This paper was presented at the Association for Education in Journalism and
Mass Communication in San Antonio, Texas August 2005.
         If you have questions about this paper, please contact the author
directly. If you have questions about the archives, email
rakyat [ at ] eparker.org. For an explanation of the subject line, 
send email to
[log in to unmask] with just the four words, "get help info aejmc," in the
body (drop the "").

(Jan 2006)
Thank you.
Elliott Parker
====================================================================

Building a Coherent Web:
Using Structure-Building Text and Hypertext To Facilitate
Engagement and Understanding of News About Complex Issues
1
JUNG-SOOK LEE COMPETITION
Building a Coherent Web:
Using Structure-Building Text and Hypertext
To Facilitate Engagement and Understanding
of News About Complex Issues.
Despite the Web's catapult of content into every corner of the country
during the past two decades, many questions remain about how Web browsers
navigate and process the potentially overwhelming amount of content. 
This study
utilizes a cognitive approach by synthesizing literature from text 
comprehension
and communication to investigate effects of text and link structure 
on the Web.
 From the outset, some scholars argued that the Web's network structure
was consistent with human memory (Bieber et al., 1997; Shirk, 1992) and that a
comprehender with the ability to control his/her own navigation will 
learn more
than those without the freedom of control (Young, 1996).
Others argued that one's interactivity eventually overwhelms the Web
user's limited capacity to comprehend content, inhibiting learning (Darken &
Sibert, 1996; McDonald & Stevenson, 1996). This claim is supported by 
data that
suggest the brain's limited capacity is quickly exceeded by a 
continuous string of
information (Lang, 2000).
Related to the level of user interactivity, some researchers investigated
differences between the more passive (or "linear") print medium and 
the arguably
more interactive (or "non linear") Internet structure. To measure 
interactivity,
some researchers recorded the choices that users made as they navigated Web
pages and links to other Web pages (Eveland & Dunwoody, 2001; Fredin, 1997;
2
Tremayne, 2004; Tremayne & Dunwoody, 2001). The outcome was that users of
more complex sites generally engaged in more interactive behavior and
demonstrated greater cognitive elaboration and subsequent recall of content
(Tremayne & Dunwoody, 2001). At the same time, others noted effects of
familiarity with a Web site on learning. For example, the time needed 
for users to
orient to novel Web sites replaces the time to actually learn content 
regardless of
users' self-reported Web expertise (Dunwoody, 2001).
Mixed results from research indicating either significant differences in
learning across media (Chen & Rada, 1996) or little difference across media
(Dillon & Gabbard, 1998) may explain why some conclude that it is the
organizational structure of Web content and reading patterns that are most
interesting (Eveland & Dunwoody, 2001). Other studies found varying effects of
story structure on readers' suspense, curiosity and enjoyment 
(Knobloch et al.,
2004) and some have correlated message structures with varying emotional
responses (Bower & Cohen, 1982; Knobloch et al., 2004; Lang, Newhagen, &
Reeves, 1996).
Based on this collection of findings, this study explores effects of the
relationship of news text with hypertext on the Web. Of particular 
interest is news
about complex issues (such as science and technology), which is exposed to Web
users with limited knowledge for the content. This exploration is 
based on claims
that four out of five Americans cannot read and understand the 
science section of
The New York Times (Miller, 2004). The general question therefore, is:
How does linearity in text hypertext links affect Web browsers'
engagement and understanding (learning) of news about complex issues?
3
Dependent variables focus on the audience's interest in and understanding
of the news content. This study has two goals. First, based on theory of text
comprehension, a model is proposed for the structure of text and hypertext.
Second, an experimental design is utilized to test the structural model by
measuring readers' engagement (interest) and understanding (comprehension).
Text Comprehension and Message Structure
Many journalism educators maintain that the traditional "inverted
pyramid" structure – which places the most current details first in a 
news story –
may be particularly important for news on the Web (Stovall, 2002). The pyramid
structure can draw scanning users in, encouraging them to read the 
story (Foust,
2005). While the pyramid structure may be effective in communicating news
that is relatively familiar to the audience, some psychologists argue 
that less
familiar news content is a form of discourse processing and not just a direct
representation of events (van Dijk, 1983). A story's lead sentence or 
paragraph is
often followed by unrelated details without clarification or 
explanation resulting
in poor coherence in the message (van Dijk, 1983, 1985, 1988). Poor 
coherence in
unfamiliar text may play a major role in how low knowledge readers develop
relations of the concepts embedded in the text (Roller, 1990).
Based on this theory and research of models of text comprehension such
as the Structure Building Framework (Gernsbacher, 1993, 1996; Gernsbacher,
1990a) and the Construction-Integration Model (Kintsch & van Dijk, 1978) this
study applies the importance of coherence to the comprehension of 
news content.
4
Fundamentally, text comprehension models predict that readers begin
building a mental representation of the text s/he is reading by first 
"laying a
foundation" and then map incoming information that relates to the foundation
(Gernsbacher, 1996).
A reader's mental structure is either enhanced by new information that
relates to the previously read information or suppressed by less coherent or
unfamiliar information (Gernsbacher & Faust, 1991a, 1991b). Readers with
limited prior knowledge may be forced to "shift" by developing substructures
(Gernsbacher, Varner, & Faust, 1990b).
Similarly, the Construction-Integration model (Kintsch & van Dijk, 1978)
posits that readers' comprehension depends on the ability to generate 
associative
links within the text to understand the content. This may be 
particularly important
when frequent topic changes in news contribute to discontinuity, 
forcing readers
to unscramble bits of information and then piece those bits together 
(van Dijk,
1985, 1988). How do these models relate to the structure of news text and
hypertext, which may be less coherent?
Linearity of news hypertext: axial vs. network
Previous studies have found that the relationship between authors and
readers suggest that structure of text and hypertext may be an 
important variable
and a significant dimension of difference across media (Eveland et al., 2002;
Eveland, Krisztina, & Seo, 2004). This interaction is further complicated by
inherit features of the Web itself (i.e. hypertext links) which may 
further affect
coherence of the message. For example, data indicate that the amount of
5
interaction on the Web - as defined by the number of link choices – are on the
increase. One analysis of 1,500 Web stories found that the average number of
hypertext links to "related content" increased significantly between 1997 and
2001 from three links per story to more than nine links per story (Tremayne,
2004). The increasing trend of links prompts further questions about the
relationships between interactivity, news content, coherence and reader
knowledge.
