AEJMC Archives

AEJMC Archives


Next Message | Previous Message
Next in Topic | Previous in Topic
Next by Same Author | Previous by Same Author
Chronologically | Most Recent First
Proportional Font | Monospaced Font


Join or Leave AEJMC
Reply | Post New Message
Search Archives

Subject: AEJ 95 BleskeG CTM Effect of headlines
From: Elliott Parker <[log in to unmask]>
Reply-To:AEJMC Conference Papers <[log in to unmask]>
Date:Fri, 16 Feb 1996 21:05:46 EST

text/plain (1964 lines)

Glen L. Bleske
California State University, Chico
Department of Journalism, School of Communication
Chico, CA 95928-0600
(916) 898-4770
[log in to unmask]
Manuscript submitted for presentation to the annual meeting
 of the Association for Education in Journalism and Mass Communication,
Communication Theory & Methodology Division,
April 1995
Newspaper headlines are conceptualized as being pre-texts that help readers
 process a news story.  Theory suggests that headlines work by cueing
       readers who use the information to organize material in the news story
          integrate the material into their long-term memories. Results from a
      controlled experiment using student subjects supported the main
 headlines helped readers recall information from a series of news stories.
  Analysis of reader protocols provided evidence that headlines helped re
          aders organize information and integrate it with prior information in
        Although numerous studies have profiled newspaper readers and their uses
          of newspapers (Stone, 1987), fewer studies have looked at how readers
       cognitively respond to various elements of the news package (Bell, 1991).
          One part of the newspaper package that has been under-researched is
         effect of headlines on readers' processing of the news (Hilliard,
        Typically, discussions of headlines conceptualize them as serving basic
          functions of allowing the reader to easily scan a page for news and
     attracting the reader to a story (Baskette, Sissor, & Brooks, 1986).  The
          purpose of this present study is to conceptualize news headlines in a
       different way: as texts that prepare readers for processing information
          the story, pre-texts that may help the reader understand the news. The
        principal question for this study looks at how headlines affect reader
        processing of news stories.
        In light of recent attempts by the industry to develop an electronic
       newspaper (Christopher, 1994; Morton, 1993), headlines may be even more
         important than they are for a traditional newspaper.  To make it easier
 readers to enter the newspaper, section fronts of the electronic newspaper
 are being designed so that they feature headlines and summary paragraphs
          (Christopher, 1994).  Under this design, a reader would then have an
 to select the complete story.  Headlines will act as electronic gates for
          the reader of the future, and this study tries to explain how those
          will work.
         The consensus of reading research is that any device that prepares
      readers for new material will increase the probability that the readers
         will be able to later produce the new information (Kulhavy, Schwartz, &
         Peterson, 1986). The question is untested and open, however, whether
      headlines function as prereadings that would prepare the typical newspaper
          reader and affect comprehension.
        For this study, headlines are conceptualized as being a form of
  pre-reading, or an advance organizer (AO).  Ausubel (1960) conceptualized
          AOs as pre-reading events that appear to help readers use background
      information they already hold.  These organizers bridge the gap between
         what the reader already knows and what the reader needs to know.
   Information in the AO becomes an anchoring post or scaffolding for ideas
          that will come later (Ausubel, 1963). A series of studies by Mayer
          1987, 1984, 1983, 1980, 1979) support Ausubel's theory and expand the
       concept of advance organizers to concrete material such as graphic
 It may be possible that, besides attracting reader attention to stories,
          headlines may be bridges, text organizers that help readers process
        Research on Headlines.  Research on the effects of newspaper headlines has
 been mixed, and there have been only a few studies. In 1953, Tannebaum
         complained that almost no research had investigated the psychological
       effects of headlines. In 1963, other researchers (MacLean, & Kao) noted
         that although the effects of page design and headline typeface had been
         subjects of extensive research, the effects of headline content on
    newspaper readers had undergone little analysis. In 1991, Hilliard found
          that little had changed since 1963: the psychological effects of
          headlines remained an unknown factor.
        Yet, the overall evidence suggests that headlines do play an important
         role in readers' information processing.  For example, headlines can
 attitudes.  Headlines that emphasized "bad war news" during World War II
          were more effective in instigating a positive attitude toward
 in the war effort (Allport & Lepkin, 1943) than were headlines with
      positive news.   One early study (Deutschmann, 1956) found that headlines
          serve as cues for readers as they seek interesting stories. Headlines
          ract attention or guide selection, but they also influence decoding of
          message, according to Tannebaum (1955), who reported in 1953 that the
       positive or negative slant of headlines affected how readers judged the
         guilt or innocence of an accused criminal in a news story.
        In a similar experiment measuring how readers respond to headlines that
          contain innuendo about a person, a sociology study (Wegner, Wenzlaff,
       Kerker, & Beattie, 1981) found that subjects rated the protagonist more
         positively when the headline was neutral, while they rated the person
in a
          negative way when the headline contained innuendo or incrimination.
But a
          study by Leventhal and Gray (1991) found that innuendos in headlines
had no
 effect on memory or attitudes toward a crime victim, the accused, or the
          criminal's sentencing. But, it appears that the sensational story
   topics--rape and abuse--may have skewed the memory tests.
        Contrary to Leventhal and Gray (1991), Pasternack (1987) found that
        libelous statements in a headline and story led people to judge a man
          likely to be a thief.  In comparing the effects of libel in the story
or in
 the headline, Pasternack found that the libelous headline showed a greater
 effect in increasing judgments of guilt.  In discussing his results,
       Pasternack (1987) made an important point:  In the real world, headlines
          may be even more important than text because readers often will look
          at a headline, skipping the story or skimming it.  Such superficial
     attention to the headline is likely to increase the effect of the headline.
        In investigating the negative effects of ethnic references in crime
      reporting, Winkel (1990) found evidence that after reading a series of 40
          headlines, subjects were more likely to overestimate the number of
          committed by an ethnic group other than the one to which the subject
      belonged.  Winkel (1990) suggested that headline readers used information
          in headlines to make generalizations about behavior of ethnic group
     members, forming stereotypical judgments about a group of people based on
          the negative characteristic emphasized in the headline.
