BREAKING UP NEWS--AN INVESTMENT IN THE ONLINE NEWSPAPER'S
FUTURE? EFFECTS OF LINEAR AND NONLINEAR HYPERTEXT FORMATS ON
USERS' RECALL, READING, SATISFACTION, AND PERCEIVED STORY
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ARTS IN MASS COMMUNICATION
UNIVERSITY OF FLORIDA
I have many people I would like to thank for their support in the writing of this
thesis. The greater part of this document was made possible by the instruction and the
inspiration of my teachers at the University of Florida and by the love and support of my
family. My most sincere acknowledgments go to:
My supervisor, Dr. Kim Walsh-Childers, for her remarkable wisdom, guidance,
and strength. I am indebted to her subtle but strong ways of encouraging me.
My professor and mentor, Professor Melinda McAdams, for sharing her
undeniable expertise in online journalism and her books with me. I am grateful for her
patience and sensitivity. While I tossed around research ideas, she recognized my feelings
about hypertext and encouraged me to write the thesis I really wanted to write.
Dr. Leonard Tipton, for answering all the questions I asked and for making me
laugh when on the verge of taking things too seriously. It was during one of our
conversations in August 2000 that I first realized the experime nt would be manageable and
valuable. In various stages, he helped me see all the elements in perspective.
Dr. John Wright, Arlindo Albuquerque, and Olivia Jeffries, for ensuring me the
access I needed to the computer labs.
Dr. Michael Weigold, for supporting my research by offering his students
incentive to participate in the experiment.
Professor Dave Carlson and Lenny Uptagraft, for making the pre-test possible.
Mr. Jeff Roslow, online editor for The News-Press, for giving me permission to
use the article of my choice in this study.
My father, Philip Galfano, for his enthusiasm about this research in its earliest
stages. His response inspired me to--as he has been known to say--"start hoppin."'
My mother, Eileen, for always listening.
My husband, Serge, for working with me on my research schedule and for helping
me code data and see this document to completion. I am ever grateful for his patience, his
faith in my ideas, his acceptance of my goals as his own, and--always--for his love.
Thoughts of the sacrifices he made to be at my side while I wrote this thesis make the
finished product so much sweeter.
TABLE OF CONTENTS
A C K N O W L E D G M E N T S ......... ................. .................................................................iii
LIST OF TABLES ..................................... ................. .......... viii
A B STR A C T .................................................. ................... ................. . ix
1 IN TR O D U C T IO N ....................... .. ........................ .. ........ ..............
P u rp o se o f th is Stu dy ...................................................................... ....................... ....... 2
Why this Study Matters to Online Newspapers........... .................................... 3
Boundaries of this Study .................. ............................. ..... .... .. .. .... ...... 4
Research Objectives.............................. ............ 5
2 LITER A TU R E R EV IEW ................................................................ ....................... 7
H hypertext: A D definition ............................................... ................ ....................... 7
H history of H hypertext ............................................................. .......... 8
Hypertext and Nonlinearity.................. ................................ .. .......... ...... 11
How Hypertext Changes the Authoring Process........................................... 14
Hypertext and N arrative Structure .................................. .............................. ...... 16
News Design: Print-oriented vs.Web-oriented ..................................................... 19
Time for Change in Online Newspaper Publishing ........................................... 20
N onlinear Form ats for O nline N ew s .................................................... .............. 22
Reasons for Building Nonlinear News Stories with Hypertext .................................. 23
Anticipated Resistance to Nonlinear Hypertext Formats ........................................ 25
Understanding the Effects of Format.................................................. ....... ....... 26
Effect of Form at on Recall.................... .................................... ........... .............. 27
E effect of F orm at on Satisfaction ..................................................... ... ................. 28
Effect of Format on Amount Read.................................................................. 29
Effect of Format on Perceived Story Credibility .............................................. 32
Research Questions and Objectives....................... ............. .......... 34
Research Hypotheses ............................................... ............ .. 35
3 M E T H O D O L O G Y ........................................................................ ..........................39
P articip an ts ....................................................................................... 3 9
Stim ulus M material .................. ................................ .. ........ .. .......... .. 40
F a c ilitie s ..................................................................... 4 2
Instrum ents and M easures........................................................... ......................... 43
P re -te st ......................................... .. .................................... ........... 4 7
P ro ced u re s ............... ... ................................ .......................................... ..... 4 9
Scale R liability A analysis ............................................ ........................ ................ .. 51
Statistical A nalysis............................................ 51
4 R E S U L T S ..............................................................................5 2
Sum m ary of Findings ................................. .............. ..................... 52
A n aly sis....................................................................................... . .. 5 2
V ariables for A nalysis............................. .............. 53
Independent Variables ...................... ............ .................. 53
D dependent V ariables............................. .............. 53
Tests of Hypotheses .............. .......... ........... .................... 54
Results Regarding the Research Questions ........................ ............. 57
5 C O N C L U SIO N S ...................... ............... .................................................59
Results Discussion ........._.... .... .. .. ....................... ...... 59
Effect of Form at on Accuracy ............. ................... ........... .. .......... .......... 60
Effect of Form at on A m ount R ead...................................... ............... ................... 61
Hypertext Comfort and User Satisfaction, Perceived Story Credibility ............ 63
Hypertext Comfort and Recall, Accuracy, Amount Read ....................................... 65
Perceived Story Credibility and User Satisfaction.......................................... .. 66
Effect of Format on User Satisfaction.............................. ......................... 68
Effect of Form at on R ecall...................... ...................... .... ................ ........ ............. 69
Conclusions .............. ................ .........7...... ....................... .. 70
Lim stations of the R research ......................................................... .............. 72
Suggestions for Future Research ............................... ............... ............... 73
A TREATMENT SCREEN EXAMPLES ............................................. ...............76
B DESIGN OF THE TREATMENTS ............................ ....... ................. 79
N onlinear T reatm ent ......................................................................................... 79
L in ear T reatm ent ......................................................................................... 8 1
C NONLINEAR TREATMENT STRUCTURE .........................................................82
D Q U E ST IO N N A IR E 1 ...................................................................... .....................83
E Q U E ST IO N N A IR E 2 ......................................................................... ...................85
F IN STR U C T IO N SH E E T ........................................ .............................................90
G DEBRIEFING NOTE ............................................................................ 92
H C O D E B O O K .................................................................................... ................ .. 93
I CHECK LIST FOR CODING THE VARIABLE "ACCURACY" ........................98
L IST O F R E FE R E N C E S ........................................................................ .................... 102
BIOGRAPHICAL SKETCH ............................................... ...............108
LIST OF TABLES
4-1: Means and standard deviations for hypertext comfort, user satisfaction, and
perceived story credibility........................................................ ............. 54
4-2: Effect of story format on time spent reading ........................................................55
4-3: H ypertext com fort correlations.......................................................................... ...... 56
4-4: Correlation between credibility and satisfaction .............................. ................56
4-5: Effect of story form at on recall............................................. .............................. 57
Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Arts
BREAKING UP NEWS--AN INVESTMENT IN THE ONLINE NEWSPAPER'S
FUTURE? EFFECTS OF LINEAR AND NONLINEAR HYPERTEXT FORMATS ON
USERS' RECALL, READING, SATISFACTION, AND PERCEIVED STORY
Chairperson: Dr. Kim Walsh-Childers
Major Department: Journalism and Communications
The call of the journalism trade press for experimentation with online news story
format remains largely ignored by most online newspapers, which continue to publish news
in linear narrative formats that mirror the layouts of printed media. In this context, this
study investigated the effects of story format on users' recall, accuracy, satisfaction,
perceived story credibility, and how much of a story they read. One-hundred thirty five
participants used one of two versions (linear electronic text vs. nonlinear hypertext) of a
newspaper feature story assumed to be appropriate for adaptation from print to online
media. While story format was not found to have significant effects on the aforementioned
variables, the nonlinear format was found to have no worse effects than the linear format.
