphasis, and exclusion, by which symbolic
handlers routinely organize discourse, whether verbal or visual" and frames
make possible for news reporters to "process a large amount of information
quickly and routinely and package the information for efficient relay to
their audience" (p. 7). Similarly, Tankard, Hendrickson, Silberman, Bliss,
and Ghanem (1991) view framing as "a central organizing idea for news that
supplies a context and suggest what the issue is through the use of
selection, emphasis, exclusion, and elaboration" (p. 5). Above definitions
of framing point out that framing involves two processes – selection and
organization. Journalists frame a story by selecting some aspects of an
issue over others and by making some of them more salient than others
within a news story.
In this perspective, size and positioning of information becomes an
important way to frame news stories. Entman (1991) also notes that the
essence of framing is sizing. Entman adds "aside from the words and images
used to depict the event, how much material on the event is available, and
how prominently is it displayed? This is the event's importance" (p. 9).
Ghanem(1997) also explains that "framing mechanism" is best captured by
placement, size, pictures, pull quotes, and other aesthetic devices. In the
cases of broadcast news, Watkins (2001) writes that two measurable indices
of importance in television news are story length and placements. A typical
television news package runs between one and half to two minutes on
average. In the case of important news stories, story length may be
extended to a longer format, and news items perceived as important are
placed in lead segments.

Entman's Four Roles of Framing
"To frame," Entman (1993, p.52) writes, "is to select some aspects of a
perceived reality and make them mire salient in a communicating text, in
such a way as to promote a particular problem definition, causal
interpretation, moral evaluation, and /or treatment recommendation for the
item described." In order to study how frames of each scene influence the
overall frame of the whole story, Entman's definition of the four roles of
framing – define problems, diagnose the causes of the problem, make moral
evaluations, and suggest remedies – is used for the study. The four roles
of framing provide a very comprehensive structure that can be applied to
study various news topics because the four frames are commonly used in news
stories.
In a study that examines how different frames were used by the U.S. and
South Korean network news in covering President Bush's visit to three Asian
countries, Entman's four frames was used as the methodological framework
(Lee & Choi, 2002). This study also uses a scene as the unit of the
analysis. The study shows that the U.S. network news used the Axis of Evil
frame, while the Sunshine frame was used in South Korean networks. By
applying Entman's four definitions of framing, the Axis of Evil frame was
divided into "rogue state (problem definition)," "causing terror (causal
interpretation)," "bring justice (moral evaluation)," and "war or sanction
(treatment recommendation)" sub-frames and the Sunshine frame was divided
into "negotiation partner (problem definition)," "isolated from world
(causal interpretation)," "anti-war (moral evaluation)," and "peace talk
(treatment recommendation)" sub-frames. Each scene was coded according to
the eight sub-frames. Among the sub-frames, the U.S. also often used the
"rogue state" sub-frame while South Korean networks often used the "peace
talk" sub-frame, which represent each country's current policy toward North
Korea.

Valence of News
Frames are not neutral but with positive or negative connotations.
Beinstein and Beinstein (1995) talks about how Gingrich led GOP Political
Action Committee taught fellow Republican candidates fro Congress to use
optimistic and positive words (e.g. share, opportunity, commitment, and
mobilize) to describe their visions and to use negative words (e. g.
collapse, destructive, and hypocrisy) to frame the policy and record of the
Democrat opponents.
In news studies, valence is often defined as how positively and negatively
a story is presented to viewers. One effective way to measure valence is to
locate the amount of conflict in a story – the higher the conflict, the
higher the valence score (Auh, 1977). Kiousis (2004) assessed the valence
in his content analysis study by asking "Ask yourself if, from the point of
view of the typical newspaper reader, the story is positive, negative, or
neutral?" to coders in the code-sheet instruction.


Hypotheses



Framing theory asserts that positioning is one way to frame a story. In a
story about two political candidates running in a race, if one candidate is
featured one the previous scenes and the other candidates on later scenes,
the candidate featured in earlier scenes dominates the overall impression
of the story. Broadcast news structure studies also suggest that lead
sentence in a story is the most important sentence because it must attract
viewer's attention. Therefore earlier scenes in a broadcast news story are
more likely to have the overall frame of the story than later scenes.
In this study, scene frame refers to frames featured in scenes, and story
frame refers to the frame of the whole story. For each story there is only
one story frame, but a story may be composed of several scenes with several
frames. Some scene frames may match the story frame, but not all scene
frames correspond with the story frame.