        Headlines are so powerful that the topic of the summary has a  greater
         affect than the slant in forming readers' attitudes toward a candidate
        (Geer, & Kahn, 1993).  Geer and Kahn suggested that this "surprising"
       amount of influence may be due to readers' attitudes being primed by
      certain topics.  Headlines, they said, may guide readers' choosing certain
          information to encode. Thus, headlines affect what people learn.
        Overall, research on newspaper headlines supports the idea that headlines
          have an effect on readers.  Winship and Allport (1943) appeared to be
 when they argued that newspaper headlines create images in the minds of
          readers and these images have important psychological effects on
        Process of Reading a Headline.   To study how readers process headlines,
          Perfetti, Beverly, Bell, Rodgers, and Faux (1987) completed six
 to test whether textual space constraints imposed on headlines affected
          the comprehension process.  The researchers noted that the Spartan
 to headline content--the omission of verbs and articles in
          headlines--created syntax that challenged a reader to interpret the
     meaning.  They hypothesized that instead of parsing headlines as if a
       headline were a sentence, a reader would heavily favor problem-solving
        techniques while processing a headline.
        By using a series of ambiguous headlines, the researchers (Perfetti et
         al., 1987) found that subjects parsed headlines in the same way that
          comprehend other written language.  Overall, headline comprehension
         very difficult, and yet readers could not overcome their reading habits
          use shortcuts or other knowledge even when those other strategies
         have been an advantage.  Ambiguous headlines took longer to read not
      because there were multiple meanings to choose from but because it took
         longer to achieve any meaning.  Readers looked at syntactic structure
 rather than semantic plausibility.
        In one of their experiments (Perfetti et al, 1987), the researchers found
          that time to read a sentence was longer when preceded by an ambiguous
       headline than an unambiguous headline.  The researchers suggested that a
          reprocessing effect--readers had to reread the sentence because it did
          meet the expectations influenced by the headline--accounts for the
          processing time.  This finding suggests that headlines can affect the
       processing time for reading a news story, and that there is a strong link
          etween a headline and the material that follows.
        Schematic Framing.  Underlying the theoretical work of Mayer and Ausubel
          in promoting advance organizers as learning devices is the need to
          how organizers work.  Ausubel (1963) tied his ideas of cognitive
          to Bartlett's ideas (1932) that "schema," a mental frame based on
         knowledge, was an organizing structure.  Geer and  Kahn (1993), too,
          the broad idea of schema theory to explain the priming result that
         The positive effects of various text features on reader comprehension
         have been explained through schema theory (e.g., Beck et al, 1991).
          Valencia, Greer, and Wardrop (1991, p. 144) have declared that the
          of schema theory on understanding reading comprehension "has been
          short of revolutionary."  Schema, according to various scholars,
          that what a reader stores in memory is determined by a type of
         process that is similar to a knowledge frame (Alba & Hasher, 1983).
          frame selects or even alters a person's experience so that the
          representation of the experience that is stored makes sense and is
    consistent with other experiences.  Schema activates a reader's prior
       knowledge, which aids interpretation of new information (Alba & Hasher,
         1983).  Generally, schema refers to the general knowledge that a person
         holds about a subject matter, event, or experience.  In general, schema
         theories hypothesize that prior knowledge that is related to new
  experiences helps improve acquisition, retention, and retrieval of new
        information (Alba & Hasher, 1983).
        Biocca (1991) has proposed a theory of schematic frames for the processing
 of television, based on the idea that viewers construct mental models of
          information.  To understand memory and processing, Biocca outlined a
      framework for studying viewers' representations of media messages.
     Although he applied his work directly to television commercials for
     political candidates, the idea of semantic frames should apply to the
       mental models readers construct as they engage a text.
        A reader forms a mental model of the meaning of a news story based on the
          model presented by the editor who wrote the headline, on the model
          by the writer of the story, on the abilities and knowledge of the
          and on the mental states of the reader as the reading goes on.  While
          words on the page may be static and fixed, the meaning in the mind of
          reader constantly changes (Biocca, 1991).
        Mental models are useful in thinking about the ways that readers build
         models of language (McNamara, Miller, & Bransford, 1991).  According to
         theories (Johnson-Laird, 1983, van Dijk & Kintsch, 1983), readers
          text at least two ways: at the propositional level (propositions
contain at
 least one noun and predicate; they are the smallest unit of knowledge that
 can stand as a separate assertion); and by constructing mental models that
 are "analogous in structure to the events, situations, or layouts de
      scribed in the text" (McNamara et al., 1991, p. 493).  Mental models do
 retain verbatim text structures, but they can help recall of events,
       relevant places, prominent characters, or objects associated with main
        characters (McNamara et al., 1991).
        Biocca's approach was to develop a theoretical model that specifies "the
          high-level semantic `programs' that appear to be operative in the
   processing" of the media message (Biocca, 1991, p.41).  He proposed that a
          selection of cognitive procedures, or schematic frames, were active in
        generating the mental models.  These frames access a person's schemata.
          "They are frames because each extracts and arranges specific
          from the semantic frames of the message.  Schematic frames are
          processes that organize application of the viewer's schemata to the
     semantic frames (codes) of the . . . message" (Biocca, 1991, p. 41).
        The frames, which represent psychological operations, were based on
      research and theory associated with psycholinguistics, discourse
  processing, and persuasion research (Biocca, 1991).  Some of their
    properties are that the various frames work at the same time while working
          out the meaning from the words.  What happens in one frame may
          what happens in another frame.  Biocca (1991, p. 42) called working
         meaning the act of "calculating semantic values from the codes."  The
       semantic values in the various frames change repeatedly as a message is
         processed.  The frames proposed (Biocca, 1991) included:  1)
          frames,  2) "Possible world" frames,  3) Actant or agent frames,  4)
          of view frames, 5) Narrative frames,  6) Ideological frames,  and 7)
      Self-schematic frames.
        After reading a headline, a person should have filled in these various
         frames, which begin to alter and change as the reader engages a text.
        These frames appear to be similar to the mental structures that Ausubel
         (1963) described as influencing meaningful learning.  Despite the
          studies of advance organizers, little effort has been made to describe
          structures created by AOs.  Biocca's theory (1991) provides a way for
       understanding the function of pretext.  AOs, by providing a schematic
 for a text, may help the reader make inferences about situations, people,
          and objects that appear in a text.
        This present study concerns itself with one basic theoretical question:
          How do pretext messages--in this case, headlines--interact with a text
          increase understanding?