Significant correlations were found between three variable pairs: comfort with hypertext
and user satisfaction, comfort with hypertext and perceived story credibility, and user
satisfaction and perceived story credibility. In addition, findings did not show a significant
difference in recall scores between the two groups: This suggests that contrary to the
findings of research published prior to 1990, users may have overcome any extra cognitive
burdens that hypertext may place on readers. Reasons why experimentation with nonlinear
story formats by newspaper professionals might be an investment in the online newspaper's
future are discussed in the conclusion.
Hypertext--information presented as a linked network of brief self-sufficient texts
that computer users may navigate in a nonlinear fashion (Keep et al., 1995)--offers
journalists a way to tell stories that take advantage of the unique characteristics of online
media. Hypertext makes it possible for users to navigate stories by association, choosing
reading paths most interesting to them or most relevant to their information needs. As more
online newspaper users come to take these characteristics of online media for granted, they
may question why online news stories are essentially electronic copies of linear articles
from the printed newspaper.
Only recently have communication scholars (Huesca et al., 1999; Mensing et al.,
1998; Vargo et al., 2000) started to explore the viability of hypertext as a mass medium for
online newspaper journalism. Hypertext enables online journalists to segment stories into
readable chunks and link the chunks in any number of ways to tell stories. Some scholars
and researchers (Berry, 1999; Deuze, 1998; Li, 1998; Tremayne, 1999, 2000) have given
thoughtful consideration to hypermedia--the presentation of information as a linked
network of image, sound, and text--for journalism. However, the challenges that hypertext
(employing no images or sound clips) poses to current news narratives remain, with few
exceptions, largely ignored.
Purpose of this Study
A considerable amount of research in which participants have been observed using
Web sites (Morkes & Nielsen, 1997; Spool, 1999b; Stanford & Poynter, 2000) has found
that users focus on text over graphics. Based on a study that tracked the eye movements of
online newspaper readers, Stanford University and The Poynter Institute (2000) concluded
that a news provider's first chance to engage users is through text. This suggests that it is
crucial to consider questions about the effectiveness of hypertext to deliver news in text
form: Does presenting newswriting in a nonlinear hypertext format enable users to recall
news content more accurately than when news is presented in a linear electronic format?
Do users find nonlinear news story formats satisfying? How much of hypertexts do users
read? Do users find nonlinear news stories credible?
This experimental study aims to find provisional answers to these questions by
investigating the effects of story format on information recall and accuracy, user
satisfaction, amount read, and perceived story credibility. The experiment tests the effects
of two World Wide Web-based story formats. One format is linear, and one is nonlinear.
The linear format presents a complex news story that is essentially an electronic
copy of the story that was printed in the newspaper. When experienced on a computer, the
story is read (conventionally) by scrolling down the screen from beginning to end. This
format often is called "shovelware"--print material "repurposed" for the Web. This linear
format is still the norm on most online news sites (Rich, 1997; Tremayne, 1999) that do
not produce content exclusively for the online environment (Deuze, 2001).
The nonlinear format displays the same complex news story but employs hypertext
links that must be clicked in order to read different parts of the article. This format contains
no additional material; only transitional phrases are changed slightly. Thus, the original
story text is rearranged into a non-sequential narrative that allows readers the freedom to
navigate the story by association.
Why this Study Matters to Online Newspapers
The status of the Internet as a primary news source for Americans has grown
considerably since 1998. In the spring of 2000, one-in-three Americans (33 percent) said
that they regularly get news online, up from 20 percent in 1998. Nearly one-in-five (18
percent) Americans who get news online at least once a week said that they now use other
sources less often, up from 11 percent in 1998. Those who now use other sources less often
typically said that the Internet is replacing newspapers or television in their lives (Pew
Research Center, 2000).
Since newspapers went online via the Web in the early 1990s, many online
newspapers simply have provided electronic copies of printed news. Perhaps burdened by
efforts to add multimedia and interactive services to their Web sites, online news staffs
have not experimented actively with presenting text in innovative, nonlinear formats. A
content analysis on emerging trends in the use of nonlinear storytelling from 1997 to 1999
suggests that "if there is a general trend toward the greater use of nonlinear storytelling, it
is most apparent at the broadcast company Web sites" (Tremayne, 2000, p. 19). If nonlinear
formats prove more satisfying to online news consumers in the long run, online newspaper
professionals might regret not having experimented more with nonlinear formats--
especially if their readers have turned instead to broadcast news Web sites. User focus on
text might be advantageous for online newspapers: Notwithstanding the importance of
photojournalism to newspaper journalism, the bulk of newspaper content is text.
