A systematic approach to analyzing the structure of news texts



Michael Schmierbach
U. Wisconsin-Madison graduate student
218 S. Charter St., Madison, WI, 53715
(608) 294-6528
[log in to unmask]
 Abstract: Despite increasing research examining media texts, there is no
systematic approach to news discourse. This paper considers some of the issues
raised by work in framing and discourse analysis and then suggests a systematic
approach to analyzing media texts - such an approach could provide data that
would be more useful for the development of theories about how the structure of
news discourse influences audiences, authors and the texts themselves.

Within the field of journalism and mass communications research, one of the
greatest ironies is how little actual media texts are studied. Studies have
instead tended to focus on the production of news or, more frequently, the
reception and effect of the media. To the extent that variation in media texts
has been evaluation, scholars have tended to look at broad shifts in the subject
of medium of the message.
This pattern dates back to some of the earliest quantitative studies in the
field. The seminal work by Lazarsfeld et al., for example, asked respondents
which candidate's messages they had heard (subject) and by which medium (1944).
Subsequent research has tended to follow this model, developing careful and
systematic measures of "exposure" and "attention" to texts but paying relatively
little attention to variation in the texts themselves.
Fortunately, not all research in the field has neglected variation in texts. At
its core, agenda setting asks how differences in the subjects covered in the
media affects the degree of importance attributed to those subjects by audiences
(McCombs & Shaw, 1997). Somewhat related to this investigation is work in media
framing, which has looked not only at the topics contained in the media but the
way those topics are presented.
Nevertheless, systematic work that evaluates media texts has remained uncommon.
In the following paper, I will address some of the existing work and the
research possibilities it suggests. Then I will propose a more systematic way of
evaluating not only the content, but also the structural form of news texts.
This proposal relies upon evaluating texts in terms of six aspects. In
suggesting these aspects, as well as proposing specific measures by which to
evaluate them, this paper aims at allowing researchers to gather more
theoretically useful data. Even though researchers have begun to look at the
details of texts, they have rarely provided measures that would allow text
structures to be compared across topics. Any complete theory of the interaction
of text structure, language, production and reception cannot be formed until a
thorough and systematic means of evaluating all these areas, including "frames"
or broad structures, is adopted.
Literature review
In order to propose a universal approach to evaluating texts, it's important to
first understand existing work dealing with text structure and language. Many
existing studies raise important issues about the nature of texts and the ways
they interact with audiences as well as the ways in which they are created. A
comprehensive approach to news texts should be able to account for the
theoretical possibilities raised by much of this working, providing data that
could potentially confirm or deny some of these theories.
Although a large number of media theories deal at some level with texts, there
are two areas that have largely dealt with the shape of texts as a primary
focus. Previously mentioned work in framing is probably the most common
currently, and it is certainly tied to many prominent theoretical ideas within
the field. Any theory that takes into account processing behaviors by readers
has to take into account the nature of the text being processed. Another area of
research has provided deeper insights into the complex shapes of texts
themselves, however. This research is less clearly defined, but it is often
referred to as "discourse analysis." Understanding the attempts to describe
media discourse within both of these fields should indicate much of what my
systematic approach tries to address.
Discourse analysis is intimately tied to linguistics and extends far beyond the
field of media research. In fact, basic texts on such analysis tend to avoid
"applied" areas such as journalism, instead preferring to evaluate relatively
simple, interpersonal texts (such as conversations, phone messages and e-mail)
(Georgakopoulou and Goutsos, 1997). Nevertheless, whether used to evaluate texts
created around the dinner table or in The New York Times, discourse analysis is
largely an attempt to find meaningful patterns within the text.
When applied to media texts, discourse analysis has typically looked for
categories of linguistic practices that are used by reader and author alike to
organize and understand the presented information. Van Dijk, in proposing a
comprehensive analysis of news texts, suggests a range of categories into which
the various parts of a news story fall (1988). Furthermore, he argues that there
is a relationship between these categories and the cognitive processes by which
authors and readers structure information - thematically central information
within the text is also central to the author's understanding of the story, for
example.
Unfortunately, there is little agreement on what these categories are. Whereas
van Dijk creates tree diagrams featuring such categories as "episode,"
"background" and "context" (1985), Gerbner provides a schema that evaluates
elements of "existence, importance, values and relationships" (1985, p. 20). In
some ways, Gerbner's efforts to create a systematic method of content analysis
are more comprehensive than van Dijk's examination of specific structural
elements, and Gerbner's schema specifically includes measures of structure.
Nevertheless, many applied efforts within discourse analysis have tended to move
away from such macro-textual concerns into micro-textual evaluations of
syntactical and lexical choices on the part of text authors. For example,
Fowler, while providing a fairly comprehensive listing of typical linguistic
categories as they exist within news texts, clearly focuses on questions of word
choice, sentence arrangement and (interestingly) typography (1991).