        Numerous studies show that students learn more from text plus
          illustrations than from text alone (Mayer, 1989).  In developing a
          for headlines, it is useful to note that headlines are similar to
   illustrations.  In fact, Waller (1979) developed the concept of "access
         structures" to describe the spatial and typographic cues that helped
      readers access and read text.  Those structures include illustrations,
        headlines, summaries, boxed materials, and even different typographic
       settings.  Headlines appear to be a specific case of a general example of
          message processing.
        Integration of Theories.  The goal of this present study is to apply
       Ausubel's (1963) theory of advance organizers to newspaper headlines.  If
          headlines are AOs, they should produce evidence of meaningful
          Ausubel (1963) stated that meaningful learning occurs when new ideas
          integrated with existing cognitive structures.  Advance organizers
         bring about that integration, according to Ausubel (1963).  An AO
brings to
 the surface those concepts already held in a reader's mind, familiar
       concepts that will help the reader understand the new material to be read
          (Ausubel, 1963).  If the reader's cognitive structure does not hold
         concepts that are relevant to learning the new material, the AO
          the needed concepts.  The AO is an anchor, a scaffold, a framework for
          reader.  Thus, a headline, if it is an AO, introduces material to be
      learned, and it prepares the reader for understanding the text that will
        Modeling the theory.  Figure 1 illustrates the various conditions that
         influence the effectiveness of headlines in helping a reader to learn
          a news story.  Information from the headline organizes the material
          the text in working memory and integrates the new material with prior
       knowledge held in long-term memory.  This process of organization and
       integration occurs repeatedly and often along with repeated connections
         with long-term memory, where a reader's schematic frame shifts and
          as the reader comprehends the text (Mayer, 1989).
Insert Figure 1 about here
        The frames become embedded in other frames or include other frames as new
          material is processed.  Successful organization and integration leads
          increased learning because the new material is accessible and
memorable for
 the reader (Mayer, 1989).  Of course, the effects of the headlines and
         text also are mediated by the characteristics and states of the reader.
        A headline by virtue of its typographic prominence and placement above a
          story provides strong cues for cognitive processes and monitors
          of reading and comprehension (van Dijk, 1986).  An effective headline
       provides a way of organizing the concepts that are being learned.
    Organization occurs in short-term memory, where material is made available
          from the senses to be integrated with prior knowledge held in
         memory.  Without organization of the new material, the reader's
learning is
 haphazard and inefficient (Mayer, 1989).  The effective headline helps the
 reader integrate concepts from the news story.  This implies that a
      schematic model formed after reading the headline is based on the
   connections between the information in short-term memory and prior
    knowledge.  The assumption is that the headline provides the reader with
          material for building a schematic model that is relevant to the text.
        Headlines, even when they are written in Spartan language, have clear
        effects on newspaper readers.  They can affect text organization and
      integration.  By affecting mental processes, they can influence learning,
          attitudes, and inferences such as judgments of guilt.  Although they
          not meet the classical definition of an advance organizer (Ausubel,
          1963), headlines may provide a frame that aids readers.
        The present study uses repeated measures and random assignment to control
          for environmental and individual differences and to focus on the
effects of
 headlines.  Further, the design will call for control of various message
          variables known to influence learning--length, topic, order--and for
      control of prior knowledge that individuals bring to the experiments.
        Theory leads to the following hypotheses:
H) Subjects will recall significantly more information from a news
         story with a headline than from one without a headline.
        If newspaper headlines function as advance organizers, they should produce
 evidence of meaningful learning.  Ausubel (1963) states that meaningful
          learning occurs when new ideas are integrated with existing cognitive
       structures.  Such integration can be measured by comparing aided recall
         scores for subjects who receive text preceded by a headline and
          without a headline.
        It would be useful to investigate the semantic frames of readers under two
 conditions: text with a headline and a text without a headline.  According
 to Biocca's theory (1991) of semantic framing, a reader's semantic frame
          will change as the reader fills in details from the text.  If readers
          a different starting point (headline vs. no headline), the details in
          various frames should be different, with the headline condition
          a more highly integrated schematic frame.
        Schema theory in general implies that schemata help readers make
   inferences, both correct and wrong, about the text they are reading.  The
          expectation of the present study is that reader protocols would reveal
 headlines affect the kinds of information that people will process and
         store in their memories. This leads to a research question:
R) Will  headlines make a difference in the schematic frames of subjects?
        If headlines make information processing easier for the reader, it should
          be measurable by comparing the protocols of  subjects who receive
 and those who do not read headlines before engaging a text.
        The aim of the following experiment was a simple one: to test whether a
          headline helped readers process a news story.  In other words, can
    headlines work as advance organizers?  The experiment looked at how
     headlines affect readers' memory of a news story.  In the main part of the
          experiment, subjects read news stories with and without headlines
         their reading times were recorded and used as the dependent variable.
        The second part of this study  was exploratory and used reader-generated
          protocols to analyze differences in the readers' memories according to
        whether they read a story with a headline or one without a headline.
        Subjects and sampling.  The subjects for this experiment were 36 undergraduate
            enrolled in the School of Journalism at a major state university.
The subjects were
         volunteers who received extra credit in one of their journalism classes
for participating
            in a volunteer subject pool.  The average age of subjects was 20.2
years, 80% were female,
 and 94% were white.  There was one African-American and one Native American in
        Materials.  An effort was made to make the news stories and headlines
consistent with the
 real world of newspapers.  Materials for this experiment were a stimuli package
of news
            stories and headlines along with a test questionnaire.
        Four stimuli news stories were selected from the local news sections of the
          Oregonian, a well-known newspaper of quality published on the Pacific
Coast (see
     Appendix).  Multiple stories were chosen to limit the possibility that some
            of one story might influence the results.  Further, exposure to
multiple stories mimics
            the natural reading environment.  People tend to read more than one
news story at a
        sitting.  It was hoped that multiple stories would make the laboratory
environment less
        The selection of the stimuli stories was not random.  The stories were
published during
            the first two weeks of October 1992, which was chosen because of the
upcoming election.