Creating effective nonlinear hypertexts is one way online newspapers could
distinguish themselves in terms of the user experience. When users are offered links to the
parts of stories they want to read, they do not have to wade through information in the
writer's preferred order. Users may feel satisfied because their time has been saved and/or
because they avoided a kind of information overload brought on by editorial choices that
would not match their own. Presenting news effectively in nonlinear hypertexts is also one
way online newspapers could set themselves apart in terms of perceived credibility. At a
time when "anyone can be a journalist online" and the distinction between reliable and
"tabloid" information is blurred (Deuze, 2001, p. 6), online newspapers may be able to use
nonlinear hypertext formats to enhance user perceptions of their product's credibility and
their organization's credibility as a news source.
In addition, publishing in nonlinear formats might help online newspapers prepare
for the emerging publishing landscape. Outing (2000) has observed:
News organizations in the new millennium will be publishing to multiple
media: print-delivered, home-printed, the Web, e-mail, PDAs [personal
digital assistants or hand-held computers], mobile phones, e-readers (or e-
book readers), pagers, Internet radio, and broadcast radio and TV. A
newspaper company of the near future will likely distribute its content to
all of those except broadcast radio and TV. ( 5)
If newspaper journalists start writing for nonlinear hypertexts--paying particular attention
to the many links or connections they can draw within their own writing, they will develop
the kind of texts that are more easily distributed to a variety of electronic devices, more
usable on those devices, and possibly more satisfying to online news consumers.
Boundaries of this Study
It is outside the scope of this thesis to be concerned with the audience-authoring
aspect of hypertext discussed by Li (1998) and described at length in important works on
hypertext (see Bolter, 1991; Landow, 1992; Nielsen, 1990, 1995). Rather, this thesis is
concerned with how online newspaper users participate in producing individualized
narratives when hypertext provides them the opportunity to read a story in an order not
completely predetermined by the journalist or news editor.
Specifically, researchers (Li, 1998; Schultz, 1999) have noted that hypertext links
can incorporate audience participation in producing newspaper content by allowing users to
add content to the online newspaper via message boards and discussion groups. It can be
recognized that hypertext links bring in audience participation in producing news content
another way: As Fredin (1997) has noted, the range of choices available in a hypertext
news story means that "each user creates a unique story through his or her sequence of
choices" (p. 4). In this respect, users participate in producing the content they consume.
So far, only a handful of researchers (Huesca et al., 1999; Mensing et al., 1998;
Vargo et al., 2000) have tested the effectiveness of hypertext for journalism. This
experimental study seeks to contribute to this small but growing body of literature. By
testing the effects of story format on recall and accuracy, user satisfaction, the amount of a
story users read, and perceived story credibility, this study aims to offer a basis for judging
how worthwhile it might be for journalists to try out different story formats.
This study is organized as follows: Chapter 2 summarizes the history and features
of hypertext as a medium for nonlinear storytelling. The chapter also reviews literature that
proposes nonlinear formats for online news. In addition, it discusses reasons why online
newspapers might experiment with new formats and why they might not. Finally, the
chapter reviews previous studies that have investigated the effects of format on information
recall, accuracy, user satisfaction, amount read, and perceived credibility.
Chapter 3 explains the experimental design and treatments used in this study.
Chapter 4 reports the findings of this research. Chapter 5 includes discussion of the
findings, the conclusions of this study, and suggestions for future research.
Hypertext: A Definition
The role of technology in effectively distributing media messages deserves serious
attention in any scholarly venture on the topic of online journalism (Deuze, 1998). The
literature on online journalism suggests that hypertextuality is an essential characteristic of
online journalism (Deuze, 2001). Yet few studies address the role hypertext can play in
delivering news online effectively.
For all the theories about its meaning and use, hypertext is a simple concept: It is a
direct connection from one position in a text to another position in a text (Aarseth, 1994).
Hypertext consists of individual blocks of text and the electronic links that join them
(Landow, 1994). It creates associations called links between the text blocks or chunks of
information called nodes. This information-representation system provides a nonlinear
semantic network with multiple paths through various texts; hence, it offers multiple
experiences of information (Guay, 1995).
Often hypertext is combined with multimedia to form hypermedia. Hypermedia can
be seen as an extension of hypertext. The main difference in hypermedia is that nodes
include multimedia content, such as photographs, graphics, audio clips, and video clips,
usually in addition to text. Because of the similarity of hypertext and hypermedia, studies
on hypermedia often raise--as well as offer insight into--questions regarding hypertext. This
study benefits from the insights provided by some research on hypermedia, but its focus is
History of Hypertext
Vannevar Bush, a mathematician, engineer, and former director of the U.S. Office
of Scientific Research and Development, was one of the first researchers to conceive of an
automated nonlinear text system similar to current implementations of hypertext. In his
well-known 1945 magazine article "As We May Think,"1 he noted a "growing mountain"
of research in the biological, physical, and psychological sciences (Bush, 1945, 6). The
problem was not so much excessive publication, given the extent and variety of scientific
interests; rather, it was that publication was "extended far beyond our present ability to
make real use of the record" (Bush, 1945, 8). He envisioned using electro-mechanical
technology as a reading and writing system in which records could be continuously
extended, stored, and consulted.
Bush proposed what he called a "memex": a mechanized device that would serve as
an interactive library and an extension of a person's memory. The memex user would be
able to build an associative trail between ideas in different texts, stored on microfilm. The
user would view two texts on adjacent screens and, with the tap of a single key on the
keyboard, forge a permanent link between two passages visible on the screens. Thereafter,
any time a passage linked to another was in view, the user could instantly call into view the
linked passage. The trails would be stored and would remain available for display as well
as for revision by the memex user, his or her peers, and future generations.
1 The original article was printed in the July 1945 issue of The Atlantic Monthly and is
included in the archives of The Atlantic Online .
The essential feature of Bush's memex was "associative indexing" ( 61). He
suggested that associative indexing essentially would enable the user to cause at will any
item to call up another. Because Bush assumed the human mind works by association, he
believed the machine would accommodate human thinking. Bush never built the memex
because the available technology was not capable of supporting its features. It was not until
the 1960s that researchers Theodor H. Nelson, who coined the term hypertext, and Douglas
C. Engelbart, who invented word processing, the mouse, and the text link, began to design
and create computer systems that implemented some of Bush's notions of linked texts
(Landow & Delany, 1991; Nelson, 1990).