Although such micro-textual analyses are potentially informative, they stray
considerably from the general structural issues laid out by early theorists in
media discourse. Furthermore, they are typically not presented in such a way as
to allow research into variation in language use within the media. This is not
to say such a process couldn't take place. As work being done within
sociolinguistics demonstrates, it is entirely possible to correlate variation in
all sorts of linguistic categories with variation in speakers (Chambers, 1995).
There is no reason many of the categories presented by Fowler, particularly
vocabulary, couldn't be correlated with both production and consumption aspects
of the media; that connection simply hasn't been drawn in any comprehensive and
systematic fashion.
Regardless, the important observation that comes from discourse analysis is that
the language and structure of media texts act to constrain the shape of those
texts (Bell, 1991). What observations by van Dijk and others begin to suggest is
that the content and structure of media discourse is constrained not only by the
news gathering environment but also by the "rules" of news discourse itself.
Although an understanding of social and economic constraints on the production
and processing of news texts remains important, a structural analysis of media
discourse should also provide measures of patterns intrinsic to the discourse
itself, thereby allowing us to map and understand the role those patterns play
in the mass communication process.
Unfortunately, most recent analysis of media texts has moved away from an
understanding of language as a self-structuring phenomenon. On the other hand,
recent analysis of "frames" within media texts has gone a long way toward
providing measures of language use that can be meaningfully correlated to
attitudes held by text recipients. Alas, the various definitions of "framing"
make it difficult to compare results across studies.
Existing summaries of framing research have tended to lament the "fractured"
nature of the field (Entman, 1993) or divide the field into various typologies
(Pan and Kosicki, 1993). Although these overviews give some sense of the variety
of work being done in the field, they do little to explain how disparate
definitions of "frames" can be consolidated into a meaningful theory of framing.
The recent explosion in work on framing has only complicated this field.
In general, recent studies tend to define frames in one of two ways. Either they
are seen as general organizing structures for a news story (or series of
stories) or they are seen as the specific "spins" actors within a story put on
their version of events. Gamson and Modigliani clearly view frames as structural
elements when they discuss the discourse of nuclear power (1989), for example.
Sadly, this account, like similar structural approaches, does little to provide
a method of analysis across content areas - the organizing frames provided by
the authors are often specific to the topic being explored. Shah et al. make an
admirable attempt to provide for more generalizable frames in their experiments
demonstrating a relationship between "ethical" and "material" frames within news
stories and the decision-making processes of readers (1996). Still, the bulk of
work on structural frames tends to be topic-specific and unrelated to linguistic
practices.
The difficulties are only compounded by the existence of a second concept of
frames - a specific "slant" on a story contained within the larger structure of
a news text. When Jasperson et al. evaluate the presence of frames on stories
dealing with the budget deficit, their computerized analysis of word proximity
would seem to generate a count of individual frames within a larger story
(1998). More so than the structural analysis of frames, this approach would seem
to capture what the authors describe as the second level of agenda setting - a
connection between the frequency of certain interpretations of an issue within
the media and the acceptance or understanding of that interpretation within the
media consumer.
Resolving the two notions of frames is further complicated by confusion over the
meaning of framing itself. In addition to existing at different levels within a
story, framing can mean different things. Benford argues that framing conveys
both the sense of structural relationships between components of a text as well
as the notion of a method by which components are included and excluded (1997).
Accordingly, it is difficult to summarize what is meant by framing, and the
approach presented in this paper rejects that term in favor of more specific
terms for a variety of textual aspects included under the broad category of
framing.
It is possible to touch on what has been captured and what has been ignored in
the literature on framing. The relationship between the structural aspects of
language and the shape of news stories, so vital in discourse analysis, has
largely eluded the work being done in framing. Furthermore, despite early
ethnographic work that used the metaphor of framing to discuss media production
(Tuchman, 1978), most recent work has tended to correlate frames only to media
effects (particularly decision making processes and perceived issue salience) if
any connections have been drawn at all. The increased interest in framing has,
however, opened up analysis of media effects to textual aspects beyond content
area and medium, with laudable results. As Iyengar has repeatedly demonstrated,
the relationship between media content and various cognitive processes on the
part of message recipients is real and complex, in part due to the way the
structure of texts selects and highlights certain messages and content areas
(1991).
Structural analysis of news texts
The compelling results obtained from a variety of approaches to news discourse
demonstrate both the existence and viability of a range of approaches.
Nevertheless, as the preceding review of existing literature demonstrates, there
are problems with some of the existing approaches. A new, systematic approach to
analyzing texts must deal with these problems. Such an approach should allow for
comparison across content areas, reducing the importance of topic-specific
categories of analysis - for example, Menashe and Siegel's analysis of various