To pursue these questions, this study proposes to measure understanding
of news and the relationship of that knowledge to the coherent structures of
hypertext. Engebretsen (2000) adapted previous work by Van Dijk (1988) and
others (Figure 1) to theorize the importance of coherence in hypertext links.
______________________________
Insert Figure 1 about here
______________________________
Engebretsen differentiates between an axial link structure and a network
link structure (Figure 2).
The axial link structure consists of a central main node or "trunk" in
which interactive links relate to specific hypertext in the story 
(left diagram in
Figure 2). Each link within the axial structure provides the reader 
with the option
for supplemental information, but the structure does not offer 
readers access to
countless and potentially disorienting other links on the Internet.
6
______________________________
Insert Figure 2 about here
______________________________
A more common structure of ubiquitous links on the Web is what
Engebretsen calls the network link structure (right diagram in Figure 
2). Network
structures consist of interconnected links that offer readers options 
more than one
link. The network structure can connect Web users to an unlimited number of
texts, topics or subjects. The reader chooses which links are accessed and the
order in which the links are accessed.
Based on the research in cognitive psychology, there are at least two
theoretical arguments why the axial structure may provide the 
strongest support
for understanding of complex news story. First, recall from the Structure-
Building Framework that incoming information during the reading process
triggers activation in a reader's mental structure. Depending on the 
subsequent
information read, a reader's mental structure is either enhanced by new
information that relates to previously read information or suppressed 
by unrelated
(i.e. unfamiliar) information. Secondly, according to the 
Construction-Integration
model, relational density is the number of associative links between a first
sentence (in a story or paragraph) and subsequent sentences. In a 
hypertext story,
an association in links does not necessarily mean an association in 
story content
(or story coherence).
7
Based on these arguments, the axial structure could more reliably expose
low-knowledge readers to coherent information that consistently enhances
activation and increases the chances of relational density of 
subsequent text that is
read online. For the network structure to satisfy such a process, 
each node or link
in the story with links would need to provide the relational density 
for activation
of a Web user's appropriate prior knowledge. It is unrealistic to 
assume that a
writer of hypertext news has the time and the resources to verify that such
parameters are fulfilled in every link of every story.
While some believe that the Internet offers proper use of contextual
information because the medium satisfies users who want shorter, fact-driven
accounts as well as users wanting context, interpretation, and 
opinion (Tremayne,
2004), it is not clear how such contextual information might be affected by
coherence of links when that information is read by those without prior
knowledge of the content.
To illustrate the potential complexity of one story, Tremayne (2004)
describes a CNN.com news story consisting of two to three paragraphs that
present eight different events in the War on Iraq. The story features numerous
links to additional material including: (1) the full 800-word story; 
(2) two other
stories related to one of the incidents; (3) two slide shows; (4) 
five video reports;
and (5) two interactive maps. Tremayne notes, "Most of these pages 
have links to
still more related material" (p. 238).
Even if the linked material related to the topic of the main story, the
question is whether one can assume that the information provides the coherence
and explanation to facilitate readers' situational understanding. In 
other words, is
8
the importance of coherence in text for situational understanding more or less
important to hypertext content?
Some might respond that the dominant path for textual coherence is
always when text is read line by line (Bolter, 1991). Others argue 
that the popular
belief that the Internet's manipulability mass of information can 
enhance human
learning is a myth (Dillon, 1996). "To date, the claims have far exceeded the
evidence and few hypertext systems have been shown to lead to either greater
comprehension or significantly better performance levels. Clearly, 
mere exposure
to information is not enough for learning to occur, which we really 
always knew"
(p. 31-32).
The assumption in this study is that regardless of the number or the
content of nodes, the next line of hypertext to be read by a Web user must
maintain explanatory structure building if the reader is to be 
engaged and if the
story facilitates application of the reader's prior knowledge. This may be
particularly important to news about complex stories about science and
technology since studies have found that only 10 percent or less of 
science stories
in 70 U.S. newspapers provided explanations of scientific terms (Long, 1995;
Long et al., 1991; Long, 1991). Accordingly:
Hypothesis 1: Explanatory structure-building news will produce greater
situational interest for stories with coherent axial links than stories in the
traditional inverted pyramid structure with network links.
Hypothesis 2: Explanatory structure-building news will produce greater
situational understanding for stories with coherent axial links than stories
in the traditional inverted pyramid structure with network links.
9
The model proposed in this study posits that a reader's interest in reading
unfamiliar news content relates to the readers situational 
understanding for that
news (Figure 3). After reading a story's lead sentence or paragraph, 
the reader's
cognitive system is prepared for decisions to either continue or stop 
reading and
interpreting subsequent text. In the case of news on the Internet, where
competition for readers' attention can be literally overwhelming, the 
structural
functions of a story's first sentence(s) could mean the difference 
between a reader
continuing or terminating his/her engagement. Coherence of text and hypertext,
therefore, should ultimately affect understanding of content.
METHOD
Design
A 2 (text structure) x 2 (hypertext structure) mixed factorial design. The
within subjects factor of text structure varied on two levels: 
original (inverted
pyramid) or ESB (explanatory structure building). The between subjects
independent variable story type consisted of the two levels axial 
hypertext vs.
network hypertext structure.
Participants
The sample included 301 undergraduates (73% female, 27% male, 48%
sophomores, 28% juniors, mean age = 19.8 years, SD = 1.75) recruited from four
courses at a large midwestern university including: an introductory 
meteorology
class, an introductory composition course, and both an introductory and
intermediate level journalism course. All students earned extra 
course credit in
exchange for their participation, which was voluntary.
10
Materials
The original text from two news stories from the New York Times were
used as stories in the traditional inverted pyramid structure (Appendix). The
content of story one (health), dated July 24, 2003, covered cancer 
research and
consisted of 642 words. Story two (technology), dated February 12, 2004,
reported nanotechnology and consisted of 775 words. These stories were chosen
because: (1) Both stories appeared on the New York Times' Web site, but not on
the home page; (2) Both stories contained content about science and technology
and; (3) Both stories appeared more than six months prior to the study.