Nelson (1990), in Literary Machines (first published in 1980), describes his idea for
Project Xanadu--a true hypertext system that would serve as a universal electronic library
and publishing system. As Bush pictured his memex, Nelson envisioned his system as one
that could better organize materials "in ways that reflect their true structure" (p. 0/8).2
Project Xanadu "is very close to Bush's memex (now computerized); and its purpose is the
augmentation of human intellect, as Doug Englebart foresaw" (p. 1/5).
Nelson defines hypertext simply as "non-sequential writing" (p. 1/17). In two
significant elaborations, he describes hypertext as "text that branches and allows choices to
the reader" (p.0/2) and as "non-sequential forms of writing connected by links" (p. 1/26).
As he notes, hypertext often contains sequential text, but the structure of documents written
with hypertext is non-sequential (see Figure 2-1).
2 Nelson used an unconventional system for numbering pages in Literary Machines. The
book begins with Chapter Zero. The page numbers represent the chapter number (the
numeral on the left of the forward slash) and the page number within that chapter (the
numeral on the right side of the forward slash). For example, the quotation cited here is
printed on page 8 of Chapter Zero. Each chapter starts on page 1.
Figure 2-1. A simplified view of a hypertext structure with seven nodes and ten links.
The parallel lines in each node represent linear text. The ovals on the origin points of the
arrows represent linked words and phrases.
When writing with hypertext, the author faces the challenge of making readers feel
comfortable and oriented (Nelson, 1990).3 Yet, two difficulties of writing sequential text
disappear: The author no longer has to decide on sequence, only on "interconnective
structure, which provides much greater flexibility" (p. 1/18); and the author no longer has
to decide "what is in or out, but simply where to put things in the searchable maze" (p.
1/18). In Nelson's view, hypertext improves the representation of thought since it can
embody all the interconnections an author, or many authors, can think of.
3 For a thorough discussion of the problem of user disorientation, see Foss, C. (1989).
Tools for reading and browsing hypertext. Information Processing & Management, 25,
Jakob Nielsen's (1990) definition of hypertext complements Nelson's. Nielsen
suggests that contrasting hypertext with a traditional textbook is the simplest way to define
All traditional text, whether in printed form or in computer files, is
sequential, meaning there is a single linear sequence defining the order in
which the text is to be read. First you read page one. Then you read page
two. ... And you don't have to be much of a mathematician to generalize
the formula which determines what page to read next. (p. 1)
In contrast, hypertext is nonsequential, meaning "there is no single order that
determines the sequence in which the text is to be read" (p. 1). Thus, hypertext presents
different options to readers, and the individual reader determines which of them to follow
when he or she reads the text. The hypertext author's task is setting up "a number of
alternatives for readers to explore rather than a single stream of information" (p. 2). It
should be noted that the author of a traditional printed text also sets up alternatives for
readers when he or she includes footnotes or cross-references (Nielsen, 1990). What
differentiates hypertext from structures such as footnotes and cross-references in printed
texts is the automation and the immediacy of calling linked texts into view. The hypertext
links that offer this immediacy also allow the nonlinear or multilinear organization of text
into formats that are convenient to explore.
Hypertext and Nonlinearity
The nonlinear order of electronic writing contrasts sharply with the linear order of
printed texts. The simplest way to understand the differences between them is to contrast
electronic texts with traditional texts. For example, a typical newspaper article is a linear
path of text, offering one pathway defined by the journalist. Even if the article has more
than one writer, a single authorial voice still paves the path. And although readers may read
the article by skipping along the path in a different order than the one paved by the
journalist or journalists, readers still must move along the established path in order to locate
the information they find interesting. The path remains fixed.
Most often, newspaper articles appear in the "inverted pyramid" narrative format
(Fitzgerald, 1996) commonly taught in journalism schools. The news narrative begins with
the conclusion and gradually offers detail and background information. Readers who enter
at the fixed entry point (the top) can stop at any point and still come away from the article
with the most important information, as defined by the journalist who fitted the details of
the event into the inverted pyramid structure. No matter where readers enter the narrative,
they cannot affect the fixity of the printed form.
A typical newspaper article also is isolated from other articles, even if the reported
events are significantly connected. Modern newspaper layout, which attempts to connect
articles via design elements including boxes, rules, shading and "packaging," does not alter
the isolated state of the articles: They remain isolated because different headlines label
them as separate articles. Newspaper layout offers readers multiple paths through the
collection of articles. The news editor or editors, who decide the placement and order of
news elements, define these paths. Layout conventions, such as the use of dominant
imagery, headlines in a hierarchy of sizes, and boxes around stories, suggest to the reader a
hierarchy of importance of events, or at least a "preferred" way to make connections among
events. Seeing newspaper layout as a non-sequential or mosaic informational format,
though it may be appreciated that way (Murray, 1997), does not alter the fixedness of
layout when it exists in printed form.
In contrast, hypertext is comparable to "a printed book that the author has attacked
with scissors and cut into convenient verbal sizes" (Bolter, 1991, p. 24). Screen-based
pages are not pages bound in a single sequence. Rather, they are "blocks of text"--termed
lexias by Roland Barthes (Landow, 1992, p. 4)--occupying a virtual space in which they
can be preceded by, followed by, and placed next to an infinite number of other lexias
(Murray, 1997). Hypertext essentially organizes lexias into nodes or modules in a network
connected by hyperlinks (Conklin, 1987; Guay, 1995; Keep et al., 1995; Landow, 1992;
Murray, 1997; Nelson, 1990; Nielsen, 1990, 1995).
Hyperlinks allow journalists to go beyond simply uploading linear articles onto
single-screen-based pages. Long, scrolling screen-based pages that offer articles in a format
suitable for printing ironically impose the constraints of linearity on readers in a medium
that is fundamentally user-driven. Nielsen (2000) says that hypertext should not be used to
present long, detailed articles in linear fashion, because text can be made short without
sacrificing depth of content by chunking information into hypertexts; furthermore,
hypertext should not be used to segment linear stories into multiple pages with a fixed
order imposed, as this would deny users the choice of reading only the pages that interest
them. Rather, hypertext can be used to relegate long and detailed background information
to secondary pages (Nielsen 2000). Hence, the journalist can make any information of
interest to a subset of readers available through a link without forcing such information on
readers who are not interested. This ability to relegate background or other information to
subsidiary pages changes the authoring process.