H1: Frames featured in early scenes in a broadcast news story are more
likely to be the overall story frame than frames featured in later scenes.

Even though a same number of scenes is allocated for two frames in a story,
if scenes about one frame are longer than other, the whole story is likely
to have the same frame as the scenes with longer duration. For example, in
a story where pro-war and anti war frames are discussed, even though the
two frames are featured in equal number of scenes, if the scenes with
pro-war frame are longer than the scenes with anti-war frame, then the
overall story is likely to be evaluated as having the pro-war frame.

H2: The longer the scene, the more likely the scene frame is to be the
overall frame of the story.

Scene's dominance in a story also plays an important role in framing.
Dominance of scene frame is defined as the proportion of a particular frame
in a whole story that is featured in scenes. If a broadcast news story is
comprised of 6 scenes, and four out of the six scenes are allocated to one
candidate and two to the other, the overall story frame is likely to be
balanced toward the candidate who takes up the larger portion of the story.

H3: The more dominant a scene frame is, the more likely the scene frame is
to be the overall story frame.

Frames can also have positive, neutral and negative valence. Similar to the
four dimensions of frames, the valence of frames in scenes also impacts the
dominant valence of the story. In this study, scene valence refers to
frames featured in scenes, and story valence refers to the frame of the
whole story.

H4: The valence of early scenes in a broadcast news story is more likely to
be the overall valence of the story than the valence of later scenes.

H5: The longer the scene, the more likely the scene valence is to be the
overall valence of the story.

H6: The more dominant scene valence is, the more likely the scene valence
is to be the overall story valence.



METHODS

A whole broadcast news story was coded based on Entman's four ways of
framing and valence. The same news stories were broken down into scene
units, and then individual scenes were coded based on the same variables.
In addition, scenes' positions within a story, duration, dominance of
frames, and dominance of scene valence, whether a scene frame matches the
story frame, and whether scene valence matches the story valence were coded
in order to examine the relationship between a whole story and scenes.

Unit of analysis
A scene in a broadcasting news story is adopted as the unit of analysis in
this study. Broadcast news stories had to be broken down into scene units
first. The scene unit was divided based on both visual and auditory
features of broadcast news. Dividing a news story into scene units involves
deconstructing news story structures established by reporters. In
structuring news stories, reporters carefully plan and arrange small
segments of sub-topics. Dividing stories into scene units just involves
understanding reporters' story logic and then disintegrating reporters' works.
The following guidelines were used in dividing scenes. First, whenever
there was a shift in theme, place, action or character in a story, it was
considered as the beginning of a new scene. Second, a beginning of a new
scene is usually indicated by a transition shots. The most routine
transition shots are wide shots that establish new settings and introduce
viewers to a new places and themes. However, a beginning of a scene does
not always accompany establishing shots. Third, it was not considered as a
scene unless it has at least one complete sentence of reporters' voice
over. Reporters try to make the picture and script correspond therefore a
paragraph or a sentence in reporter's narration usually finishes within one
scene. Fourth, the anchor shots in the studio, reporter's standing up in
the location, and sound bite interviews were considered as separate scenes.

Sampling
A one-month period of broadcast news stories aired on three networks' prime
time news were sampled from February 1 to February 2004. News stories from
the three networks, NBC, ABC and CBS were collected in alternation based on
a three-day term such that Day 1 would be NBC, Day 2 would be ABC, and the
Day 3 would be CBS. Out of the collected news stories, only 'package' news
stories were examined. Stories that are introduced by an anchor and told by
reporters on a separate segment are called the package news. Stories that
are read by anchors were excluded in the study because they are short and
usually not carefully structured by reporters. As the result, a total
number of 87 stories were sampled.