Operationally, each original New York Times story was modified into the
explanatory structure-building format by performing theoretically motivated
operations. In accordance with van Dijk's analysis (1988), each 
paragraph of the
story was categorized by content (i.e. history, context, reaction, 
explanation, etc.)
Paragraphs were then rearranged to conform to the concepts of coherence and
explanation previous discussed in the literature. For example, number of
paragraphs in the health story was reduced from 15 to 13 and 
"credits" (or details
about the researchers and their institutions) and, in the ESB 
structure, the credits
were moved to after the story's situation is explained to the reader. In other
words, explanatory text was added to facts there were originally presented in
order of importance.
After arranging paragraphs for structure building, sentence-level relational
density and explanation was analyzed by dividing each sentence into individual
"idea units," where each unit conveys a single event, state, or action (Mayer,
1985). Mayer's goal is to manipulate reading strategy by applying 
operations to
11
target idea units. Specifically, instead of presenting ideas in order of
"importance," Mayer manipulates text design to encourage a reader's 
reflective or
elaborative processing. Mayer does this by organizing text around key 
idea units
instead of dropping idea units into a text. According to Mayer, 
breaking a text into
idea units helps to identify which terms and processes need 
explanation so that
problem solving can occur.
Idea units may or may not supply the information for low knowledge
readers to make appropriate inferences for situational understanding. 
Therefore,
the third and final step is to identify which idea units are "explanative."
To illustrate, by using the New York Times health story used in this study,
the original contained 814 words and was dissected into 222 idea units. Within
those 222 units, 55 scientific terms were identified. Of the 55 
scientific terms,
only two were accompanied by explanatory text, indicating that approximately
96% of the unexplained scientific terms in the story required low knowledge
readers to use prior science knowledge to generate the appropriate 
inferences for
situational understanding. The rewriting of original sentences or 
paragraphs was
minimal. Modified versions of each story appear in appendix.
Participants were asked to indicate their level of interest in continuing to
read the story after the story's lead paragraph by clicking to select 
one of five
statements representing a scale of 0-6 (with 0 = participant wishing 
to terminate
reading and 6 = participant interested in reading more). After 
reading the story in
its entirety, participants again selected one of five statements to 
indicate how
interesting they found the story (with 0 = not at all interesting and 6 = very
interesting).
12
Measurement of situational understanding included a battery of different
question types, given the theoretical notion that a situational understanding
includes both recall of explicit text base information and free 
recall or sorting of
details requiring reader inferences. Therefore, in addition to 12 true/false
questions and a detailed multiple choice question about explicit 
details presented
by the text base, participants were required to apply prior knowledge and
generated inferences to accurately sort 15 different terms and processes into
contextual categories.
So that both stories were approximately the same word length, the last
sentences of the nanotechnology story (the longer of the two stories) were
removed. Second, two sets of five links (one set representing axial 
links and the
other set network links) were added to both stories. Both sets of 
links related to
the same content within the story except that the network links 
connected to an
external Web site related to the story. Axial links (embedded within the text)
connected to separate window that displayed a brief explanation of 
that particular
term. None of the axial links took readers to an external site with 
additional links.
Readers only needed to close the link window to proceed with the main 
story text.
In terms of content, based on comments from study one related to reader
interest in the health story, most occurrences of the word "tumor" 
were changed
to either "growth" or "abnormal tissue." There were structural 
modifications to
the texts of either story.
13
Procedure:
Approximately two weeks prior to the experiment, participants registered
online and completed the same 12 items of civic scientific literacy 
used in study
one plus self-reported levels of interest and levels of familiarity 
for various types
of news, including news about health, technology, sports, and politics. Each
online registrant received an E-mail confirming one of six 
experimental sessions
to which they were assigned and a "subject ID" number.
Upon arrival at one of two windowless computer labs, each containing
twelve identical Macintosh iMAC computers, participants were presented a
consent form for signature and told to choose any one of the computer 
stations to
begin. Each consent form included a number corresponding to a numbered menu
displayed on each computer. (Each number on the computer menu represented the
text version of the stories to be read.) After signing their consent 
form, each
participant read the same on-screen instructions then clicked the menu item
indicated on their consent form.
Before clicking a second button to begin the experiment, each participant
entered their "subject ID number" so that their data could be 
associated with their
online pre-test completed two weeks prior. Each participant was then 
exposed to
the health and technology stories, one in the original (inverted 
pyramid) structure
and the other story in the modified (explanatory structure building) 
format. To
reduce the effects of order, the presentation of stories was counter-balanced.
Similar to study one (but on a scale of 0-5), participants indicated their
level of interest after reading each story's lead paragraph and after 
the entire
story. After reading both stories, each participant completed an online
14
questionnaire, including how many of the five links the participant 
clicked in each
story and, if clicked, how useful they found the links to be. 
Identical to study one,
participants completed the same sequence of free recall, sorting and 
true/false
questions used in study one with the exception of changes of a few 
questions to
accommodate the shortening of technology story and a final question 
asking each
participant for their cumulative grade point average. After clicking 
the "submit"
button participants were exposed to a "thank you" screen, the option 
of reading a
debriefing of the experiment if desired, and instructions to please 
quietly leave the
lab.
Results
As a manipulation check for story familiarity on a scale of 0 to 5 (with 0
equating to no familiarity and 5 representing much familiarity with 
the content),
scores were less than one for both stories (M = .98, SD =1.31 for the 
health story
and M = .42, SD = .91 for the technology story). 56% of the 
participants reported
"0" familiarity with the stories. Only 3 participants reported a "5" 
and these three
were excluded from the study.
To test for homogeneity across participants, an analysis of variance
(ANOVA) detected no significant differences between the axial and network
hypertext groups in pre-test measures of civic scientific literacy, 
the amount of
time on the Web, the amount of time browsing online news, the number of either
science courses or math courses and grade point average.
15
Table 1 first presents data each of the four repeated measures of text
structure. The four groups are then divided by the between subject factor of
network links versus the axial link condition.
Network links
For group, 1, situational interest, (M = 1.10, SD = 1.31) for the technology
(structure building) story was significantly less than interest in 
the original health
inverted pyramid story (M = 2.39, SD = 1.26). ANOVA indicated an F (1,78) =
121.816, p = < .001, eta .61. Situational understanding for the 
technology story in
the structure building format (M = 28.56, SD = 9.09) was also less than the
understanding of the original health text (M = 37.32, SD = 8.08). The 
difference
was also statistically significant with F (1,78) = 1400.685, p = < 
.001, eta .95.