How Hypertext Changes the Authoring Process
The authoring process changes for the newspaper journalist who writes now with
hypertext.4 With hypertext, the authoring process is no longer only a word-and-sentence-
level activity; authoring now is the design of a document (Conklin, 1987).5 In Murray's
(1997) terms, authorship in electronic media is "procedural":
Procedural authorship means writing the rules by which the texts appear as
well as the texts themselves. It means writing the rules for the interactor's
involvement, that is, the conditions under which things will happen in
response to the participant's actions. It means establishing the properties
of the objects and potential objects in the virtual world and the formulas
for how they will relate to one another. The procedural author creates not
just a set of scenes but a world of narrative possibilities. (pp. 152-153)
In electronic narrative, Murray says, "the procedural author is like a choreographer who
supplies the rhythms, the context, and the set of steps that will be performed" (p. 153).
While Murray's definition of procedural authorship readily applies to the process of
authoring virtual worlds, such as the visual landscapes of video arcade games or earlier
text-based virtual reality games such as Zork, her definition also applies to the electronic
journalist's writing process. Though the order of steps in hypertext news design is open to
debate, the process of authoring a hypertext includes writing multiple story components
and linking them together in a structure that offers users multiple ways to access all
components. The journalist decides how story components should relate to one another and
4 See McAdams and Berger (2001) for an in-depth description of the processes involved
in authoring hypertexts.
5 It should be noted that Conklin (1987) drastically oversimplifies the authoring process
by overlooking the author's process of ordering information. However, his point that
authoring with hypertext involves document design is critical in a discussion of how
hypertext changes the authoring process.
Contrary to postmodernist notions that the author is dead (Barthes, 1968), as well as
any fear that journalists may lose all control of their reporting, the author is--as Bolter
(1991) has suggested--very much alive and in control. It just so happens that journalists
who write with hypertext now share this control with readers (called users) who actively
click to read their way through stories. Through careful link creation, the journalist shapes
the hypertext into a structure that preserves user agency--"the satisfying power to take
meaningful action and see the results of [one's] decisions and choices" (Murray, 1997, p.
126). In more concrete terms, agency can be defined as the power of users to choose their
own paths through a story and be rewarded by an appropriate and satisfying response.
Hypertext stories, in contrast to print articles, generally offer users more than one
possible entry point, many internal paths, and no distinct ending (Murray, 1997). This
changes traditional journalism in two ways. First, the journalist is no longer making all of
the decisions about text flow. Users decide which links to pursue and may exercise the
freedom to interact directly with chunks and establish new relationships between them
(Conklin, 1987). Second, user control of story order shifts the space where journalistic
standards will be manifest. Traditionally, the singular, linear narrative is judged in terms of
its fairness, balance, objectivity, accuracy, and completeness. But Fredin (1997) argues that
the same standards now will appear in "labeled links between fil/e" (p. 22). For example, a
story about a police shooting with a marked "entry page" that links to only one of five eye-
witness accounts that exist in the hypertext is comparable to quoting that same one
eyewitness in the second paragraph of a linear story, just after the lead. Both of these
choices may be perceived as a bias toward a particular frame for that story. Fredin's
argument that traditional journalistic standards will be manifest in links makes way for
hypertext to break down the journalistic narrative as we know it.
Hypertext and Narrative Structure
Koch (1991) and Manoff (1986) have argued that the narrative forms used by
journalists shape their coverage of the events they report. In his book chapter "Writing the
News (By Telling the 'Story')," Manoff suggests that the choice of narrative structure
is among the most important journalistic decisions ... since it determines
the shape of the event to be judged and thereby often the judgment that is
to be rendered. In fact, journalism can be seen as an activity that judges
events while it reports them by juxtaposing, amalgamating, or separating
facts, events, and opinion in order to find in them their "story." (p. 218)
Citing reports from major U.S. daily newspapers including The New York Times and The
Washington Post, Manoff points to two deceptive qualities of modern newswriting. First,
modern newswriting operates according to the ideal of objectivity with an emphasis on
facts. But it does so within the requirement that it represent events as stories--stories that
seek to convey that, for each event, there is one story and one correct way to tell it (Manoff,
1986). This quality is deceptive from a postmodern stance, which insists that there are
multiple ways to tell a story, all of which may be "correct." Second, Manoff says that
modern newswriting conforms to the genre of ironic narrative, which "seeks to establish a
bond between writer and reader in order to persuade the latter that he has discovered
judgments for himself" (p. 28). The singular authoritative voice, which forges this bond
while presenting events in a fixed order, no matter whether they occurred in that order, is
also a deceptive quality of modern newswriting.
The linear, fixed order of modem newswriting conflicts with the variable, nonlinear
order of hypertext. As Fitzgerald (1996) suggests, hypertext does not change the essentials
of j ournalism--"gather and report the news accurately, fairly and responsibly, giving
readers ... balance, perspective and specific information" (p. 72). But it does alter the basic
format of newspaper journalism: the inverted pyramid. In light of current practice in online
newspaper publishing, namely the wholesale repurposing of print for the Web, Fitzgerald
raised the question of whether the inverted pyramid would remain the dominant model.
One answer to his question is that journalists will keep writing inverted pyramids--
just more of them, and the texts will be shorter than current versions. Online journalism
practitioners who embrace this view believe that each story element will have its own
inverted pyramid, and journalists will link the elements according to the connections they
feel are appropriate to the stories they tell (Fitzgerald, 1996). Web usability research
(Morkes & Nielsen, 1997) suggests that this way of writing may be preferable for online
news delivery. Nielsen (2000) says that hypertext pages written in inverted pyramid style
are optimized for usability. Due to questions about users' motivation to scroll,6 starting
with a short conclusion containing the most important information "gives users the gist of a
page even if they do not read all of it" (p. 112).