Entman's four ways of framing
Problem definition: stories and scenes defining an event, a social problem,
and an issue were considered as the problem definition frame. Stories
focused on the three Ws, what happened, where it happened and who did it
were coded in this category.
Causal interpretation: Stories and scene concentrating on explaining the
causal interpretations of issues or social phenomena were classified in
this category. In depth investigations of why and how some incidents, such
as depression, high unemployment rate, and low voter-turn-out rate, happen
belong to this category.
Moral evaluation: Stories and scenes making moral judgments on issues were
considered as having the moral evaluation frame. Stories and scenes that
try to justify an action or criticize some issues belong to this category.
For example, showing pictures of tortured victims in the Sadam Hussein's
regime is making moral evaluation of the war in Iraq, and frames the war as
a morally justifiable incident. Similarly, when a politician is featured in
a story as an incompetent candidate who misses many votes and does not keep
his or her campaign promises, the story has a moral judgment frame.
Treatment recommendation: Stories and scenes suggesting remedies for a
problem, and providing future prospects were included in this category.
Many stories provide solutions to problems at the end. A typical fire story
ends with comments about improving the fire alarm systems and giving tips
to avoid future fires. Many murder stories talks about improving the
security in the neighborhood and the reinforcements of the police. Also
those stories and scenes that predict outcomes and looking toward the
future were included in this category.

Valence in scenes
Stories and scenes were coded as positive, neutral and negative according
to their affective valence of reporting. If objects, themes or ideas in a
story or a scene are negative in nature or, portrayed negatively, the story
or the scene was coded as negative. When objects, themes, or ideas featured
in a story or a scene are positive in nature, or portrayed positively in
reports, the scene is coded as positive. Those stories and scenes with
neither positive nor negative valence were coded as neutral.

Scene frames matching the story frame
Scenes with frames that match the story frame were coded as 1, and scenes
with frames that do not match the story frame were coded as 2.

Scene valence matching the story valence
Scenes valence that matches the story valence were coded as 1, and scenes
with valence that do not match the story valence were coded as 2.

Position of scenes
A broadcast news story usually has 5 – 15 scenes depending on the length of
the story. In order to equally code scenes to three positions with an
approximate of 33% each, the total number of scenes in a story were divided
by 3.3, and then rounded up to the nearest number. The calculated number
became the number of early scenes and ending scenes for that story, and the
rest of the scenes were counted as the middle scenes. For example, if a
story has 8 scenes, 8 is divided by 3.3 which yields the value of 2.6. In
this case, 3 the nearest number rounded up from 2.6 became the number of
early and ending scenes. This means that the first, second, and third
scenes are coded as the early scenes, the fourth and fifth scenes are coded
as the middle scenes, and the sixth, seventh, and eighth scenes are coded
as the ending scenes.

Duration of scenes
Duration of scenes was counted in seconds.

Dominance of scene frame
The dominance of scene frame refers to the proportion of a particular scene
frame in a story. There are four possible frames used in this study. A
story may contain all four possible frames, while some stories may have
only one or two frames. Therefore, the dominance of scene frame was
calculated by dividing the number of scenes with a particular frame with
the total number of scenes in a story. For example, when there are eight
scenes in a story, and among the eight scenes if four scenes have the
"problem identification" frame, three scenes have the "causal
interpretation" frame, and one scene has the "moral evaluation" frame, each
of four scenes with the "problem identification" frame was coded as 50%
because they take up half of the story. The three scenes with the "causal
interpretation" frame were coded 37.5% and the one scene with the "moral
evaluation" frame was coded 12.5%. The scene coded as 50% means that the
scene has a frame that takes up 50% of dominance in the whole story.

Dominance of scene valence
The dominance of scene valence refers to the proportion of a particular
scene valence in a whole story. The dominance of scene valence was
calculated using the same methods of measuring the dominance of scene
frame. When out of four scenes, three scenes have positive valence and one
scene has negative valence, then the three positive valence scenes have 75%
of dominance and the negative scene has 25% of dominance.

Inter-coder reliability
Scene
Two coders conducted the analysis. The researcher coded the sample stories
with the scene unit, and the other coder coded with the story unit. Scene
segmentation was conducted by the researcher.

Frame and valence
10 stories with 93 scenes which compose 11.49 % of total sample were
randomly selected with SPSS, and then exchanged between the two coders to
test inter-coder reliability. Scott's pi was used in measuring reliability,
correcting for agreement due to chance. The inter-coder reliability for
scene frame was .75, story frame was .74, scene valence was .81, and story
valence was .84.