frames in media coverage of tobacco (1998) may be useful in understanding that
specific issue, but categories such as "non-smoker's rights" and "free
speech/legal product" are of little use to researchers interested in other areas
of the media or even other aspects of health policy. Although content-specific
categories cannot be eliminated, their importance can be reduced by including
other, standardized aspects of texts in the analysis.
Those other aspects should address some of the issues brought up by discourse
analysis as well as other research into the production and use of media texts.
Analysis should not only consider the relationship between content variation and
effects but also between structural and content variation and between variation
in producers and in the texts they produce. Research along those lines would
provide more universally applicable (and therefore theoretically useful) data as
well as potential insights into the way the structure of discourse helps shape
media texts and their impacts.
With these considerations in mind, I propose that media texts be analyzed in
terms of six important aspects: What claims are being made in a text?; Who is
making those claims?; How do the claims relate to the structure of the news
story?; Where are the claims located within the news story?; Where are the
claims located relative to other claims?; and What language is being used to
present the claims?
In the remainder of this article, I will consider the theoretical rationale
behind each of these components as well as the specific measures that might be
used to evaluate them. Furthermore, I will suggest ways in which each component
relates to the others and to existing theory. Finally, I will draw upon examples
from a specific news text (see appendix) to demonstrate the analytical process.
What claims are being made?
Claims are closely related to the second notion of frames discussed in the
literature review - specific approaches to a subject provided by actors and
contained within a larger story. Categorizing claims requires an understanding
of the issue being discussed within the story or stories under analysis - the
range of possible claims is the range of viewpoints being presented within the
issue. As much as possible, these viewpoints should be divided into general
categories representing sets of related propositions about the issue.
In many ways this is both the most and least problematic of the six categories.
Despite my earlier caveat, the classification of frames still requires
issue-specific analysis. In some ways, this would appear to thwart efforts to
create an analytical method that would allow comparison of variation across
content areas. On the other hand, because it relies upon issue-specific
categories, the notion of claims is closely tied to most existing framing
research and therefore easily applicable to work in that field.
To some extent, issue-specific analysis is both inevitable and desirable. There
is no question that the range of possible approaches to a topic, as well as the
approaches within that range that are selected for inclusion by the media, are
important to understanding the subject being studied. Textual analysis is used
to gain an understanding of a political or social subject as much as it is
intended to gain an understanding of the media process; furthermore, different
topic areas may be approached and understood in drastically different ways by
both message producers and consumers. Rejecting any topic-specific categories
would prevent any sort of topic-specific understanding.
Nevertheless, there are two ways in which the analysis of claims within this
schema can be used to gain a more general understanding of media discourse.
First, patterns may begin to emerge in the sorts of claims that are used in
texts. If we define and evaluate claims consistently, research is more likely to
reveal consistent patterns in the types of claims articles contain - such
patterns might be related to the "framing" types (ethical, material) proposed in
the Shah et al. article described earlier (1996). Second, it is important to
remember that claims are not the only thing being analyzed in this approach.
They are instead evaluated in conjunction with other aspects of media texts,
making it easier to see patterns in both claim types and media discourse as a
whole.
Because claims are closely tied to existing research approaches, the theoretical
justification for analyzing them should be fairly clear. The concept of claims
is tied to a vast body of research concerning both inputs and outputs of media
texts. On the input side, a systematic measurement of claims would provide one
test for a variety of theories concerning media constraints. If, as numerous
scholars have suggested, certain messages (namely those tied to the dominant
ideology of society) are more likely to find their way into the media, this
should manifest itself in quantifiable ways within texts (examples of constraint
theories include Bagdikian, 1992; Gandy, 1982; Herman & Chomsky, 1988). On the
output side, claims are clearly linked to agenda setting and "typical" framing
theories. Media texts do appear to determine (in part) what issues are ranked as
salient in decision making by readers (Iyengar & Simon, 1993). If we count which
claims appear more or less frequently within texts, we will have some idea of
what issues are being pushed to the top of the agenda for the topic in question.
Obviously, if the types of claims to be coded for vary according to topic, it is
impossible to specify them in this context. General guidelines to claims can be
presented, however, and these guidelines should allow comparison in text
structure between different topics. Claims should fall into discreet categories,
each of which represents a philosophically coherent understanding of a topic.
For example, in articles on the economic cost of logging, one category might be
claims that logging's costs are transparent and accurately reflected in U.S.
Forest Service accounting (see appendix). Another category of claims might be
that costs are not accurately reflected by such accounting due to dishonest or
inaccurate practices on the part of the Forest Service. And yet another category
might be that economic principles are incapable of reflecting the price of
environmental damage and that debates over the monetary cost of logging are