            About 40 stories were selected from this time period based on their
content.  The main
           attributes were that the story have information that was unfamiliar
to students in our
           geographical area, that the stories have information that was
learnable (such as election
            stories), and that the stories be of adequate length, about 200-300
        Story topics were chosen that focused on issues of low involvement that were
unique to
            Oregon.  This geographic limitation was designed as a control for
the prior knowledge that
 a typical college student would have about the information that would be tested
in the
            memory questions.  Government action and elections were considered
desirable stories for
            the experiment because their content is strongly associated with the
type of information
            provided by newspapers that is suitable for reader learning.  On the
other hand, feature
            stories, editorials, and other types of entertainment content (such
as sports) were
        considered to be beyond the scope of this study.
        The 40 stories were reduced to eight stories.  The main criterion for reduction
was based
 on the qualitative judgment of the experimenter, who decided whether a story
        generate an adequate number of meaningful questions for the memory test
and how well the
            story fit into a pattern of news representation.  From this list of
eight, four were
         randomly selected for this experiment: the water story, the Hispanic
story, the sign
         story, and the fishing story.
        Each of the four stories had two treatments: with and without a headline, for a
total of
            eight stimuli.  The headlines were written in a style typical of
modern newspapers: a
          two-line main head along with a two-line subhead in smaller type (see
Appendix).  Three
            experts judged that the stories and headlines represented typical
news stories.  Each
          judge received the four stories plus 16 other distracter stories and
judged them on a
          scale of 0-10, with 10 signifying that the story represented the kind
of story that is
           typically printed in a newspaper.  Each scored a perfect 10, while
some of the distracter
            stories scored a 0.
        The news stories were printed on 8 1/2 by 11-inch paper to look as if they were
     photocopies of real news stories.  They included bylines, datelines, and
multiple legs of
            10-point type.  The headlines were printed on separate sheets of 8
1/2 by 11-inch paper.
            The main head was 30-
point type and the subhead was 16-point (see Appendix).
        As suggested by previous experiments, headlines are unlikely to have effects on
            under four conditions: 1) if the subjects have previous knowledge or
have formed opinions
            about the subject (Tannebaum, 1953); 2) if the topic of the news
story is highly involving
 or more likely to increase the reader's attention (Mayer, 1989); 3) if the
material is
            not potentially meaningful (Mayer, 1989); and 4) if the reader does
not need help in
         organizing the material (Mayer, 1989).  The selection of news stories
was designed to
          control for the first three of these possible conditions.
A pre-test of the stories included a focus group session with seven students who
 that the stories were unknown to them, that the stories were boring (low
            and that the stories contained material that they believed was worth
         A series of four packets--each containing the four news stories, half with and
          without headlines--were prepared so that each subject read a total of
four messages, two
            from each condition (with or without a headline).  Story order in
the packets was based on
 Wagenaar's (1969) method for constructing N X N "digram balanced" Latin
Squares.  (See
            Table 4.1).  The packets were designed as a between-subject
manipulation.  This
    control-manipulation was designed to balance potential topic, order, and
        effects, which were tested in a post-hoc analysis.
Insert Table 1 about here
        General directions for the experiment informed subjects that they would read a
series of
            news stories, and that after reading the news stories they would be
asked about what they
            read.  The directions emphasized that they should read the news
stories carefully, but at
            a normal speed.
        Measurement.  A questionnaire, also on 8 1/2 by 11-inch sheets of paper,
included 20
          multiple-choice questions. (five for each story).  The last page asked
subjects about
          basic demographic data: age, gender, race (see Appendix B).
        A series of 32 questions (eight per story) were written.  Each question focused
        important information central to the stories, which was confirmed
through a pre-test of
            the questionnaire.  There was one correct answer for each question:
To eliminate amb
        iguity, the answer was an exact or nearly exact passage from the text.
Three plausible
            alternatives--both semantically and structurally different from the
correct answer--were
            constructed for each question.  A fourth alternative (to discourage
guessing) was "I am
            not sure or I do not remember."
        The questions and stimuli were pretested by a group of seven subjects, all from
         volunteer subject pool.  Questions that were answered correctly or
wrongly by more than
            five people were discarded or rewritten as needed to reach the goal
of 20 questions for
            the experiment.
        The 20 chosen questions were carefully constructed to capture those ideas that
         related to the major concepts of the stories.  Mayer (1989, pp. 48-49)
notes that valid
            major dependent variables in experiments designed to measure
learning "must involve recall
 of conceptual information, retention of material in verbatim format, and/or
          problem-solving transfer performance rather than the more traditional
measures of overall
            amount recalled and/or overall performance on comprehension tests."
        To assure that the 20 questions did reflect important concepts in the stories,
         judges, taken from the volunteer subject pool, read each of the four
stories and then
          rated the 20 questions along with 12 additional questions for their
importance to the
          story.  A bi-polar 7-point scale, (Not Important Idea vs. Important
Idea) was used.  None
            of the 12 additional question was rated as more important than the
20 used in this
       experiment.  The lowest score among the questions used was 3.75, slightly
below the middle
 score of 4, while the highest score for the questions not chosen was 3.25.
        Another pre-test was used to assure the validity of the questionnaire.  A group
of three
            judges, all professional journalists, read each of the headlines and
the questionnaire to
            determine whether any headline provided an  answer to any of the
questions.  The judges
            determined that none of the headlines provided information that
helped answer the
      questionnaire, except for one, which was rewritten.  The judges also noted
that they had
            no prior knowledge that helped them to answer the questions.
        Each subject answered all 20 questions. The reliability score for 20 questions
was Alpha
            = .68.
        Procedure.  Each subject received four news stories to read.  Subjects were
          assigned to one of the four stimuli packets. Instructions emphasized
that the subjects
           were to read the stories as if they were carefully reading a
newspaper.  Subjects were
           told that after reading the stories they would be asked some
questions about what they
           remembered about the stories.
        Subjects also were told that they had a 30-minute time limit in reading the
stories (this
 time limit was determined by the pretest).  A time limit was necessary because
            have shown that by slowing reading speed, comprehension increases
(Kintsch & Vipond,
         1979).  The goal was to have the subjects read carefully, but at a
normal rate of speed.
        After reading the four stories, the subjects returned their stimuli packets to
        experimenter.  The average reading time was about 16 minutes.  Subjects
were then given a
            questionnaire (Appendix B) to complete, and the order of questions
was the same as the
           story order in the stimuli packet.  The questionnaires were marked so
that the
   experimenter knew which of the four stimuli packets the subject received, but
           remained anonymous.