Another answer to Fitzgerald's question is that the inverted pyramid will survive in
a different context. In 1996, Fitzgerald interviewed the multimedia services editor for the
Associated Press, Ruth Gersh, who suggested that the form might take the shape of an
inside-out matryoska, the Russian nesting dolls. Starting with "the little doll," journalists
would use the structure to present increasingly larger segments of a story. Another
6 In an online column, Nielsen (1996) writes: "On the Web, the inverted pyramid
becomes even more important since we know from several user studies that users don't
scroll, so they will very frequently be left to read only the top part of an article. Very
interested readers will scroll, and these few motivated souls will reach the foundation of
the pyramid and get the full story in all its gory detail" ( 4). Ziff-Davis AnchorDesk
is a good example of an "inverted-pyramid site"
(Nielsen, 1996, 8; 2000).
possibility is that online newspapers would adopt a form of what Kevin McKenna, editorial
director of New York Times Electronic Media in 1996, called "serial storytelling"
(Fitzgerald, 1996, p. 72). In this form, journalists would tell the story in 400- to 500-word
installments, and readers could read as deep as they wanted into the sequence (Fitzgerald,
1996). Both of these suggestions are innovative, compared to presenting stories written for
print in a long scrolling file. However, conceptually, they resemble expandable versions of
a fixed, linear inverted pyramid structure. Both formats offer users little choice in terms of
story-reading order, making the formats less appropriate for online news delivery--
especially if they require users to click through the text that represents "inside dolls" to
read the perspective they seek, or to click "deep" into a sequence to find it. Users likely
would be frustrated for at least two reasons: (1) their perspective is buried, and (2) such
"forced" extra clicks waste their time.
Despite some discussion of innovative concepts in the trade press (Dube, 2000;
Fitzgerald, 1996; Johnson, 2001; Meyer, 1996; Scanlan, 2000; South, 1999), online
newspapers have not experimented extensively with new narrative designs that hypertext
makes possible. Harper (1996) found that online newspaper content had offered little more
than an electronic version of the print product. Gubman and Greer (1997) found that most
online newspapers had not adapted well to the digital environment in terms of news
presentation style. Their content analysis of a sample of 83 U.S. online newspapers found
that only 13 (15.7 percent) of those newspapers were using any kind of nontraditional
storytelling or linked text blocks. Martin's (1998) research serves as a check on Gubman
and Greer's earlier result. Her observations of practices at two major metro U.S. daily
newspapers showed that online staff members rarely questioned or changed print copy in
substantive ways before publishing news articles on the Internet.
News Design: Print-oriented vs.Web-oriented
Today, a typical online newspaper contains long-form text articles (Palser, 2000).
Traditional linear format is the norm (Rich, 1997). Mass communication scholars and trade
press writers alike have noted the influence of the print tradition on current Web design
principles (Lowrey, 1999; Martin, 1998; Palser, 2000; Rich, 1997); the general sense is that
online newspaper editors have hesitated to let go of print design principles. But, as these
authors have noted, certain print design principles might not translate to the Web. For
example, Web designers must present news with users' technical limitations in mind--
which means designing for small screens and keeping scrolling text to a minimum
(Lowrey, 1999; Morkes & Nielsen, 1997). "It seems clear ... that in some ways, Internet
news design must transcend the modernist design-as-map metaphor" (Lowrey, 1999, p. 23).
Ingrained in the design-as-map metaphor is the idea that designers package and label
related stories "to ensure readers make proper connections between issues, as
predetermined by editors" (p. 14). As suggested earlier, the editorial hierarchy provided by
newspaper page layout conventions may disappear online, where users are equipped with
multiple windows for reading text and thus empowered to make connections between
issues for themselves.
In contrast to studies that have suggested that print design principles might not
translate to the Web, research on Web usability (Morkes & Nielsen, 1997) and a study
conducted by Stanford University and The Poynter Institute (2000) suggest that print
design principles plainly do not translate to the Web. As Nielsen (1999) has noted, the Web
designer's smaller canvas leaves less room for staples of print design such as dominant
photos, display art, and fancy typography. On the Web, these design elements can require
long download times, which might turn users away. Longer texts, another remnant of print
(as opposed to shorter texts), are also "unpleasant to read online" (Nielsen, 1999).
Furthermore, Spool (1999a, 1999b) found that graphic design is not as important as people
think: Time required to generate attractive graphics that download quickly does not pay off
in terms of increased usability. For these reasons, the Web is more amenable to shorter, no-
frills text presentations. In fact, some of the most popular Web sites are "very text
intensive" (Spool, 1999a).
The Stanford and Poynter (2000) study supports Nielsen's and Spool's ideas about
the importance of text (as opposed to graphics) online: Researchers tracking eye
movements and surfing behaviors of 67 online news users found that users entering a Web
site focus on text earlier and more often than photos or graphics. This finding is the
opposite of typical print behavior. While the Stanford and Poynter report on the study has
acknowledged critics of its methodological limitations--for example: small sample size and
failure to track peripheral vision (Palser, 2000), the study results fundamentally challenge
the notion that print design principles translate to the Web.
Time for Change in Online Newspaper Publishing
Current research suggests the time has come for online newspaper publishing to
challenge its print orientation. Martin's (1998) research speaks to the "tedious" and "time-
consuming" nature of online newspaper publishing as practiced in the newsrooms she
observed (p. 72). Her research documents a newsroom culture in which online editors
spend hours each day selecting stories from print editions and stripping them of pagination
markups for online presentation; rarely are these editors writing or producing original
content for online news consumers. Though the study is based on observations at only two
newsrooms in 1996, given the proliferation of long-form text articles noted by Palser
(2000), it is reasonable to assume that current routines resemble the practices documented
by Martin. It seems such routines would pose two major risks to the online counterparts: (1)
staff burnout and (2) failure (by the organization) to differentiate online articles from
printed presentations enough to make newspaper readers also want to visit the Web site for
a different news-reading experience.
Research suggests that the long-term viability of online news could depend on
whether online services offer news consumers a different experience than that provided by
print. Based on a survey of 489 Austin-area residents (random sample), Chyi and Lasorsa
(1999) found 76 percent of Web users surveyed said they would prefer print when asked to
imagine being provided with both print newspapers and online newspapers with the same
news content at the same price. The researchers have stated that this result "implies that
online newspapers will find it difficult to compete effectively with the print format in the
local market where both editions are available without differentiating their products" (p.