These data failed to support hypotheses 1 and 2.
For the second network link group, the original technology inverted
pyramid story produced greater situational interest, (M = 2.31, SD 
=1.32) than the
modified health story (M = 1.91, SD = 1.33). The difference was significant F
(1,73) = 22.652, p = < .001, eta .24. H1 is not supported. Conversely, the
original technology inverted pyramid story produced less situational
understanding (M = 30.20, SD = 8.00) than the modified health story 
(M = 34.54,
SD = 8.71), and this difference was significant F (1,73) = 1331.872, 
p = < .001,
eta .95. H2 is supported. Figure 4 illustrates the trends in interest and
understanding for groups 1 and 2.
Axial links
participants in the axial link structure produced results more 
consistent with what
one would expect when investigating coherence in text and hypertext.
technology story (M = 1.33, SD = 1.26) was less than the 
self-reported interest in
the health story modified for coherence (M = 2.43, SD = 1.35) and 
this difference
was significant F (1,73) =138.16, p = < .001, eta .651. Similarly, situational
understanding of the original (IP) technology story (M = 27.12, SD = 8.66) was
also less than the understanding of the modified health version (M = 
36.85 SD =
7.84) representing a statistically significant difference F (1,73) = 
1404.33 p =
<.001 eta .95. Both H1 and H2 are supported.
= 1.80, SD = 1.34) was greater than interest in the original health 
IP story (M =
1.57, SD = 1.40), and the difference was significant F (1, 73) = 
43.43, p = < .001.
Similarly, the modified technology story with axial links (M = 32.59 
SD = 8.38)
scored higher in understanding than the original health news text (M 
= 27.47, SD
= 8.26) producing a significant difference F(1,73) = 1187.69 p = <.001. As in
group 3 with the axial link condition, H1 and H2 are both supported 
by the results
of group 4. . The means of both groups are summarized graphically in Figure 5.
16
Unlike the results from the first two network link groups, data from
Specifically, for group 3, reader interest in the original (inverted pyramid)
Finally, in group 4, situational interest in the technology ESB version (M
17
DISCUSSION
This research proposed that unfamiliar and complex news (in this case,
news about science and technology), structured in the traditional 
inverted pyramid
with network links might reduce interest and understanding of the 
news content.
The rationale was that such a format may limit explanation of key 
concepts to the
audience, provide greater incoherence and impute reader familiarity for the
content.
Results from two stories in two link structures elucidate how message
structure can have profound effects on the level of readers' understanding and
interest in important news issues. Therefore, based on a rich history 
of research in
cognitive psychology about how individuals process textual 
information, a claim
could be made that, collectively, news reporting information about complex
issues, structured for more coherence and explanation, could enhance the
engagement of low-knowledge readers and their subsequent understanding of
situations (terms and processes) in news. Clearly, link structure 
appears to play a
major role in whether a given text structure is effective.
Regardless of a consistent preference for health news over technology
news by readers tested in this study, explanatory structure building 
with axial
links appeared to significantly enhance situational understanding of content.
One cannot, however, avoid the possible explanation from previous 
research that
more interactivity may also affect the level of content-specific 
recall (Tremayne
& Dunwoody, 2001). Based on this possible explanation, a post-hoc analysis
was conducted on the total number of links clicked by each of the four groups
18
tested. Interesting, as represented by Figure 6, the axial hypertext structure
appeared to spark more link engagement (as measured by the total number of
story links clicked).
Overall, participants exposed to stories with axial links were at least three
times more likely to click on a link than an a participant reading 
the same story
with a network hypertext link structure.
A closer look at interactivity by story and link condition, however (Figure
7), reveals that in the axial condition (n = 149), the difference in 
self-reported use
of the five links between the original stories (MaxialIP = .82, SD = 
1.32) and the
modified ESB text (MaxialESB = .69, SD = 1.23) was not statistically 
different F (1,
148) = 3.185, p < ns. In the network condition (n = 146), the 
difference in selfreported
use of the five links provided in the original ESB text structure
(MnetworkESB = .51, SD = .81) was more than twice the number of links used for
stories in the original IP text (MnetworkIP = .24, SD = .60) 
resulting in a significant
difference in link use F (1, 145) = 17.053, p < .001.
Also, Pearson correlations of reader interest with link use revealed that the
number of links clicked for the IP story was positively related to 
interest in the IP
stories r (146) = .26, p = <. 002, and number of links used for the 
ESB stories was
positively related to interest in ESB stories r (145) = .17, p = <. 046. Only
interest in the original IP stories (where fewer clicks were used) correlated
significantly with situational understanding stories r (148) = .35, p 
= <. 001. This
suggests that although readers reported more interest in the original inverted
pyramid stories (and consequently appeared to experience a deeper 
understanding
of the inverted pyramid stories it appears that more link use in the structure
19
building (ESB) story condition meant more reader engagement. A weaker
argument, based on these data, is that greater use of network links 
(independent of
the structure building process in text) appears to inhibit 
situational understanding.
In summary, the difference in link use for the stories with network links is
significant. This significance (based on the fact that more links 
were used for both
stories in the structure building version) may suggest that the greater use of
network links resulted in less interest and less understanding of the 
stories with
network links. If so, this suggests that the coherence of text can 
provide only so
much benefit to a reader on the Web. If this correct, the structure 
of links and the
corresponding interactivity related to these links would appear to be 
paramount to
a user's interest in and understanding of news on the Web.
Limitations:
As with any experimental study, exposing two science and technology
stories to participants in a controlled environment offers limited 
external validity.
Although this poses a limitation to this study, it is important to place into
perspective the goal of this study. The goal was not to simply 
measure how well
the audience comprehends a news story about science and technology, but how
the audience responds to unfamiliar news content using variations of text and
hypertext structures.
Although it could be noted that this study, like so many others, focused on
college students, it is this segment of the online population that, 
arguably, should
be of primary interest to those committed to more public engagement 
science and
technology (Miller, 1987, 1995).
Conclusions
with limited knowledge about the content? Results from this study suggest that
the answer is 'no," and some assumptions about the level of public 
understanding
of content such as science and technology may be premature without first
verifying the effectiveness of the message structure to engage the 
low-knowledge
audience for deeper understanding.
a story is structured textually and with hypertext links may reap substantial
benefits. Without such an effort, one might be content with the 
assumption that
news about complex issues, such as science and technology, should generally be
directed to only the "science attentive public." This research suggests that
opportunities may exist to increase the population of that audience.