therefore irrelevant or even harmful.
Not all claims are easily categorized, of course. Analysts should remain open to
the possibility that they have not accounted for all possible categories of
claims. They should also work to make very explicit the boundaries of
categories. Furthermore, it may help to remember that each paragraph in a news
story, due to the conventions of journalistic writing, will usually represent a
single claim. Finally, nearly all claims will include some kind of warrant, a
piece of evidence (however weak) backing the advocacy of the claim.
There are exceptions to these typical advocacy claims, and they come in two
kinds. First, for every claim there is the possibility of a direct refutation of
that claim, an "anti-claim" if you will. For example, the Forest Service denial
of an environmentalist claim (paragraph seven) represents such an anti-claim. It
doesn't not present any aspect of a philosophically coherent position; instead,
it simply refutes a coherent claim. There is a subtle distinction between a
claim that refutes another claim and a merely contradictory claim. Paragraph
three and paragraph nine represent contradictory claims. The third paragraph
falls into the category of claims that Forest Service costs are open and
accurate - this is the "first time" a loss has existed, expenditures are
consistently "reported" and the loss is only "$14.7 million." The ninth
paragraph, on the other hand, represents the claim that accounting methods are
inaccurate and even dishonest. The Forest Service relies up "tricks to hide the
loses," in part by "not count[ing]" certain expenditures. Each position could
and does exist on its own - individuals do not need to hold one belief for the
other to be tenable. On the other hand, the claim in paragraph seven would be
meaningless unless it refuted the claim in paragraphs five and six. The
importance of such anti-claims is unclear, but accounting for it in analyses
could help determine if readers and authors treat them differently than coherent
claims. For example, is the position of a group weakened in the perception of
text consumers if that group's claims are primary in refutation?
The other type of claim is a "neutral" claim that does explicitly advocate a
position. This is not to say that such claims do not have philosophical
importance or that the warrants contained in such claims do not affect audience
opinions. Rather, it is simply that as worded, the claims do not take a position
on the issue at hand. The tenth paragraph is a good example of this - as
positioned in the article, there is no reason the amount of timber sold by the
government would affect the validity of accounting methods or the cost of
selling timber. Obviously, this information could be used in an advocacy claim;
an environmental group might suggest the amount of timber sold demonstrates the
reluctance on the part of the Forest Service to development non-timber-based
policies. But in the context of the story, this is a neutral claim.
Again, there is no firm rule as to how many categories of claims there are, how
narrowly or broadly to set the boundaries of those categories or how to resolve
differences between coders. These are things that, to some extent, will vary
depending on the subject and the goal of the research. Overall, though, if
future analysis of texts evaluates claims as defined in this proposal and
connects these claims to some of the other important aspects, the data provided
should allow development of theories beyond those facilitated in current
research.
Who is making the claims?
Although probably the most researched aspect of next texts, claims are also the
most complex. The remaining five aspects are simpler to categorize and explain,
although they are no less important. Furthermore, all five aspects are coded
according to universal standards that should not shift from topic to topic.
Because these are universal categories, analysis of variation in these
categories across topic areas, authors and effects should prove extremely
fruitful.
The first such set of universal categories applies to the people making claims,
the actors in a news text. Although actors have received little attention in
evaluations of media effects, they are probably the most important focus of
studies that look into the production of media texts. For example, Tuchman
argues that journalists rely upon a web of known sources (1978), while Herman
and Chomsky suggest that elite sources rely upon "flak" to punish those
journalists who turn to sources and claims that differ from mainstream
ideologies (1988). Similarly, Gandy claims that sources subsidize certain kinds
of information by making it less expensive for journalists, both in terms of
money and time spent retrieving it (1982). As was the case with theories of
constraints concerning claims, constraints on the actors available to
journalists can, in part, be tested by a systematic analysis of which actors
actually appear in news stories, especially if the number and range of actors
within mainstream texts is compared to other catalogs of available sources, such
as alternative media. Additionally, despite a relative lack of research on the
subject, there is no reason to believe that actors have no influence on the ways
claims are received. A systematic study of the types of actors who appear in
certain stories would allow for correlation of variation in actors not only with
production of texts but also with variation in the ways texts are received. For
example, are texts with fewer privileged actors (i.e. "elite" sources) less
likely to be perceived by readers as truthful or useful?
Actors are simply defined as the source given for a claim. These may be people,
groups or even documents. If no source is given, such as the claim about the
amount of timber sold in paragraph ten of the sample story, or if the source is
the journalist themselves (some form of personal observation) then the source is
the text author. Otherwise, authors fall into one of several categories:
Government - any officials associated with a public body; business - anyone
explicitly related to the management and operation of a business (such as Frank
Stewart in paragraph eleven, because of the mention of industry); activist - any
group that works (outside of business or government) to achieve certain
political and social goals (this would include unions, as well as the