11). Ideally, beyond what Chyi and Lasorsa have said, online newspapers would compete
effectively with the print products not because they offered different content but because
the presentation of news content would meet the demands of the medium and users would
find the online experience of the content satisfying and credible.
Along with Chyi and Lasorsa, Henderson and Femback (1998) have noted that
online newspapers could differentiate their online products by developing interactive
services and multimedia presentations instead of presenting shovelware. Henderson and
Femback added that to offer substantial multimedia services in the near future would force
newspapers to build alliances with broadcasters. But building alliances with broadcasters in
order to offer multimedia might not be the answer online newspapers need right now. The
Internet is a medium that is driven by users who, as research suggests (Morkes & Nielsen,
1997; Spool, 1999b; Stanford & Poynter, 2000), focus on text. For online newspaper
editors, embracing text first goes hand-in-hand with taking advantage of user focus on text
in a user-driven medium. By constructing effective nonlinear formats for online news,
online newspapers would enable users to choose their own paths through stories--paths that
presumably would meet their needs and prove satisfying.
Nonlinear Formats for Online News
In "Rethinking the News Story for the Internet," Fredin (1997) proposed several
basic formats for news stories with internal choices--which he calls "hyperstories" (p. 2).
Fredin's hyperstory prototypes feature links that lead to other sections of the same article,
thus allowing users to construct their own stories through making choices. The key to a
good hyperstory is that it "[keeps] users in an active state of mind so that the choices they
make keep building the story effectively" (p. 2).
While Fredin's prototypes have been defined and described as hypermedia formats
(Fredin, 1997; Tremayne, 2000), they work as hypertext fornats too; after all, hypermedia
is "multimedia hypertext" (Nielsen, 1990, p. 5). A single hyperstory consists of the user
interface plus a network of files (Fredin, 1997). In other words, a hyperstory would not be
contained in a single hypertext file (or on a single Web page), as is often the case with a
printed story "repurposed" for the Web. The network of files would be constructed by the
In Fredin's prototypes, "screen layout consists of two windows side by side.... One
window, usually on the left, is the main story window; the other window is the related-file
or context window" (p. 9). Both windows contain scrollable text and are simultaneously
visible. While current online newspapers usually present stories in only one window,
Fredin has explained that having two windows is "highly advantageous"--one reason being
that "it can help orient the user by reminding the user of material recently seen, or by
serving as a sort of local home page in the story" (p. 9). The use of two windows might be
disadvantageous (at least for now) because current users are likely accustomed to reading
stories in a single browser window. However, Fredin's prototypes can be used to
conceptualize nonlinear story designs that can be tested out on users.
Reasons for Building Nonlinear News Stories with Hypertext
The literature suggests several reasons why online editors should begin to build
stories in the nonlinear formats that hypertext makes possible. The most urgent ones appear
related to user satisfaction. It has been suggested that in the recent race to publish on the
Web, newspaper professionals seem to have forgotten their readers: They have rushed to
incorporate everything from love-letter services to virtual museums on their Web sites,
instead of trying to find out "what really works" and "what readers really want in an
electronic newspaper" (Vargo et al., 2000, p. 40). Perhaps the development of nonlinear
formats that effectively satisfy users has suffered because of the race to incorporate such
Unlike the first users of hypertext, the next generation will take nonlinear formats
for granted (Murray, 1997). In addition, once most users of this generation see the
navigation-by-association approach as a fundamental reason to access information online,
they may begin to feel dissatisfied with online news in its current linear format, which
gives them little freedom to access instantly the parts of news stories that interest them.
Once users develop the expectation that online content can be tailored to offer easy access
to their interests and perspectives, they may be disappointed that they have to scroll
through linear stories derived from print media in search of the perspectives they are
seeking--perhaps only to be let down further if perspectives they sought are missing from
the story. A well-labeled hyperstory would make clear up front whether the story includes a
certain perspective, so users would not have to waste time looking for it if it is not included
in the story.
Additional reasons to embrace hypertext are related to the fate of online
newspapers, especially as that fate is related to both the role they play and the public's trust
in them. Starting to build hypertext stories now may help safeguard the position of online
newspapers as reliable gatekeepers in an era during which the universe of online
information expands quickly and indefinitely. As more and more massive databases (e.g.,
government databases of documents that journalists may link to as related files) become
widely available, the public's need for interpretations and explanations of the databases and
their contents will increase. Fredin (1997) suggested that the hyperstory may become a
continually developed interface for larger databases of story files and related files. Given
that a key part of interpreting and explaining data would include linking the databases to
other information about ongoing events in the world, hyperstories can greatly challenge and
expand the role of journalism in society" (p. 39). Online newspapers--with plenty of text at
the editors' disposal and with an audience of users found to focus on text--are uniquely
positioned to meet the challenges that hypertext poses to journalism as we know it.
Finally, there are philosophical grounds for building news stories with hypertext.
More than 20 years ago, Herbert Gans (1979) called for "multiperspectival journalism,"
which would change conventional story formats in a way that hypertext and the endless
digital news hole make possible:
When several perspectives must be taken into account on any given topic,
stories will naturally become longer. Moreover, journalists would be
required to organize these perspectives and in some cases relate and
interpret them; consequently, news analyses would be necessary more
often. When the news contains greater diversity of opinion from sources,
additional journalistic commentary may also be desirable, thus allowing
for personal and advocacy journalists in national news organizations. In
the process, the news would become more ideological, with explicit
ideological diversity replacing the near-uniformity that now prevails.
Furthermore, because hyperstories theoretically embrace the pluralism of opinions and
events, rather than offering singular accounts that purport to record the "truth" of events,
they are more compatible with postmodern perspectives. As the 20th century recedes, "we
no longer believe in a single reality.... Yet we retain the core human desire to fix reality on
one canvas, to express all of what we see in an integrated and shapely manner" (Murray,
1997, pp. 161-162). By combining the advantages of hypertext with newspaper journalists'
skill with text, online newspapers could be that canvas.
Anticipated Resistance to Nonlinear Hypertext Formats
No matter how well online newspapers are positioned to embrace and express
postmodern perspectives, the industry probably will oppose change for three reasons.