20
Can we to assume that all news is appropriately structured for an audience
Data collected in this study suggest that paying particular attention to how
21
Figure 1. van Disk's hypothetical structure of a news schema.
Source: van Dijk (1988). Reprinted with permission from the publisher 
and author.
Figure 2. Connective axial link structure(left) vs. associated 
network links (right).
22
Source: Engebretsen (2000).
23
Figure 3: Proposed model of reader engagement
and situational understanding of complex news.
Group 1: Network Links
Situational Interest
Situational Understanding
Group 2: Network Links
Situational Interest
Situational Understanding
Group 3: Axial Links
Situational Interest
Situational Understanding
Group 4: Axial Links
Situational Interest
Situational Understanding
Table 1. Within subject means of text vs. hypertext structure
* = highest means per group
M
STORY 1
SD
Tech ESB
1.31
9.09
1.10
28.56
Health ESB
1.33
8.71
1.91
*34.54
Tech IP
1.26
8.66
1.33
27.12
Health IP
1.40
8.26
1.57
27.47
24
M
STORY 2
SD
Health IP
* 2.39
* 37.32
1.26
8.08
Tech IP
* 2.31
30.20
1.32
8.00
Health ESB
*2.43
*36.85
1.35
7.84
Tech ESB
*1.80
*32.59
1.34
8.38
Figure 4. Situational interest and understanding with network links
25
Figure 5. Situational interest and understanding with axial links
26
\
Figure 6. Total links used by text and hypertext structure
27
28
Axial Links Network Links
Figure 7. Total links used by story, text structure and hypertext structure
29
APPENDIX
Benign Viruses Shine on the Silicon Assembly Line
ORIGINAL New York Times version
Reprinted with permission from The New York Times Company.
(Page 1)
Living organisms do a fine job of growing crystals, like the ones 
that make up abalone shells, for
example. But there are lots of other inorganic materials, including 
those that make up
semiconductors, that living things haven't gotten around to 
producing. That may change, though,
with some help from a tiny benign virus and a professor at the 
Massachusetts Institute of
Technology.
[ Participant clicks to indicate level of interest in story (on a 
scale of 0-6)
before proceeding to next page ]
(Page 2)
Taking over where nature left off, the professor, Angela M. Belcher, 
has induced the virus to
produce, at last count, roughly 30 inorganic materials with semi 
conducting or magnetic
properties.
"We have forced organisms to grow some of the technologically 
interesting materials that nature
hasn't had the opportunity yet to work with," said Dr. Belcher, an 
associate professor of materials
science.
Now she and her team report in the journal Science that they have 
selectively altered the DNA in
their viruses to generate a variety of tiny wires made of magnetic 
and semi conducting materials.
Such wires may one day be part of the extremely small circuitry in 
the next generation of evershrinking
high-speed electronic components. "It's amazing," she said. "Not only 
does the virus
make this nice semiconductor wire at room temperature, but all the 
crystals are aligned."
The shape of the new wires offers not only beauty but utility as 
well, said William S. Rees Jr.,
director of the Molecular Design Institute at the Georgia Institute 
of Technology and currently on
leave at the Department of Homeland Security in Washington. "The 
entire field of nanoelectronics
depends upon the ability to mass produce cost-effective components," 
he said, "and she has
opened the door to this."
Most nano wires currently tend to be less than uniform in shape, he 
said, but Dr. Belcher has
produced highly regular forms. "Hers are all the same diameters and 
length," he said. "It's similar
to grass growing in your front lawn when it's well manicured and each 
blade is the same height."
Dr. Belcher's team uses the virus as a temporary scaffolding on which 
the crystals grow. The
viruses are first altered by the insertion of a few amino acid 
chains, called peptides, so that they
attract a particular material like zinc sulfide or cadmium sulfide. 
As the material starts to form a
crystal on the virus, Dr. Belcher adds elemental components of zinc 
sulfide or cadmium sulfide in
a solution, and the crystals grow into individual nanowires. Then the 
virus is baked away.
It is all a matter of affinity, molecular recognition and genetic 
programming, Dr. Belcher said.
"We programmed the virus to grow a particular material at a 
particular length," she said. "Then we
burned off the virus and were left with single-crystal semiconductor wires."
"Viruses are nice stuff to work with," she said, but she and her 
colleagues are starting to use the
technique with other organisms, too. Thomas N. Theis, director of 
physical sciences at I.B.M.
Research in Hawthorne, N.Y., describes Dr. Belcher's work as part of 
a relentless drive toward
miniaturization of complex electronic structures, and predicts that 
it will have a big impact. "This
kind of chemistry could revolutionize many manufacturing processes by 
making them less
expensive," he said.
Dr. Belcher has jointly founded a company, Semzyme, with Evelyn L. 
Hu, a professor of electrical
engineering at the University of California at Santa Barbara, to try 
to bring the technology to the
marketplace.
Dr. Belcher started her research into this method for making 
nanowires in early 2000, when she
took a library of viruses that were identical except for one DNA 
insert coded for the production of
a random peptide. She wanted to see whether, by virtue of the 
peptides they produced, some of
those viruses would attract and bind with semiconducting materials.
Most had no effect and were discarded. When they did find a promising 
virus, copies of it were
made by inserting it into a living bacterial cell. "Then we had a 
population of a virus with some
affinity to a semiconductor," she said.
The group tested for attraction under increasingly stringent 
conditions, replicating successful
candidates again and again. The procedure gradually grew more 
streamlined. "By now it only
takes us two weeks," Dr. Belcher said.
That accomplished, she began using the viruses to create nanowires. 
The process is independent of
the virus used. "This is actually a genetic tool kit for growing and 
organizing nanowires for these
semiconducting and magnetic materials," Dr. Belcher said.
Dr. Belcher pointed out that the self-assembling her materials did 
was quite different from the
dreaded self-replication so often evoked by foes of nanotechnology. 
"These materials don't
replicate themselves," she said. They can be programmed only to 
assemble in a particular place in
a particular shape.