     Data analysis.  The dependent variable, Memory, was based on the number of
            answers to the multiple- choice answers.  Each subject contributed a
score from 0 to 10
            for each condition.  A paired t-test was used to determine if the
subjects remembered
          significantly more key ideas from a news story with a headline as
opposed to a story
         without a headline.  The conventional significance level of .05 was
        Main analysis.  The mean score for the headline condition was 5.472 with a
 standard deviation of 1.934.  The mean score for the no-headline condition
 was 4.889 with a standard deviation of 1.997.  The paired t-test, with a
          difference in the means of .5833 and a standard deviation of 1.811 and
        standard error of .302, indicated that the t-value was 1.93 with 35
 of freedom.  The one-tailed probability was .03.
        Order effects.  The experiment was designed to allow a post-hoc analysis
          of effects that might be due to story order and the repetition of the
       stimulus condition (headline or no headline). The analysis indicated no
         evidence of a systematic bias.
        Discussion.  The null hypothesis is rejected and Hypothesis 1 is
   supported: Headlines did help readers recall significantly more information
 from news stories.  Headlines appeared to help readers organize
  information.  Headlines did more than the traditional functions (Baskette
          et al., 1986) of attracting attention, summarizing the story, helping
 the news page, depicting the mood of the story, depicting the tone of the
          newspaper, and providing typeface relief.  Headlines helped readers
          and remember information from the news story.
        Materials for protocol analysis.  After a rest and particapting in  a
        distracter task, the subjects read two more stories.  The materials for
         this part of the experiment--instructions, headlines, news stories, and
         questions--were presented on computer screens of IBM-PCs and generated
          Micro Experimental Laboratory (MEL) software.  The stimuli included
         first three paragraphs from two news stories, an election story and a
       hospital story, along with a headline for each story (See Appendix).  The
          stories, a campaign story and a hospital story, were written by the
     experimenter, who based them on actual news stories from a regional
        Each subject read the two stories, one with a headline and the other
       without. There were four stimuli created: Story 1 with headline; Story 1
          without headline; Story 2 with headline; Story 2 without headline.
          treatment groups, with two stories each, were created so that story
          and condition were counterbalanced.
        After the second story, an open-ended question appeared on the screen.
          Subjects were given a sheet of paper to record their answer to the
    question, which asked them to retell the story in their own words.  The
         goal of asking the open-ended question was to require the subjects to
       report their memory of the SECOND story they read.  For half the
          this story would be preceded by a headline, for the other half, there
 be no headline.
        Instructions for the experiment told subjects that they would have to
        answer some questions about the stories they read.  Before each headline
          appeared on the screen, subjects were asked to carefully read the
 After the headline, instructions told subjects that they would read a news
 story, one paragraph at a time.
        Equipment.  IBM-PC computers were used along with MEL software.  The
       software is well-known in psychology and is often used in experiments.
        Measurement.   The headline or no headline treatment was the independent
          variable of this experiment.  Two judges, blind to the treatment
  conditions, analyzed the content of subjects' protocols.  There were nine
          protocols (one from each subject) available for each of the four
          or 18 protocols for each of the treatments (with headline or without
      headline).  Contents of the protocols were analyzed by proposition (van
         Dijk, & Kintsch, 1983; Kintsch, & Vipond, 1979):  Propositions contain
          least one noun and predicate; they are the smallest unit of knowledge
          can stand as a separate assertion.  The judges then used a scale
          by Cooke (1994), and modified for this experiment, for categorizing
         propositions as schematic frames.  Overall, in coding, the judges
agreed on
 86% of the propositions.  The disagreements by the judges were resolved
          through consensus.
        Protocol analysis.  An analysis of the protocols also was conducted.  The
          protocols had been expected to provide information about the
differences in
 mental models of readers who read a headline before writing the protocol
          and those readers who did not have a headline.  Coding of the
protocols was
 based on schematic frame theory (Biocca 1991) using a system proposed in
          Cooke (1994).
        A preliminary analysis of the protocols allowed for the creation of a
        coding sheet that condensed the categories for schematic frames to the
        * Possible World frame, where readers assign characters or actions to some
 possible world (e.g., this story took place in Portland).
        * Discoursive frame, where readers determine that a story is about a topic
 (e.g., the story is about feuds) or genre (e.g., this story is a tragedy).
        *  Agent frames, where readers state the role of agents, describe them in
          some way, discuss their internal states, or state the motivations or
          of the agents.  Agents in this case can be institutions and objects as
 as people.
        *  Point of view frames, where the reader uses a statement that
  acknowledges that the reader is viewing the world through some agent or
         other point of view.  This category was eliminated from the judges'
          sheets because of the limited nature of the protocols, which contained
          statements reflecting a point of view other than as a third-person
          someone who appears to be not really involved with the action.
        * Narrative frames, where the viewer references either an "event" (a
       single action that has occurred within the story) or an episode (the
 relation among a series of events or the consequence of an event).
        * Ideological frames and Self-schematic frames were eliminated from this
          analysis because there were no examples of these type of frames in the
        Thus, coding of the protocols was limited to the following frames:
     Possible World Spatial, Possible World Character, Discoursive Topic,
      Discoursive Genre, Agent Role, Agent Description, Agent Motivation, Agent
          Internal State,    Narrative Event, Narrative Episode.  Further, each
          the frames was given one of the following subcodes that described the
       dynamics of the organization of the message within the frame (Cooke,
  These included: Non-judgmental, which was the default coding and the most
 likely coding for a proposition that simply related a fact direct from the
 text; Confused, which was the subcode for mistakes by the reader;
    Value-laden, in which the comment reflected some negative or positive
       judgment; and Inferred, which described those comments that were not
      explicit in the text.
         Post-hoc analysis of protocols.  Coding of the protocols led to the
       classification of 602 propositions: 312 for the headline condition, 290
 the non-headline condition.  About 80% of the propositions mimic the text
          of the original stories.  Of interest in this analysis are the 114 p
     ropositions (about 19%) that represent the schema-based inferences of the
        In this section of the analysis, numbers mean less than the qualitative
          insights.  A clear pattern emerges from the coding: subjects who
          no headline provided protocols that recalled the stories in ways
          from subjects who read a headlined story.  The headlines clearly
          the mental models formed by the readers.