environmental groups mentioned in the article); media - other media sources (for
example, the Washington Post article cited in the story); unaffiliated experts -
for example, members of the academy, who are seen not as associated with a group
involved in the issue but are seen as possessing unique and privileged knowledge
about the situation. In some rare cases it may be important to break down the
activist category; for example, in discussions of same-sex marriage, the
distinction between religious activists and gay-rights activists may be
important. However, because both could be aggregated into the activist category,
comparisons can still be made between studies, since the categories are
sufficiently broad to allow for meaningful analysis regardless of the topic. Two
final categories are also important: affected and unaffected individuals. These
groups are distinguished by the fact that they are associated only with a broad
public (in the case of unaffected individuals) or with a narrower social group
somehow affected by the debate (for example, a logger in the case of debates
over timber). Typically, these individuals are not seen as having any special
insight into the situation, even if they are directly and immediately affected
by it. Many theorists suggest that such individuals are far less likely to
receive positive coverage (or any coverage) from the media despite the useful
knowledge they might have, due in part to the constraints on news gathering
(Eliasoph, 1998). Again, by using the proposed categories, this proposition
could be tested by a systematic analysis.
How do the claims fit into the structure of the text?
Although this is probably the most vital aspect of structural analysis, it is
also the least explored within the existing literature. Despite numerous framing
studies that claimed to deal with the way news discourse was structured, the
reality is that such discussions tended to focus more on the acceptable range of
claims within a text than on the actual linguistic techniques used to arrange
news stories. This aspect does not deal with the "real" relationship between
different claims but rather with the relationship implied by the structuring
language used in the text. This is an important distinction. A successful
analysis of this aspect (and the narrative structure of a news text) should not
simply group the claims contained within a text into a category and label the
story's structure accordingly - this would provide a meaningless measure, since
the types of claims are already represented by analysis of the claims
themselves.
Instead, narrative structure is closely related to the concept of ideology. Like
framing, this is a term that has various definitions and is accordingly
problematic. Following in part the model proposed by van Dijk (1998), I argue
that ideology is neither a set of individual beliefs nor a hegemonic set of
beliefs enforced upon an entire society. Instead, an ideology is a widely held
"map" of acceptable beliefs and the relationships between them. This notion of
ideology is conceptually important because it illustrates the possibility that
some beliefs are more likely to be held within a given ideology and that some
ideas are related to other ideas within a given ideology (regardless of the
logic behind that relationship - consider the common link between free-market
capitalism and social conservatism). In this sense, ideology is intimately
related to the structural practices of language (and rationally so).
Finding empirical evidence to support the concept of ideology at a societal
level is both difficult and fairly meaningless; ideology is more useful as a
model of the way beliefs are organized than an actual set of ideas that could be
somehow mapped and documented. Within news texts, however, the ideological
structure can be measured. If some degree of "ideology" exists in news
discourse, there ought to be patterns in the syntactic and structural
arrangement of that discourse. In short, claims ought to be linguistically tied
not simply because of their "rational" relationship but because of some
discursive, potentially ideological structure that helps shape news texts.
Measuring the role each claim plays in the structure is simply a matter of
determining how the claim is related to other claims within the text. Typically,
the general arrangement of claims is set in the lead paragraph of a news story;
reporters distinguish between, for example, one-sided announcements and
two-sided debates. Identifying the relationship of each claim to that structure
is simply a matter of evaluating the language used to introduce and relate the
claim to other claims. Although this could be done in a number of ways, this
paper proposes a standardized, quantifiable measure of structure to facilitate
using this analysis in conjunction with other measures. Under this standardized
approach, stories are broken into a number of sides, where each side represents
a series of linguistically linked claims.
All stories include at least one side - a neutral, informational side.
Typically, this side contains the neutral claims I previously described,
although depending on the story nearly any claim could fall into this side. The
neutral side contains all claims that are not positioned as being part of a
debate or argument - regardless of whether the claims actually constitute
advocacy or not. In the sample story, the tenth paragraph describing total
government timber sales falls into the neutral side - there is no language
linking it to any of the other sides in the story or to any sort of unique
advocacy. In some stories, for example most obituaries, there is only one side,
even though some of that information could constitute an advocacy claim. If a
dead woman is described as a "skilled artist" there is likely nothing in the
language of the obituary that would treat that as anything other than a fact.
However, if a living artist is described in an article on a gallery opening as
skilled, the language of the story will likely structure this as representing an
opinion - a non-neutral side, in other words.
A story can contain anywhere from zero to hundreds of distinct non-neutral
sides. In most cases, they contain only one or two. In the example of the
gallery opening story, there might be several claims about the artist's skill