30
31
Benign Viruses Shine on the Silicon Assembly Line
Modified ESB VERSION
(Page 1)
Living organisms do a fine job of growing crystals, like the ones 
that make up some seashells, for
example. But there are lots of other materials, including those that 
make up semiconductors, that
living things cannot yet produce. That may change, though, with help 
from a tiny particle that
lives as a parasite in plants and animals. The particle is a harmless virus.
[ Participant clicks to indicate level of interest in story (on a 
scale of 0-6)
before proceeding to next page ]
(Page 2)
Professor Angela M. Belcher and her research team at the 
Massachusetts Institute of Technology
report in the journal Science that they have modified the genetics in 
viruses to generate a variety of
tiny wires called nanowires, which are made of magnetic and 
semiconducting materials.
Nanotechnology is a collective term that refers to developments on 
the nanometer scale, or one
millionth of a millimeter in size. Such tiny wires may one day be 
part of the small circuitry in the
next generation of ever-shrinking high-speed electronic components. 
"It's amazing," Professor
Belcher said. "Not only does the virus make this nice semiconductor 
wire at room temperature, but
all the crystals are aligned."
Taking over where nature left off, the researchers, has caused the 
virus to produce, at last count,
about 30 materials that have semiconducting or magnetic properties.
Dr. Belcher started her research into this method for making 
nanowires in early 2000, when she
experimented with a library of viruses that were nearly identical. 
She wanted to see whether,
some of the viruses would attract and bind with semiconducting 
materials. Most had no effect and
were discarded.
However, when researchers did find a promising virus, copies of the 
virus were made. "Then we
had a population of a virus with some affinity to a semiconductor," 
she said. The group tested for
attraction under strictly controlled conditions, replicating 
successful candidates again and again.
The procedure gradually grew more streamlined. "By now, it only takes 
us two weeks,"
Dr. Belcher reported. That accomplished, Dr. Belcher, an associate 
professor of materials science,
began using the viruses to create nanowires. The production process 
is independent of the virus
used. "This is actually a genetic tool kit for growing and organizing 
nanowires for these
semiconducting and magnetic materials," Dr. Belcher said.
She pointed out that the process her materials completed was quite 
different from the selfreplication
process often noted by opponents of nanotechnology. "These materials 
don't replicate
themselves," she said. They can be programmed only to assemble in a 
particular place and a
particular shape.
"We have forced organisms to grow some of the technologically 
interesting materials
that nature hasn't had the opportunity yet to work with," said Dr. 
Belcher. Most nanowires
currently tend to be different in shape, said William S. Rees Jr., 
director of the Molecular Design
Institute at the Georgia Institute of Technology, but Dr. Belcher has 
produced highly regular
forms."Hers are all the same diameters and length," said Rees. "It's 
similar to grass growing in
your front lawn when it's well manicured and each blade is the same height."
It is all a matter of similarity recognition and programming, Dr. 
Belcher said. "We programmed
the virus to grow a particular material at a particular length," she 
said. "Then we burned off the
virus and we're left with single-crystal semiconductor wires."
32
"Viruses are nice stuff to work with," Dr. Belcher said, but she and 
her colleagues are starting to
use the technique with other organisms, too.
Thomas N. Theis, director of physical sciences at I.B.M. Research in 
Hawthorne, N.Y.,
describes Dr. Belcher's work as part of a relentless drive toward 
miniaturization of complex
electronics and predicts that it will have a big impact. "This kind 
of chemistry could revolutionize
many manufacturing processes by making them less expensive," he said.
The shape of the new wires offers not only beauty but utility as 
well, said William Rees "The
entire field of nanoelectronics depends upon the ability to mass 
produce cost-effective
components," Rees said, "and she has opened the door to this."
Dr. Belcher's team uses the virus as a platform on which the crystals 
grow. The viruses are first
altered so that they attract a particular material. As the material 
starts to form a crystal on the
virus, Dr. Belcher adds a solution of elemental components and the 
crystals grow into individual
nanowires. Then the virus is baked leaving only the wires.
In the future, Dr. Belcher hopes to make materials that are not only 
self-assembling but selfhealing,
too. "You design a circuit and if there's a break, it can heal 
itself," she said.
Dr. Belcher has jointly founded a company, Semzyme, with Evelyn L.
Hu, a professor of electrical engineering at the University of 
California at Santa Barbara,
to try to bring the technology to the marketplace.
A Budding Tumor Unmasked by the Vessels That Feed It
ORIGINAL New York Times version
Reprinted with permission from The New York Times Company.
(642 words)
(Page 1)
For a tumor to grow, it needs a good supply of blood, which it gets 
by switching on the body's
process of blood-vessel making, known as angiogenesis. Researchers 
are trying to develop drugs
to inhibit angiogenesis as a way of fighting tumors, but they need 
ways to make sure the
inhibitors, which have so far had mixed results, are effective early 
in therapy, long before the
vessels affect the tumor itself.
[ Participant clicks to indicate level of interest in story (on a 
scale of 0-6)
before proceeding to next page ]
(Page 2)
One computer-based imaging technology may have the potential to 
detect changes in the blood
vessels in and around tumors, signaling the power of a particular 
inhibitor. The technique, an
adaptation of conventional magnetic resonance imaging, or M.R.I., 
captures up to a thousand
images taken serially of a tumor before, while and after dye is 
introduced. Software analyzes the
images, characterizing what the dye (called a contrast agent) has 
revealed on its journey into and
out of the tumor - leakiness, for example, a hallmark of vessels that 
are being formed.
The technology, called dynamic contrast-enhanced M.R.I., is largely 
confined to research
institutions conducting clinical trials and should be considered 
experimental, said Dr. Peter L.
Choyke, a radiologist and chief of M.R.I. at the National Institutes 
of Health in Bethesda, Md. For
the past three years, Dr. Choyke has been working on refining the 
technique in collaboration with
33
Dr. Michael Knopp, a radiologist at the Ohio State University 
Comprehensive Cancer Center, and
other researchers.
Dynamic contrast-enhanced M.R.I. is one of several technologies that 
provide noninvasive images
of the creation of new blood vessels in animals and humans. It has 
shown particular potential in
analyzing extremely small blood vessels, Dr. Choyke said, and might 
therefore one day find wide
use in identifying tumors and monitoring therapies that inhibit 
angiogenesis. The process yields a
loop of images that can be viewed one after another. "This process 
reveals a more complete map
of regional vascular properties of a tumor than single snapshots 
taken with M.R.I. could," Dr.