        These differences are seen in the following examples.  They are taken
 from two of the 18 subjects who read the story about political
      candidates.  One reader had no headline.  The text of this reader's
          protocol follows:
        Republican candidate (male) Meeker, 67, supports free trade with
             Mexico. He says it will create new jobs for the region. When asked
               a group of Hispanics (Hispanics for Unity) if he would remember
               once he was elected, he said he would consider hiring a Spanish
           speaker as a member of his staff.
        (Female) Democratic candidate Kelley, 35, is opposed to free trade
               with Mexico because it would cause American jobs to be lost, and
 unions would suffer. She said she hopes labor will speak up at
           election time as it usually does.
        This protocol was the most complete of any, recalling and listing 27 facts
 of the 44 in the story (average recall was 12 facts).  This organization
          was typical for the nine subjects who read a non-headlined version of
          political story.  Five subjects used this organization: listing all
         aspects of one candidate and then all the aspects of the other
          The other four subjects in the non-headline condition borrowed the
    organization of the story, which directly contrasted the candidates by list
          ing one candidate and then the other and then back to the first and
     finishing with the second.
        The following text is the protocol of a reader who had a headline on the
        It's election time and the Republicans and Democrats are sharing
              their views.  The Republican (Meeker) has promised Hispanics that
 will be an integral part of the future.  Meeker will hire a Hispanic
               if elected and make sure that jobs are available for them.  The
           Democrat is opposed to the Republican's idea of free trade with
           Mexico. The Democrat is looking after the labor groups.  He promises
               that their jobs are secure and that no foreigners will infiltrate
             their towns.  The Republican and Democrat agree that the real views
               will come out on election day. The majority's voice will be
        This protocol began with the inference that it was election time.  None of
 the non-headline protocols made that inference, while five of the nine
         with the headline did and did so at the start of the protocol.
          the headline did not mention that the story is an election story, it
          say that a candidate was seeking votes.  Yet this is not much more
    descriptive of an election than was the first paragraph of the story, which
 strongly implied that it was election time.  The main subjects of the
        stories were introduced as candidates for Congress and as foes.
        It appeared that for some of the headline readers, the headline helped
         them access a schematic frame for elections and this frame helped them
          organize the story.  But, as suggested by Mayer (1987, 1984), advance
       organizers, which headlines appear to be, affect the type of information
          recalled.  Likewise, research on mental models indicated that they
were not
 good structures for recalling verbatim text, but instead helped readers
          integrate themes and ideas into established schema (McNamara et al.,
  Therefore, subjects are more likely to make inferences representative of
          schema intrusions when the headline works as an advance organizer
         Mayer [1987] called assimilative encoding), while subjects without the
        advance organizer are more likely to recall facts such as ages, names,
          gender (what Mayer called additive encoding).
        Without a headline, readers began the story with a task orientation: "I
          will memorize facts because I know that this is an experiment and I
will be
 required to recall facts."  Instead of an election schematic frame, it
         appeared that the readers of the non-headline story either used the
     organization of the story or focused on the main agents in the story: a
         Republican, a Democrat, Hispanics and Labor.
        The other thing that happened was that seven of the nine headline
    protocols of the election story made some sort of value-laden comment about
 the agents in the stories, while only one such comment was generated in
          the non-headline version.  It should be noted that the headline
          no value-laden information.
        The influence of schematic frames also was apparent in the transplant
        story.  The headline in this case stated that a feud has ended with an
        agreement.  A schematic frame for feuds was used by some of the headline
          readers in their protocols.  Only three of nine in the non-headline
     condition mentioned the animosity among hospitals, which is mentioned in
          the final paragraph of the story, while seven of nine in the headline
       condition mentioned the animosity.  Further, the three in the
          condition mentioned animosity at the end of their protocols, the same
       organization as in the story, while in the headline condition, three of
 protocols mentioned the animosity in the first paragraph.
        Also, two other protocols in the headline condition began with a phrase
          taken directly from the story, "collaborative effort."   This phrase,
       mentioned by no one in the non-headline condition, was linked directly to
          the animosity paragraph, where it served as a bridge between the
          that there is animosity and why there is animosity.
        Analysis of the protocols suggests that a schematic model formed after
         reading the headline was based on the connections between the
          in the reader's short-term memory and prior knowledge.  The headline
      appeared to provide the reader with material for building a schematic
 that is relevant to the text.
        Schematic frames.  According to Biocca's theory (1991) of semantic
     framing, a reader's semantic frame will change as the reader fills in
       details from the text.  If readers have a different starting point
    (headline vs. no headline), the details in the various frames should be
         different, with the headline condition exhibiting a more highly
          schematic frame.
        In this study, it was hoped that schematic frame theory would provide a
          method for quantitatively indicating differences among protocols from
       readers of headline and non-headline stories.  But because of the short
         length of the stories, there were limited findings for the framing of
          stories between conditions. See Table 3 for a chart of the codings.
Insert Table 2 about here
        The coding for semantic frames describes the types of inferences that
        readers made in their protocols.  In the hospital and the campaign
          (especially in the campaign story), readers were more likely to infer
       cause-and-effect relationships in their narrative frames when they were
         furnished the headline.  And, in the candidates' story only, readers
          made inferences about descriptions of the agents at a much higher rate
 in the non-headline story.
        Two interesting patterns emerge when the campaign and hospital story are
          compared.  Because the campaign story has much more description of the
        agents, readers who did not have a headline focused on description (96
        compared to 62 for the headline condition).  It appears that readers who
          lacked a headline and thus a clue about how to organize the story
          on the text for a clue and fixated on agent descriptions as an
          device.  But in the hospital story, without as much description and
more ab
          stract agents (institutions vs. people in the campaign story) readers
          more likely to focus on the narrative frame of the story, regardless
of the
        The election story, for example, included the following inferences from
          the headline condition: readers described work shipped to Mexico as
 the Democratic candidate was motivated by her support of labor groups, and
 she talked to labor groups because she wanted their votes.  Nothing like
          this appeared in the non-headline protocols, where the most common
    inference was the fact that the Republican in hiring a Spanish-speaking
         staffer might hire a Mexican.
        With longer news stories, the results of the schematic frame coding might
          have been more extensive and allowed for a more rigorous analysis.
     Subjects appeared to be unwilling to stray beyond a straight narrative of
          the story.  About 98% of the comments were from the point of view of a
        third person who has no role in the story.  A more involving, longer
          (involvement was controlled for in this experiment) might have
         richer differences in protocols.