that clearly constitute a side, but there might be no opposing claims that are
presented as a separate side. So long as the language introducing a claim
suggests that it supports an existing side, the new claim is also a part of that
side (even in cases where the language of the claim itself may not seem to
support that side). Such language could include phrases such as "X agrees with
that assessment," "one indication of this is _," "Y concurs, saying," or it
could include a clear linkage between actors (for example, if Y "also of group
X" was quoted, we could assume they represented the same side). Or, the
continuity of a statement could indicate that two claims were linked; for
example, the fourth paragraph in the sample story refers to "this pattern" from
the third paragraph and contains the pronoun "he". Because the author uses
language that requires reading information from the third graph into the fourth,
in the absence of contradictory language, we can safely place both of these onto
the same side.
Not all language is supporting, of course. The fifth paragraph of the sample
story is plainly opposed to the fourth, using language such as "charged"
(indicating it is not part of the neutral side) and "actually," suggesting a
contradiction between the two claims. In this case, we have another side coming
into existence. In many, but not all, cases, two claims that are both opposed to
the same side will themselves both be part of the same side. However, this is
not necessarily the case - it is important to examine the language carefully.
Ideally, linked claims will use supporting language (for example, the "long" in
the ninth paragraph, which implies a continuity with other claims of a similar
nature cited in the story and thus with the side those claims were a part of).
Still, oppositional language is an important indication of claims; examples of
such language might include statements such as "Z opposes," "B group has long
disagreed with this policy" or "many scientists are skeptical."
It is important to remember that the key is the language connecting claims, not
the reasoning of the claims themselves. For example, the last paragraph of the
sample story clearly places the blame for increasing timber costs on
bureaucratic practices by the government - as a claim, it falls into the same
category as the fourth paragraph of the story. Structurally, however, there is
no language linking it to anything else in the story; I would code this as a
distinct side within the story. In another story on the same subject, the last
paragraph used referential language, explaining an actor was "not commenting
specifically on the contents of the memo." There, language links the claim to
earlier claims by other Forest Service officials about the memo; here, while
philosophically similar, the claim is structurally free from earlier claims.
Conversely, the third and fourth paragraphs, described above as part of the same
side, make different categories of claims; one is about the openness and
accuracy of the accounting methods, while the other is about the source of
financial loss. Yet structurally these are linked as part of the same side.
In my view, these structural measures are the most important of the six
attributes. Largely, this is because they have been poorly examined in the
existing research. Unfortunately, the shortage of research into textual
structure also makes it difficult to speculate on how variation in structure
might affect readers or be affected by writers. If structure does in fact
represent the ideology contained within the story, however, it could be a
fruitful area of study. Potentially, the way audiences relate various opinions
could be affected by the structure of news stories; the resources available for
oppositional readings could be provided and also restricted by the shape of the
media discourse. Additionally, the process by which reporters select claims
could be shaped by the self-perpetuating ideology of the story - as the story
develops, it may be the claims that adopt themselves to the structure rather
than the structure shifting to account for the range of claims. All of this is
mere speculation, of course, but it suggests why analysis of news structure,
particularly in conjunction with the other attributes described in this paper,
could prove valuable.
Where does the claim appear in the article?
Compared to earlier measures, this one is remarkably simple. What paragraph does
the claim appear in? Van Dijk argues that claims which are more central to the
theme of the story tend to appear closer to the beginning of the story (1988).
Furthermore, he demonstrates that people are more likely to recall information
that was part of that central theme and located close to the top of the story.
If the prevalence of claim types within a set of stories is being used as a
predictor of audience beliefs, accounting for the position of that claim within
the story could greatly increase the accuracy of such predictions.
Where does the claim appear relative to other claims?
This attribute is closely related to both the general structure of the story and
the way in which claims are sorted (top to bottom) in the story. Independently,
this is simply a matter of determining whether a claim appears near other,
similar claims and whether actors appear near other, similar actors. Measuring
this would allow some understanding of whether messages gain or lose persuasive
power when they appear near supporting and opposing messages.
By tying this to the preceding two attributes, we can begin to thoroughly map
the overall structure of a news story, independent of the information contained
within. Not only could a researcher test whether certain topics were more likely
to elicit certain structures and whether structures emerged as dominant over
time (as suggested by Jamieson and Capella in their examination of media
coverage of the health care debate (1998)) but also how the structure of a news
story and type of a claim and actor appear to influence the placement of claims
within a text. Are certain actors or certain claims more likely to appear by
themselves, near the bottom of a story? Are those claims and actors typically
left unconnected to the dominant side(s) in a news story? Are these patterns
related to the type of issue being reported? To the experience of the reporter?