Results

The sampled 87 stories were first segmented into scenes. As the result, a
total number of 706 scenes were sampled for the study. As the Table 1
shows, there are at average about 8 scenes in a broadcast news story and
the average length of a scene is 17.56 seconds (SD= 27.03).

[Insert Table 1 about here]

Table 2 shows the percentages of the four frames and three valence for
scenes and stories. About half of the scene frame is the 'problem
definition' frame (47.7%) while the "causal interpretation' frame is the
least adopted frame in scenes (12,00%). For scene valence, there are more
negative valence scenes (33,70%) than positive valence scenes (14.70%), and
51.60% of scenes are neutral.
The story frame and valence also showed a similar pattern. In case of story
frames, the 'problem definition' frame is the most common frame with 48.9%
and the 'causal interpretation' frame is the least common with 8.10%. For
story valence, there are more negative stories (31.70%) than positive
stories (7.2%).

[Insert Table 2 about here]

The first hypothesis posits that the early scenes are more likely to have
the overall story frame than middle and ending scenes. A chi-square
analysis conducted between three scene positions, and the number of scenes
matching the story frame and not matching the story frame. As Table 3
indicates the chi-square analysis failed to produce a statistically
significant result. The first hypothesis was not supported. Scene positions
do not have relationships wit the number of scenes matching the story frame.

[Insert Table 3 about here]

The second hypothesis predicts that longer the scene is, the more likely
the scene frame is to have the overall frame of the story. A t-test was
conducted to compare the means of scene length between scenes matching the
story frame and scenes not matching the story frame. As Table 4 indicates
the t-test result was not statistically significant. The result shows that
the length of scene does not have any relationship with whether the scene
frame matches the story frame.
The third hypothesis predicted that the more dominant a scene frame is, the
more likely the scene frame is to be the overall story frame. A t-test
result indicates that scenes matching the story frame have higher dominance
of scene frame (Mean = 67.13, SD = 21.70) than the scenes not matching the
story frame (Mean = 31.89, SD = 16.00) (t=23.56, df=7-4, p< .001). The
third hypothesis is supported. The scene frames that compose a high
proportion in a story are likely to be the overall story frame.

[Insert Table 4 about here]

The dependent variable, whether the scene frame matches the story frame is
a dummy variable with two values. Therefore, a multiple logistic regression
analysis was conducted to examine the influence of all three variables –
scene position, length and dominance of frame – on the dependent variable
simultaneously.  As the results are shown in Table 5, the model chi-square
value is significant (_2= 405, df= 3, p < .001). The model explains the
variance of the dependent variable significantly. The result also shows
that scene position and scene length do not have significant relationship
with the variance of the dependent variable, but the dominance of scene
frame statistically significantly contributes to the variance of the
dependent variable (_2= 182.94, df= 1, p < .001). The odds rations
coefficient indicates that when one unit (percentage) of the dominance of
scene frame increases the probability of matching the story frame over not
matching the story frame increases 1.10 times. Figure 3 illustrates the
relationship between the dominance of scene frame and the possibilities of
a scene frame matching the story frame over the total number of stories.

[Insert Table 5 about here]
[Insert Figure 3 about here]

The hypothesis 4 through 6 addresses questions about the scene valence. The
hypothesis 5 posits that the valence of early scenes is more likely to be
the overall story valence than middle and ending scenes.  As Table 6 shows
the result of a chi-square analysis does not support the hypothesis. The
position of scene valence does not make difference in the in the number of
scene matching the story valence.

[Insert Table 6 about here]

An independent t test was conducted to test whether there is a difference
in length between the scenes matching the story valence and the scene not
matching the story valence. The t-test is not statistically significant,
thus the hypothesis 5 is not supported as shown in Table 7.
The hypothesis 6 predicts that the dominant scene valence is more likely to
be the overall story valence than less dominant scenes. A t-test result was
supported (t=23.56, df=704, p < .001) (Table 7). The scenes with frames
matching the story frame had higher dominance (Mean = 67.13, SD= 21.70)
than scenes with frames not matching the story frame (Mean = 31.89, SD =
16.00).