        Headlines helped readers recall information from a news story, and they
          also affected the type of information that readers process and store
         their memories.
        According to a post-hoc analysis, headlines led subjects to create mental
          models of the stories that differed from the models formed by subjects
          did not have headlines precede the stories.  The differences in the
          models were manifested by organization and content of the protocols.
          general, subjects who read headlines appeared more likely to  take
         from the headline in organizing information from the story and
          prior knowledge with knowledge from the stories.
        A theory of news headlines would suggest that headlines help readers frame
 the news they are about to read.  This type of framing also can be
     inferred from previous studies that suggested that headlines affected how
          people judged the subjects of stories, such as the guilt or innocence
of an
 accused criminal (Winkel, 1990; Pasternack, 1987; Tannebaum, 1953).  The
          results of the current study go beyond the concept that headlines bias
        thinking about the news subjects and suggest that headlines can affect
 readers remember from the stories.  This is a new finding and adds
     understanding to the role of the headline in newspapers.
         This study provides some clues about how headlines help readers process
          information.  Evidence from the reader protocols supports Mayer's
   theoretical explanation (1987, 1979) that AOs build external connections
          between the reader's prior knowledge and the text, integrating
         information with the reader's existing knowledge.  The protocols
          numerous examples of readers' real world knowledge intruding into the
       protocols when a headline is part of the news.  Aided by headlines,
 integrated the story information, which led to inferences and a mental
         model that reflected the "real world experience" of the subjects.
        When news stories were preceded by headlines, readers were more likely to
          include inferences and value judgments about the main agents of the
     stories.  Instead of just providing facts from the story, the readers
       showed that the facts had become integrated with information that had
          stored in their long-term memories. It appears that without a headline
          help integration and organization, some subjects memorized details.
        The analysis of the protocols according to Biocca's theory of schematic
          frames (1991) was exploratory.  It appeared successful, creating a
          picture of the mental processes affected by a headline.  The
 of each proposition of the reader protocols suggests that headlines create
 schematic frames for readers and the frames affect what readers remember
          about the story.  The pattern of memory supports Biocca's contention
that a
 theoretical model could specify the cognitive procedures, or schematic
         frames, that are active in generating mental models.  These frames
access a
 person's schemata and help organize the information from the semantic
        frames of the message (Biocca, 1991).
        Biocca's schematic frames allowed a quantitative analysis of the readers'
          protocols.  The analysis picked up differences between the two news
 used: one that had abstract agents (hospitals) and more concrete agents
          (political candidates).  In the hospital protocols, the agents are
          in a spatial world by all the subjects, who tended to remember a
          of events influenced by headline cues.  But when the agent is a
          candidate, the spatial world is less likely to be recalled while
          descriptions of the agents are more likely.  When a headline is
          fewer text-based descriptions are present, but there is an increase in
        descriptions that are based on reader inferences about political
        From this study, it is possible to sketch a theory for newspaper
   headlines.  A headline has typographic prominence and placement above a
         story and provides strong cues for cognitive processes.  An effective
       headline provides a way of organizing the concepts that a reader needs to
          understand a news story.  This organization occurs in short-term
          where material is made available from the senses to be integrated with
        prior knowledge held in long-term memory.  The effective headline helps
 reader integrate concepts from the news story and a schematic model is
         formed while reading the headline.  The headline provides the reader
          material for building a schematic model that is relevant to the text.
        Limitations.  This study has four major limitations.  First is the
     challenge to validity created by the laboratory setting.  Second is the
         nature of the stimuli, which, despite efforts to make the stories and
       headlines read as real newspaper stories, have low ecological validity
 compared with a real newspaper. The third major limitation is the use of
          student subjects, who are very different from the average newspaper
  Future studies--in the field using real newspapers--may yield interesting
        The fourth major limitation rests on the sample of messages.  Although two
 messages were used to limit the possibility that an idiosyncrasy of any
          one message was responsible for the results (Jackson, 1992), the
          sample was not random nor was it designed to be representative of
        Other limitations affect the understanding of the results.  There may be
          idiosyncracies in the messages, or the manipulations, or the subjects
          led to the positive and negative results reported in this study.  For
       example, the headlines used may be more effective than the typical
 used in a newspaper, or the content of the news stories may be a critical
          factor in bringing about the reported effects.
        Future studies.   As newspapers move toward their electronic futures, the
          role of headlines may become more important as readers are faced with
          increased number of messages.  Further studies of headlines could
          other content features of the headline summaries that might make them
          effective in helping readers learn.  The current study suggested that
 is a strong linkage between the headline and the lead paragraph as they
          help the reader form a mental model of the story.  Timely probes,
based on
          Biocca's schematic frame theory (1991),  could unravel how different
      message features affect reader processing of the news.
        Final summary.  This study supports the general idea, echoed in many of
          the studies involving headlines:  Readers take cues from headline,
          are more important than journalists have thought.  Although this study
          limitations, it clearly indicates that headlines help readers organize
 stories and learn from those stories.  Newspaper headlines appear to
       function as organizers, which integrate new ideas with readers' existing
          cognitive structures.
Alba, J. W., & Hasher, L. (1983). Is memory schematic? Psychological
      Bulletin, 93,
Allport, F. H., & Lepkin, M. (1943). Building war morale with
          news-headlines. Public
        Opinion Quarterly, 7, 211- 221.
Ausubel, D. P. (1960). The use of advanced organizers in the learning and
          retention of  meaningful verbal behavior. Journal of Educational
          , 51, 267-272.
Ausubel, D. P. (1963). The psychology of meaningful verbal learning. New
        Grune & Stratton.
Bartlett, F. C. (1932). Remembering: A study in experimental and social
        Cambridge, UK: Cambridge University Press.
Baskette, F. K., Sissors, J. Z., & Brooks, B. S. (1986). The  art of
      editing (4th ed.). New    York: MacMillan.
Beck, I. L., McKeown, M. G., Sinatra, G. M., & Loxterman, J. A. (1991).
        social studies text from a text- processing perspective: Evidence of
       improved                         comprehensibility. Reading Research Quarterly,
          26, 251- 275.
Bell, A. (1991). The language of news media. Oxford, UK: Blackwell.
Biocca, F. (1991). Viewer's mental models of political commercials: Towards
 a theory       of the semantic processing of television. In F. Biocca (Ed.),
          Television and
        political advertising: Vol. 1, psychological processes. Hillsdale, NJ:
Christopher, L. C. (1994). Closing the gap: Anxiety rises as news
        scramble to fill electronic niches. Quill, January/February, 27-29.