Choyke said.
Characterized by chaotic flow patterns and tortuous paths, blood 
vessels in tumors are markedly
different from those in healthy tissue. Leaks are common. "Tumor 
vessels are full of holes, and
that allows the contrast agent to leak out readily," Dr. Choyke said. 
"That's one of the things we
measure." A judgment on how aggressive a tumor is can be based in 
part on this permeability, he
said. "You can characterize a tumor as highly vascular - that is, 
amenable to an angiogenic
inhibitor," Dr. Choyke said, in contrast to a lesion that does not 
have many blood vessels. The
process might be helpful in determining whether a biopsy is necessary.
Dr. Choyke cited a woman with a high risk for breast cancer whom he 
had examined recently.
"We saw a little area in the breast, a nodule," he said. "But it 
didn't enhance with the contrast agent
to suggest that it was a highly permeable vascular area, so it didn't 
have a pattern suggesting
malignancy." In such a case, he said, it would be possible to 
postpone a biopsy.
Dr. Knopp said that dynamic contrast-enhanced M.R.I. might prove 
useful in preventing incorrect
biopsy results. "We are recognizing that tumors are not a single 
entity, but a heterogeneous array
of features," he said. Dynamic contrast-enhanced M.R.I. can help 
guide where the biopsy is
performed. "If you have a bulky tumor, we can show where there is 
active tumor tissue and areas
not as representative of the tumor."
In the experimental method described by Drs. Knopp and Choyke in 
recent papers, a dye is
injected and scanning is repeated until about 10 minutes of data have 
accumulated. Algorithms
analyze the images and map how permeable the blood vessels are, how 
much blood is flowing and
the vessels' volume. Workstations with high-resolution displays can 
present colorized images of
the data in views that create a composite of many scans.
A Budding Tumor Unmasked by the Vessels That Feed It
Modified ESB VERSION
(638 words)
(Page 1)
For a tumor to grow in your body, it needs blood, which tumors obtain 
from your body's ability to
make blood vessels. Researchers are trying to stop tumors with drugs 
that stop the growth of blood
vessels, and computer images may show if the drugs are working. The 
process might help in
determining whether a surgical biopsy in needed, according to Dr. 
Peter L. Choyke of the National
Institutes of Health.
[ Participant clicks to indicate level of interest in story (on a 
scale of 0-6)
before proceeding to next page ]
34
(Page 2)
Dr. Choyke cited a woman with a high risk for breast cancer whom he 
had examined recently.
"We saw a little area in the breast," he said. "It didn't have a 
pattern suggesting malignancy." In
such a case, Dr. Choyke said, it would be possible to postpone a 
surgical biopsy.
Dr. Michael Knopp at the Ohio State University Comprehensive Cancer 
Center said that the
computer imaging might prove useful in preventing incorrect biopsy 
results. "We are recognizing
that tumors are not a single entity, but a heterogeneous array of 
features," he said. This imaging
technology can help guide where the biopsy is performed. "If you have 
a bulky tumor, we can
show where there is active tumor tissue and areas not as 
representative of the tumor."
The computer images, taken as dye is injected into the body's tissue 
can show the journey of the
dye into and out of a tumor. Up to one thousand computer images taken 
before, during and after
the dye is introduced into the body can indicate if new bloods 
vessels are being formed. In the
experimental method described by Drs. Knopp and Choyke in recent 
papers, the dye is injected
and scanning is repeated until about 10 minutes of data have 
accumulated. The images are
analyzed for how porous the blood vessels are, how much blood is 
flowing and the vessels'
volume. High-resolution color images create a movie of many scans.
"This process reveals a more complete map of regional vascular 
properties of a tumor than single
snapshots," Dr. Choyke said. Characterized by chaotic flow patterns, 
blood vessels in tumors are
markedly different from those in healthy tissue. Leaks are common. 
"Tumor vessels are full of
holes, and that allows the contrast agent to leak out readily," Dr. 
Choyke said. "That's one of the
things we measure."
The technology, called dynamic contrast-enhanced M.R.I. is one of 
several technologies that
provide images of the creation of new blood vessels in animals and 
humans. The technology has
shown particular potential in analyzing extremely small blood 
vessels, Dr. Choyke said, and might
therefore one day find wide use in identifying tumors and monitoring 
therapies that creation of
blood vessels. The technology is not yet in wide use, partly, Dr. 
Choyke said, because different
research groups use different software to analyze their data. He 
expects a consensus to emerge in
the next few years as standard software becomes widely available and 
research groups move
toward a universally accepted way of analyzing the data.
Dynamic contrast-enhanced M.R.I. is largely confined to research 
institutions conducting clinical
trials and should be considered experimental, said Dr. Peter L. 
Choyke. For the past three years,
Dr. Choyke has been working on refining the technique in 
collaboration with Dr. Knopp, at the
Ohio State University Comprehensive Cancer Center, and other researchers.
"Dynamic enhanced-contrast M.R.I. has the greatest potential - as yet 
unrealized - to monitor
therapy early on," Dr. Choyke said. He expects that when drugs stop 
new blood vessels from
forming, the M.R.I. will reveal changes in blood vessels that occur 
before the tumor responds to
the changes by shrinking or stabilizing. But Dr. Michael O'Reilly, 
who did pioneering research
with Dr. Judah Folkman at Children's Hospital in Boston said that 
even if such monitoring became
possible, considerable research would still be needed. Even after 
studies with the mice are
completed, Dr. O'Reilly predicted, it will be difficult to apply the 
results to people.
35
References
Bieber, M., Vitali, F., Ashman, H., Balasubramanian, V., & Oinaas-Kullonen, H.
(1997). Fourth generation hypermedia: Some missing links for the world
wide web. International Journal of Human-Computer Studies, 47, 31-65.
Bolter, J. D. (1991). Writing space: The computer, hypertext, and the 
history of
writing. Hillsdale, NJ: Earlbaum.
Bower, G. H., & Cohen, P. R. (1982). Emotional influences in memory and
thinking: Data and theory. In M. S. C. S. T. Fiske (Ed.), Affect and
cognition (pp. 291-331). Hillsdale, NJ: Erlbaum.