[Insert Table 7 about here]

A multiple logistic regression analysis was conducted with all three
variables – scene position, scene length, and dominance of scene valence to
predict the dependent variable, whether scene matches the story valence or
not. As shown in Table 8, the model chi-square was significant (_2= 434.33,
df =3, p < .001), which indicates that the model explains the variance in
the dependent variable. The result shows that scene position and length do
not predict whether the scene valence matches the story valence, but the
dominance of scene valence is a significant predictor of the number of
scene matching the story valence (_2= .10, df=1, p < .001). The odds ratio
coefficient indicates that when one unit (percentage) of the dominance of
scene valence increases the probability of matching the scene valence over
not matching the scene valence increases 1.11 times. Figure 4 illustrates
the relationship between the dominance of scene valence and predicted
probability of the number of valence matched scene over the total number of
scenes.

[Insert Table 8 about here]
[Insert Figure 4 about here]

Discussion

This study explicated the nature of scene by examining the relationship
between scene position, scene length, and dominance of scene frame and
scene valence with the overall story frame and valence. First, the study
suggests that early, middle and ending positions of scenes do not affect
the overall evaluation of the story. Frames and valence expressed in early
scenes of a story do not particularly become the overall frame and valence
of the story.
One plausible explanation is that, in broadcast news, important information
is presented both at early scenes and ending scenes. Unlike newspaper
stories that use the inverted pyramid style, broadcast news stories are not
supposed to decline in interest. They should build up to a climax
(Stephens, 1993). In this sense, scenes in all three positions, early,
middle and ending, should equally contribute to the formation of the story
frame and valence.
Early scenes are also often devoted to introduce the overall situations of
a problem and provide location sketches rather than to present important
perspectives of the story. Often stories about the suicide bombing in Iraq
and stories about the Israeli-Palestine conflict began by showing street
demonstrations and rallies, which depicts overall seriousness of the
problem to the viewers rather than provide view points and interpretations.
This result is consistent with a previous experimental study, which
examined the effects of two scene-orders on viewers' memory and attitudes
(Choi & Lee, 2004). The study suggests that the order of positive and
negative valence scenes and the order of image and issue scenes do not
affect viewer's memory and affective measures. Participants who viewed
positive information about a political candidate first and then negative
information second, and the participants who viewed negative information
first and positive information second gave similar evaluations on the
candidates. The order of scene did not affect the evaluation of the whole
story. This study also suggests that the scene position is not a very
important predictor of the story frame.
Second, the study also found that the length of scene is not related to the
overall frame and valence of the story. This suggests that frames featured
longer scenes do not particularly influence the overall story frame. One
reason for this is because a lot of important perspectives are often
discussed in short scenes rather than in long scenes. Reporters who want to
present a story in an unbiased way possible try to present a story
neutrally for most parts, but make short remarks at the beginning or at the
ends, which changes the overall frame of the story. Therefore frames
presented in longer scenes ddo not determine the whole story frame.
Third, the study reveals that the dominance of scene frame and valence,
defined as the proportions of a scene frame or valence in a whole story,
has impacts on the overall frame and valence of a story. Scenes with high
dominance of frame and valence are more likely to have the overall story
frame and valence than scenes with low dominance of frame and valence. This
means that frames and valence discussed in many scenes in a story are
likely to be the overall story frame and valence. For example, in a story
about a gay marriage, if scenes supporting the gay marriage out number the
scenes criticizing the gay marriage, the overall frame of the story
supports the gay marriage.
According to the study, dominance of a scene in a story (frequency of
scenes) determines the overall frame of the story rather than scene
position (placement) and length (size). From the media framing perspective,
the results of this study imply that when reporters produce news in
newsrooms, they should be careful about the frequency of presenting certain
viewpoints more than position and length in order to accomplish unbiased
reporting.
There are some limitations in the study. First, the study does not confirm
that stories can be broken down into scenes in broadcasting news studies.
This can only accomplished by establishing a solid coding book and reaching
high inter-coder reliability in scene segmentation. For this paper,
reliability was assessed by an agreement between this researcher and a
coder. A further study should find a methodological framework of assessing
high reliability between two independent coders.
Second, because only three top stories were sampled from news programs,
most of the topics were concerned with the 2004 Democrat primaries,
political upheavals in Haiti, and the gay marriage. More diverse topics of
news should be included in the sample in future analyses.
Third, this study doesn't tell the causal relationship between attributes
of scenes and the story frame. Some questions concerning the nature of
relationship between the scene position, length, and dominance of frame and
valence are better answered with experimental studies. More experimental
studies should be conducted to measure the impact of scene's attributes on
the overall evaluation of stories.



