Clark, H. H., & Clark, E. V. (1977). Psychology and language. New York:
        Brace & Jovanovich.
Cooke, C. (1994). Signification, sex, and environmental ideology: An
      analysis of the
        semantic processing of a        televised environmental message. Master's
        School of Journalism and Mass Communication, University of North Carolina
        at Chapel Hill.
Deutschmann, P. J. (1956). The relationship between interest and perception
        headline-type stimuli. Dissertation,  Stanford University.
Dole, J. A., Valencia, S. W., Greer, E. A., & Wardrop, J. L. (1991).
      Effects of two types
        of prereading instruction       on the comprehension of narrative and
        text. Reading Research Quarterly, 26, 142-157.
Geer, J.G. & Kahn, K.F. (1993). Grabbing attention: An experimental
     investigation of   headlines during campaigns.  Political Communication, 10,
Hilliard, R. D. (1991). Nut grafs: A powerful tool for attracting, holding
        Newspaper Research Journal, 12(3), 76-90.
Jackson, S. (1992). Message effects research. New York: Guilford.
Johnson-Laird, P. (1983). Mental models: Toward a cognitive science of
        inference, and consciousness. Cambridge, MA: Harvard University Press.
Kintsch W. & Vipond D. (1979). Reading comprehension and readability in
        educational practice and psychological theory. In L. Nilsson (Ed.),
      Perspectives      on memory research (pp. 329-365). Hillsdale, NJ: Erlbaum.
Kulhavy, R. W., Schwartz, N. H., & Peterson, S. (1986).  Working memory:
        encoding process. In G.D. Phye, & T. Andre (Eds.), Cognitive Classroom
        Learning: Understanding, thinking, and problem solving (pp. 115- 140). San
        Diego: Academic Press.
Leventhal, G., & Gray, S. J. (1991). Can innuendos in headlines affect
        Psychological Reports, 69, 801-802.
MacLean, M. S., Jr., & Kao, A. L. (1963). Picture selection: An editorial
        Journalism Quarterly, 40, 230-232.
Mayer, R. E. (1979). Can advance organizers influence meaningful learning?
        of Educational Research, 49, 371-383.
Mayer, R. E. (1980). Elaboration techniques that increase the
          meaningfulness of
        technical text: An experimental test of the learning strategy hypothesis.
        Journal of Educational Psychology, 72, 770-784.
Mayer, R. E. (1983). Can you repeat that? Qualitative effects of repetition
         advance organizers on learning from science prose. Journal of Educational
        Psychology, 75, 40-49.
Mayer, R. E. (1984). Aids to text comprehension. Educational Psychologist,
          19, 30-42.
Mayer, R. E. (1987). Instructional variables that influence cognitive
        during reading. In Bruce K. Britton & Shawn M. Glynn (Eds.), Executive
        control processes in reading (201-216). Hillsdale, NJ: Erlbaum.
Mayer, R. E. (1989). Models for understanding. Review of Educational
      Research, 59,
McNamara, T. P. Miller, D. L., & Bransford, J. D. (1991). Mental models and
        comprehension. In R. Barr, M. L. Kamil, P. B. Mosenthal, P. D. Pearson
        Handbook of reading research, Volume II, (pp. 490-511). New York: Longman.
Morton, J. (1993). Growth of electronic media has made newspapers better.
          Newspaper Research Journal, 14(2), 18-22.
Pasternack, S. (1987). Headlines and libel: Is the "unit" approach most
        Newspaper Research Journal, 8(2), 33-39.
Perfetti, C. A., Beverly, S., Bell, L. Rodgers, K., & Faux, R. (1987).
        newspaper headlines. Journal of Memory & Cognition, 26, 692-713.
Stone, G. (1987). Examining newspapers: What research reveals about
        newspapers. Newbury Park: Sage.
Tannebaum, P. H. (1953). The effect of headlines on the interpretation of
        stories. Journalism Quarterly, 30, 189-197.
Tannebaum, P. H. (1955). The indexing process in communication. Public
        Quarterly, 19, 292-302.
van Dijk, T. A. (1986). News schemata.  In C. R. Cooper, S. Greenbaum
        Studying writing: Linguistic approaches. Beverly Hills: Sage.
van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse
        New York: Academic Press.
Waller, R. H. W. (1979). Typographic access structures for instructional
          text. In P. A.
        Kolers, M. E. Wrolstad, &H. Bouma (Eds.), Processing of visible language.
          New   York: Plenum.
Wegner, D. M., Wenzlaff, R., Kerker, R. M., & Beattie, A. E. (1981).
        through innuendo: Can media questions become public answers?" Journal of
        Personality and Social Psychology, 40, 822-832.
Winkel, F. W. (1990). Crime reporting in newspapers: An exploratory study
          of the
        effects of ethnic references in crime news. Social Behaviour, 5, 87-101.
Winship, E. C., & Allport G. W. (1943). Do rosy headlines sell newspapers?
        Opinion Quarterly, 7, 205-210.
 Table 1
Latin Square Design for Experiment 1
Story Topic and Condition Order
 (H= with head; NH=no head)
Reading order   1st      2nd     3rd     4th
1              A         B        C       D
               NH        H        NH      H
2              B         D        A       C
               NH        H        NH      H
3              C         A        D       B
               NH        H        NH      H
4              D         C        B       A
               NH        H        NH      H
A = Water Shortage Story
B = Hispanic Protest Story
C = Sign Ordinance Story
        D = Fishing Proposal Story
 Table 2
Number of Propositions by Schematic Frames by Story and Treatment
                HOSPITAL                CAMPAIGN
              Head     No Head        Head   No Head
               n=9       n=9           n=9     n=9
Event          68         58           41       50
Episode         1          7            0        4
Inferred       20         10           35        11
Confused        4          3            3         3
Descriptive    18         17           62        96
Inferred        4          0           17         7
Confused        2          5            2         2
(Inferred)      0          0            7         0
Possible World
Spatial          9          9            5         5
* Coded 588 propositions, another 14 propositions were not relevant to the

Back to: Top of Message | Previous Page | Main AEJMC Page



CataList Email List Search Powered by the LISTSERV Email List Manager