Chen, C., & Rada, R. (1996). Interacting with hypertext: A meta-analysis of
experimental studies. Human-Computer Interaction, 11, 125-156.
Darken, R. P., & Sibert, J. L. (1996). Navigating large virtual spaces.
International Journal of Human-Computer Interaction, 8, 49-71.
Dillon, A. (1996). Myths, misconceptions, and an alternative perspective on
information usage and the electronic medium. In J. J. L. J.F. Rouet, A.
Dillon & R. J. Spiro (Ed.), Hypertext and cognition (pp. 25-42). Mahwah,
NJ: Erlbaum.
Dillon, A., & Gabbard, R. (1998). Hypermedia as an education technology: A
review of the quantitative research literature on learner comprehension,
control, style. Review of Educational Research, 68, 322-349.
Dunwoody, S. (2001). Studying users of the why files. Science Communication,
22(3), 274-282.
Eveland, W. P., Cortese, J., Park, H., & Dunwoody, S. (2002). How web site
organization influences free recall, factual knowledge, and knowledge
structure. Paper presented at the National Communication Association,
New Orleans, LA.
Eveland, W. P., & Dunwoody, S. (2001). User control and structural isomorphism
or disorientation and cognitive load? Learning from web versus print.
Communication Research, 28, 48-78.
Eveland, W. P., Krisztina, M., & Seo, M. (2004). Moving beyond "just 
the facts":
The influence of online news on the content and structure of public affairs
knowledge. Communication Research, 31(1), 82-107.
Foust, J. C. (2005). Online journalism. Scottsdale, AZ: Halcomb Hathaway,
Publishers, Inc.
36
Fredin, E. S. (1997). Rethinking the news story for the internet: Hyperstory
prototypes and a model of the user. Columbia, SC: Association for
Education in Journalism and Mass Communication.
Gernsbacher, M., A. (1993). Less skilled readers have less efficient 
suppression
mechanisms. Psychological Science, 4, 294-298.
Gernsbacher, M., A. (1996). The structure building framework: What it 
is, what it
might also be and why. In B. K. B. A. C. Graesser (Ed.), Models of text
understanding (pp. 289-311). Hillsdale, NJ: Erlbaum.
Gernsbacher, M. A. (1990a). Language comprehension as structure building.
Hillsdale, N.J.: L. Erlbaum.
Gernsbacher, M. A., & Faust, M. E. (1991a). The role of suppression 
in sentence
comprehension. In G. B. Simpson (Ed.), Comprehending word and
sentence (pp. 97-128). Amsterdam: North Holland.
Gernsbacher, M. A., & Faust, M. E. (1991b). The mechanism of suppression: A
component of general comprehension skill. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 17, 245-262.
Gernsbacher, M. A., Varner, K. R., & Faust, M. (1990b). Investigating 
differences
in general comprehension skill. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 16, 430-445.
Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension.
Psychological Review, 85, 363-394.
Knobloch, S., Patzig, G., Mende, A.-M., & Hastall, M. (2004). Affective news:
Effects of discourse structure in narratives on suspense, curiosity, and
enjoyment while reading news and novels. Communication Research,
31(3), 259-287.
Lang, A. (2000). The limited capacity model of mediated message processing.
Journal of Communication, 50, 46-70.
Lang, A., Newhagen, J., & Reeves, B. (1996). Negative video as structure:
Emotion, attention, capacity and memory. Journal of Broadcasting &
Electronic Media, 460-477.
Long, L. (1995). Scientific explanation in u.S. Newspaper science 
stories. Public
Understanding of Science, 4, 119-130.
Long, M., Steinke, J., Kalter, J., Schwaab, A., Savagian, A., Jaffer, 
N., et al.
(1991). Explanation in newspaper science section news articles. Paper
presented at the Association for Education in Journalism and Mass
Communication, Boston, MA.
37
Long, M. A. (1991). An analysis of the organizational structure and 
explanatory
content of newspaper science stories (science writing). Unpublished
doctoral dissertation, University of Wisconsin, Madison, WI.
Mayer, R. E. (1985). How to analyze science prose. In B. K. B. 
Britton, John B.
(Ed.), Understanding expository text: A theoretical and practical
handbook for analyzing explanatory text (pp. 305-313). Hillsdale, NJ:
Erlbaum.
McDonald, S., & Stevenson, R. J. (1996). Disorientation in hypertext: 
The effects
of three text structures on navigation performance. Applied Ergonomics,
27, 61-68.
Miller, J. D. (1987). The scientifically illiterate. American 
Demographics, 9(6),
26-31.
Miller, J. D. (1995). Scientific literacy for effective citizenship. 
In R. E. Yager
(Ed.), Science/technology/society as reform in science education (pp. 185-
204). New York: State University of New York Press.
Miller, J. D. (2004). Public understanding of, and attitudes toward, 
scientific
research: What we know and what we need to know. Public
Understanding of Science, 13, 273-294.
Roller, C. M. (1990). The interaction between knowledge and structure 
variables
in the processing of expository prose. Reading Research Quarterly, 25,
78-89.
Shirk, H. N. (1992). Cognitive architecture in hypermedia instruction. In E.
Barrett (Ed.), Sociomedia: Multimedia, hypermedia, and the social
construction of knowledge (pp. 79-93). Cambridge, MA: MIT Press.
Stovall, J. G. (2002). Writing for the mass media. Needham Heights, MA: Allyn
& Bacon.
Tremayne, M. (2004). The web of context: Applying network theory to the use of
hyperlinks in journalism on the web. Journalism & Mass Communication
Quarterly, 81(2), 237-253.
Tremayne, M., & Dunwoody, S. (2001). Interactivity, information processing and
learning on the world wide web. Science Communication, 23(2), 111-132.
van Dijk, T. A. (1983). Discourse analysis: Its development and 
application to the
structure of news. Journal of Communication, 33(2), 20-43.
van Dijk, T. A. (1985). Structures of news in the press. In T. A. v. 
Dijk (Ed.),
Discourse and communication: New approaches to the analysis of mass
media discourse and communication (Vol. 10, pp. 69-93). New York:
Walter de Gruter.
38
van Dijk, T. A. (1988). News as discourse. Hillsdale, NJ: Erlbaum.
Young, J. D. (1996). The effect of self-regulated learning strategies on
performance in learner controlled computer-based instruction. Educational
Technology Research & Development, 44, 17-27.