Table 1            Means and standard deviation for stories and scenes

Variables
Mean
SD
N
Story
Story length in seconds
142.48
27.03
87
Scene
Scene length in seconds
17.56
8.01
706
Dominance of scene frame*
52.60
26.14
706
Dominance of scene valence**
57.27
25.35
706
Number of scene in a story***
8.11
1.76
87
* Ranges from 0 to100 percents.
** Ranges from 0 to 100 percents
*** The number of scene in a story was counted.



















Table 2        Percentages for frame and valence of stories and scenes

Variables
%
Is the scene fame the story frame?
      Yes
      No

58.80
41.20

100.00%
(N= 706)
Is the scene valence the story valence?
      Yes
      No

64.60
35.40
100.00%
(N=706)
Position of scene in a story
      Early scenes
      Middle scenes
      Ending scenes

29.60
40.80
29.60
100.00%
(N=706)
Scene frame frequency
   Problem definition
   Causal interpretation
   Moral evaluation
       Treatment recommendation

49.70
12.00
22.50
15.70
100.00%
(N= 706)
Scene valence frequency
   Positive
   Neutral
       Negative

14.70
51.60
33.70
100.00%
(N=706)
Story frame frequency
   Problem definition
   Causal interpretation
   Moral evaluation
       Treatment recommendation
       Other Frames

48.90
8.10
28.50
14.60
100.00%
(N=87)
Story valence frequency
   Positive
   Neutral
       Negative

7.20
61.00
31.70
100.00%
(N=87)









Table 3      Cross-tabulation of scene position by the number of scene
frame matching the story frame

Scene position
Whether scene frame matches the story frame
Early scene
Middle scenes
Ending scenes
Matching
62.20%
55.90%
59.30%
Not matching
37.80
44.10
40.70
100%
(N =209)
100%
(N =288)
100%
(N =209)

_2 = 2.02, df = 2, ns
Cramer's V = .05



Table 4   Independent t test for scene length and dominance of scene frames
by whether scene frame matches the story frame

Whether scene frame matches the story frame
Matching
Mean
(&SD)
Not matching Mean
(&SD)
Variable
(N =415 )
( N =291 )
t value
df
Significance
Scene length in seconds
17.85
(8.41)
17.15
(7.42)
1.14
704
ns
Dominance of scene frame*
67.13
(21.78)
31,89
(16.00)
23.56
704
p < .001
* Ranges from 0 to 100 percents










Table 5   Multiple logistic regression analysis of scene position, scene
length, and dominance of frame on whether the scene frame matches the story
frame
N = 706
Variables
B
p
Expected _
Odds ratio
Scene position
.20
ns
.09
1.22
Scene length
.03
ns
.11
1.02
Dominance of scene frame*
.09
P < .001
1.33
1.10
* Ranges from 0 to 100 percents.
_2 =405.01, df = 3, p < .001









Table 6      Cross-tabulation of scene position by the number of scene
valence matching the story valence

Scene position
Whether scene valence matches the story valence
Early scene
Middle scenes
Ending scenes
Matching
67.00%
64.20%
62.70%
Not matching
33.00
35.80
37.30
100%
(N = 209)
100%
(N =288)
100%
( N =209)

X2 = .87, df = 2, ns
Cramer's V = .04








Table 7    Independent t test for scene length and dominance of scene
valence by whether                scene valence matches the story valence

Whether scene valence matches the story valence
Matching Mean
(&SD)
Not matching
Mean
(&SD)
Variable
(N =456))
( N =250 )
t value
df
Significance
Scene length in seconds
17.59
(7.83)
17.50
(8.35)
.15
704
ns
Dominance of scene valence*
70.36
(18.74)
33.41
(17.10)
25.83
704
P < .001
* Ranges from 0 to 100 percents









Table 8   Multiple logistic regression analysis of scene position, scene
length, and dominance of valence on whether the scene valence matches the
story valence
N = 706
Variables
B
p
Expected _
Odds ratio
Scene position
.01
ns
.004
1.01
Scene length
.0004
ns
.02
1.00
Dominance of scene frame*
.10
P < .001
1.39
1.11
* Ranges from 0 to 100 percents.
_2 =434.33, df = 3, p < .001