Do they appear to influence audience perceptions of the issue? By providing
standardized measures of the structure of news stories as well as the components
(claims and actors) that fall into that structure, this methodology proposal
hopes to make is easier for researchers to produce data that could help answer
some of these questions.
What sort of language is used to present the claim?
This last attribute is a potentially massive category, and it receives little
attention here because this article is primarily intended to encourage a more
standardized approach to textual analysis by media researchers, rather than to
encourage linguists to conduct more research on media texts. An enormous amount
of information about next texts could be obtained if only a small number of
researchers conducted studies that included the measures already listed. Adding
standardized examinations of the specific language used in the media would only
enhance the richness of that information.
Furthermore, issues of language use are included here with a note of caution.
Too often, evaluation of language has been connected to a highly non-standard
approach. While these studies have often been insightful, they provide data that
is almost useless to researchers hoping to make generalizations about patterns
in news texts. For example, Fowler's book on "Language in the News" raises a
vast number of linguistic fields that could provide valuable insight into the
shape of news texts (1991). Unfortunately, the work done by Fowler is somewhat
inconsistent; syntactic and lexical choices are highlighted (or ignored) as they
fit into the argument being made. This doesn't mean the argument is wrong, but
it does mean there is little data to relate to other observed patterns in texts,
audiences or reporters.
The stigma of linguistic investigation as inconsistent is unfortunate. So too is
the (less frequent) stigma of quantitative linguistic investigation as
pointlessly tedious and obsessed with minute details. A systematic investigation
of vocabulary in news stories, for example, may not indicate very much by
itself. But in conjunction with other measures (such as those presented
elsewhere in this paper) it could provide useful insight into how texts are
constructed and understood; it might be, for example, that variations in
vocabulary help explain why certain claims are more likely to appear in news
stories. The question of what gives a claim cognitive power (an issue
insightfully raised by Schudson (1989)) is largely unanswered; addressing that
issue will require some understanding of the differences in language use between
claims.
Clearly, the various fields within applied linguistics provide for a vast number
of potential measures. Rather than attempting to proscribe what measures might
be most useful or describing the range of available fields, this paper will
merely suggest a couple possible avenues of research that could particularly be
informative in conjunction with the measures already provided.
Obviously, questions of vocabulary are both easily answered and infinitely
valuable, if combined with other measures and treated as a variable. It is the
treatment of language use as variable that is key. Merely stating that
minorities are treated unfavorably in the press is inadequate. Rather,
researchers should be able to chart differences in the type of vocabulary used
to describe different ethnic groups and then compare those shifts in vocabulary
to shifts in the likelihood members of the group will be actors in news stories,
to the types of claims about those groups included in news stories, and to the
way news stories structure and position those claims and actors.
Similarly, syntactic observations should not merely note patterns; they should
demonstrate variance in the syntax used in news stories. If agents are typically
left out of stories about state-sponsored violence, that should be compared to
stories about other types of violence. Furthermore, it should be linked to other
measures of news texts - for example, does the lack of agency imply that state
actors are also less likely to appear in such stories? Finally, in any analysis
of language variance, it should be possible to compare such variance to
differences in the people producing and consuming the news. For example, are
minority reporters less likely to use "derogatory" vocabulary in describing
members of their minority? Is that the only difference, or do they also
structure their stories differently? Do minorities use minority actors more
frequently? And do any of these differences affect audience opinions about
minorities? Or state-sponsored violence? Or any other subject under
investigation?
Conclusions
Many of the measures suggested in the proposed methodology have been used
before, and many of the questions raised have been investigated. Existing work
on framing, in particular, is designed in part to answer questions about the
impact certain structures have on the audience. The problem, as mentioned in the
literature review, is that these measures have been inconsistent and have
usually only looked at one or two of the six attributes described above.
Each of the six attributes has potential theoretical implication when considered
as a variable unto itself. Furthermore, each of them can and should be compared
not only to production and consumption but also (and especially) variation
within the text. This possibility - comparing various components of the
structure of texts in a meaningful fashion - is the first advantage of a
standardized examination of textual structure in the manner proposed. The second
advantage is the possibility of comparing data across studies, thereby granting
us a better understanding not only of how texts interact with their creators and
readers but also of how the discursive rules of news affect the texts
themselves.
A single study of most or all of the six attributes, in the manner suggested,
even without any outside measures, would provide some insight into the
particular topic under investigation. More important, however, is the fact that
such a study, when combined with others and when designed to include other
measures, would allow us to build a more comprehensive theory about the
relationship between discursive structure and the entire media process, a
relationship that in turn has significant implications for the shape of society
as a whole.