Figure 1. A story about a political candidate

Scene 1

Candidate's stance on economic issue
Scene 2

Candidate's stance on foreign policies
Scene 3

Candidate's stance on environmental issues
Scene 4

Candidate's past voting records
Scene 5

Criticisms about the candidate








Figure 2. A story about the abortion issue

Scene 1

Pro-life
Scene 2

Pro–choice
Scene 3

Pro-life
Scene 4

Pro-choice










   [--- ???  Graphic Goes Here  ---]


Figure 3. The relationship between frame proportion and predicted
probability of the number of frame-matched scenes over the total number of
scenes


   [--- ???  Graphic Goes Here  ---]


Figure 4. The relationship between valence proportion and predicted
probability of the number of valence-matched scenes over the total number
of scenes
References


Auh, T. S. (1977). Issue conflict and mass media agenda setting during the
Bayh-Lugar senatorial campaign of 1974. Paper presented to the Association
for Education in Journalism and Mass Communication (AEJMC) annual conference.
Beinstein, A. D., & Beinstein, P. W. (1995). Quotations from speaker Newt:
The little red, white and blue book of the Republican revolution. New York:
Workman Publishing.
Choi, Y. J., & Lee, J. H. (May, 2004). Effects image-issue and
positive-negative scene orders in broadcast news stories. Paper accepted to
the annual conference of International Communication Association, New
Orleans, LA.
Coates, C. (1994). Professional's TV news handbook. Chicago, Il: Bonus
Books, Inc.
Entman, R. M. (1993). Framing: Toward clarification of a fractured
paradigm. Journal of Communication, 43, 51-58.
Ghanem, S. (1997). Filling in the tapestry: The second level of agenda
setting. In M. McCombs, D. L., Shaw, & D., Weaver (Eds.), Communication and
democracy (pp. 3-14). Mahwah, MJ: Erlbaum.
Gitlin, T. (1980). The whole world is watching: Mass media in the making
and unmaking of the new left. Berkeley, C. A.: University of California
Press.
Green, M. (1969). Television news: Anatomy and process. Belmont, CA:
Wadsworth Publishing Company.
Iedema, R. (2001). Analyzing film and television: A social semiotic account
of hospital: An unhealthy business. In T. V. Leeuwen & C. Jewitt (Eds.),
Handbook of visual analysis. London: Thousand Oaks.
Keller, T., & Hawkins, S. A. (2002). Television news: A handbook for
writing, reporting, shooting, & editing. Arizona: Holcomb Hathaway.
Kiousis, S. (2004). Explicating media salience: A factor analysis of New
York Times issue coverage during the 2000 U.S. presidential election.
Journal of Communication, 54, 71-87.
Lee, J. H., & Choi, Y. J. (July, 2002). Axis of Evil vs. Sunshine frames:
U.S. and S. Korean TV news coverage of President Bush's visit to S. Korea.
Paper presented at the annual conference of the Association for Education
in Journalism and Mass Communication, Miami, FL.
Mascelli, J. V. (1965). The five C's of cinematography. Hollywood, CA:
Cine/Graphic Publications.
Metz, C. (1974). Film language: A semiotics of the cinema. New York: Oxford
University Press.
Monaco, J. (1977). How to read a film: The art technology, language,
history and theory of film and media. New York: Oxford University Press.
Newhagen, J. E., & Reeves, B. (1992). The evening's bad news: Effects of
compelling negative television news images on memory. Journal of
Communication, 42 (2), 25 – 41.
Papper, R. A. (1995). Broadcast news writing stylebook. Needham Heights,
MA: A Simon & Schuster Company.
Stephens, M. (1993). Broadcast news. Orlando, FL: Harcourt Brace Jovanovich
College Publishing.
Tankard, J. W. Jr., Hendrickson, L., Silberman, J., Bliss, K., & Ghanem, S.
(1991). Media frames: Approaches to conceptualization and measurement.
Paper presented at the annual meeting of the Association for Education in
Journalism and Mass Communication, Boston, MA. .
Watkins, S. C. (2001). Framing protest: News media frames of the Million
Man March. Critical Studies in Media Communication, 18, 83-101.
Whitaker, R. (1970). The language of film. New Jersey: Prentice-Hall.