 Appendix

November 21, 1997
By SCOTT SONNER, Associated Press Writer

(1) The Forest Service is acknowledging for the first time that taxpayers are
losing money logging national forests.

(2) A draft of a report due out next month says the government spent nearly $ 15
million more on the logging operations than private timber companies paid to
purchase the wood in fiscal 1996.

(3) "For the first time since we have reported such information, expenditures
for the program as a whole exceeded revenues ... by some $ 14.7 million," Robert
Joslin, deputy chief of the Forest Service, said in a copy of the draft obtained
Friday by The Associated Press.

(4) "This pattern can be expected to continue in the future as we place more and
more emphasis on using timber sales as a management tool for achieving
objectives other than fiber production," he said.

(5) Environmentalists charged that the agency actually reached the conclusion
months ago but kept it secret while Congress debated and narrowly defeated
proposals this fall to slow construction of logging roads.

(6) "They've known this since March," said Michael Francis, a former
congressional aide now at The Wilderness Society. "They have been sitting on the
numbers for their own political reasons."

(7) Forest Service officials denied that.

(8) The Washington Post first reported Friday on a Forest Service memo outlining
the losses.

(9) Environmentalists long have accused the Forest Service of using accounting
tricks to hide the losses. For example, the Forest Service does not count as a
cost the one-fourth of timber sale receipts that are returned to rural counties
housing the forests. That mounted to $ 240 million in 1996.

(10) The government sold 3.7 billion board feet of timber from national forests
last year.

(11) "The administration has taken a profitable program and made it unprofitable
for the taxpayers," said Frank Stewart, spokesman for the industry's American
Forest & Paper Association. He blamed it on "all the red tape this program has
to go through."
 References

Bagdikian, B. (1992). The Media Monopoly. Boston: Beacon Press.

Bell, A. (1991). The Language of the News Media. Oxford: Blackwell.

Benford, R. (1997). An insider's critique of the social movement framing
perspective. Sociological Inquiry, 67, 4, 409-430.

Chambers, J. K. (1995). Sociolinguistic Theory: Linguistic variation and its
social significance. Oxford: Blackwell.

Eliasoph, N. (1998). Avoiding Politics. New York: Cambridge University Press.

Entman, R. (1993). Framing: Towards clarification of a fractured paradigm.
Journal of Communication, 43, 51-58.

Fowler, R. (1991). Language in the News: Discourse and Ideology in the Press.
London: Routledge.

Lazarsfeld, P. F., Berelson, B. & Gaudet, H. (1944). The People's Choice: How
the Voter Makes Up His Mind In a Presidential Campaign. New York: Columbia
University Press.

Gamson, W. A. & Modigliani, A. (1989). Media discourse and public opinion on
nuclear power: A constructionalist approach. American Journal of Sociology, 95,
1-37.

Gandy, O. (1982). Beyond Agenda Setting: Information Subsidies and Public
Policy. Norwood: Ablex Publshing.

Georgakopoulou, A. & Goutsos, D. (1997). Discourse Analysis: An Introduction.
Edinburg: Edinburg U. Press.

Gerbner, G. (1985) Mass media discourse: Message system analysis as a component
of cultural indicators. In van Dijk, T. (ed.) Discourse and Communication.
Berlin: Walter de Gruyter.

Herman, E. & Chomsky, N. (1988). Manufacturing Consent. New York: Pantheon
Books.

Iyengar, S. (1991). Is Anyone Responsible: How Television Frames Political
Issues. Chicago: University of Chicago Press.

Iyengar, S. and Simon, A. (1993). News coverage of the Gulf Crisis and public
opinion: A study of agenda-setting, priming and framing. Communication Research,
20, 365-383.

Jamieson, K. & Capella, J. (1998). The role of the press in the health care
reform debate 1993-1994. In D. Graber, D. McQuail & P. Norris (Eds.), The
Politics of News, The News of Politics (pp. 110-131). Washington: Congressional
Quarterly Press.

Jasperson, A., Shah, D., Watts, M., Faber, R. & Fan, D. (1998). Framing and the
public agenda: Media effects on the importance of the federal budget deficit.
Political Communication, 15, 205-224.

McCombs, M. & Shaw, D. (1997). The evolution of agenda-setting research:
Twenty-five years in the marketplace of ideas. Journal of Communication, 47,
58-67.

Menashe, C. & Siegel, M. (1998). The power of a frame: An analysis of newspaper
coverage of tobacco. Journal of Health Communication, 3, 4, 307-336.

Pan, Z. & Kosicki, G. M. (1993). Framing analysis: An approach to news
discourse. Political Communication, 10, 55-75.

Schudson, M. (1989). How culture works: Perspectives from media studies on the
efficacy of symbols. Theory and Society, 18, 153-180.

Shah, D., Domke, D. & Wackman, D. (1996). "To thine own self be true": Values,
framing and voter decision-making strategies. Communication Research, 23, 5,
509-561.

Tuchman, G. (1978). Making News: A Study in the Construction of Reality. New
York: Free Press.

Van Dijk, T. (1985). Discourse and Communication. Berlin: Walter de Gruyter.

Van Dijk, T. (1988). News As Discourse. Hillsdale: Lawrence Erlbaum Associates.

Van Dijk, T. (1998). Ideology: A Multidisciplinary Approach. London: Sage
Publications.