First of all, the dynamics of electronic writing marginalize what has been the central
quality of print for the last 500 years: "the fixed and monumental page of print ... that
exists in thousands of identical copies and ... resists change" (Bolter, 1991, p. 60). In
online newspapers, content is neither fixed nor monumental. News articles change with
constant updates, at least on some sites. What is more, with printed newspapers, it is much
more expensive and laborious for readers to compare the coverage of various news sources.
In an online environment, users can easily compare articles from different online
newspapers on the same computer screen. This makes it easier for readers to notice
discrepancies that can threaten the perceived integrity of content being compared.
A second reason the industry might stand firm against change is what may be a
profound new ethical obligation on journalists for more complete reporting. The
thoroughness of reporting possible is no longer limited by the size of the news hole.
Furthermore, when reporting with hypertext, journalists are no longer required to fit the
details they gather into a preconceived narrative structure. Hence, the reporting process is
transformed from one in which the journalist decides which details go "in" and which can
be left out to a process in which the journalist builds a structure into which all the details
can fit, linking pages in a way that demonstrates journalistic values of fairness, balance,
objectivity, accuracy, and completeness. This practice is labor-intensive, to say the least.
The process probably would be less time-efficient than conventional journalistic routines.
The third reason industry probably will resist the shift to hyperstories is focus on
profits. Huesca (2000) has argued that most journalists will never produce a hypertext
because of the media industry's fixation on making money: Even though the end product
could actually revitalize civic life and enhance understandings of social, cultural, and
political issues, the process would eat away at the bottom line (Huesca, 2000).
Understanding the Effects of Format
It is reasonable to assume that the industry will hesitate to spend resources on
hyperstories until the effects of nonlinear story format are better understood. This thesis
attempts to provide some understanding by exploring the effects of a linear story format
versus a nonlinear story format on recall, accuracy, user satisfaction, the amount of a story
users read, and perceived story credibility.
Effect of Format on Recall
Mensing, Greer, Gubman, and Louis (1998) tested whether information recall and
enjoyment is affected by online news presentation style. (This sub-section of this thesis is
concerned only with the recall findings.) Sixty-one participants read two news articles from
a simulated version of an online newspaper, with each participant receiving either a linear
treatment or a nonlinear treatment. The researchers found no differences in recall between
participants reading linear news articles that required scrolling down multiple screens and
participants reading the same articles in a nonlinear hypertext format. Though this result
runs contrary to the often cited study by Gordon (1988, cited in Charney, 1994; McKnight,
1996; and Nielsen, 1990), it points to the possibility that nonlinear narratives will benefit
users with the same recall power as linear narratives. This possibility is supported by
Wenger (1996), who found in two experiments that processing hypertext is not more
demanding overall than processing linear text.
A weakness of the Mensing et al. (1998) study is a potential problem with article
selection. As the researchers note, the two articles used in the experiment might have been
so interesting to the participants that the participants read them more carefully than they
would have under non-experimental conditions: The articles were selected "to have high
appeal for the subjects" and the researchers believed that the article topics would be
"relatively interesting and reader friendly to the majority of subjects." While it would have
been impossible for researchers to change the artificiality of experimental conditions, the
article-selection problem might have been mitigated by a more thorough justification of
article choice that also considered what kinds of content is best suited for online delivery.
One more potential problem noted by the researchers is that the manipulation of the
presentation style might have been too subtle. It might have been that the linear and
nonlinear treatments were so similar that differences could not easily be found. As the
researchers suggest, perhaps using longer, more complex articles would have made
differences between treatments easier to discern.
A potential problem ignored by the researchers is that because the participants were
journalism students, their understanding of journalistic narrative might have aided their
recall efforts. This problem might have been remedied by selecting a sample of participants
with various academic backgrounds.
Effect of Format on Satisfaction
A study by Kamerer and Wilcox (1993) has tested the effect of hypermedia on user
satisfaction. However, very little (if any) research has tested the effect of hypertext on user
satisfaction. Mueller and Kamerer (1995) examined the correlates of satisfaction with and
preferences for electronic newspapers. Findings showed that media use correlated
positively with satisfaction variables. Huesca et al. (1999) used a combined quantitative-
qualitative approach to explore user responses to linear and nonlinear hypertext news
stories. The results showed varying levels of satisfaction for each format. Still, it appears
that no research to date has tested, in a controlled experiment using only quantitative
methods, the effects of format on user satisfaction. This variable is worth testing because
Fredin (1997) and Murray (1997) have suggested that the nonlinear hypertext is an
emotional medium that, if constructed and used well, can provide a strong sense of
emotional satisfaction. What is unknown is whether it more satisfying to read nonlinear
hypertext than it is to read linear, scrolling text. As Mensing, Greer, Gubman, and Louis
(1998) note, if the newspaper industry is going to invest the time and resources to take
advantage of hypertext, the investment should pay off in terms of improved user
Effect of Format on Amount Read
The effect of format on how much readers read, scan, scroll, and/or click through
screen-based articles is still an open question. In 1994 and 1995, Nielsen (2000) found that
few users ever scrolled; perhaps 10 percent of users would scroll beyond the information
available when the page loaded. The one exception was users who arrived at a page with an
article they found interesting or important to their work: They would scroll as they scanned
long articles. On navigation pages,7 however, users would choose from among visible
options--meaning that "even if they would be willing to scroll, many users will make their
selection from whatever options are visible 'above the fold' if they see one that looks
promising" (p. 115). More recent studies have partly confirmed these findings (Nielsen,
2000). Yet, more users have started scrolling (Nielsen, 2000; Spool, 1999b).8 Nielsen
(2000) attributes this "willingness" to scroll to the prevalence of "badly designed, long
pages on the Web [having] inured most users to some amount of scrolling" (p. 115).
Nielsen's discussion of his research hints that a user's willingness to scroll through
an article might depend on the user's interest in the article. Therefore, Nielsen's research
suggests that an experiment testing the effects of linear, scrolling formats on amount read
should consider user interest in an article as a possible intervening variable.
7 Since Nielsen's book Designing Web Usability: The Practice of Simplicity is about
designing usable Web sites, it is assumed that he is discussing navigation options that
allow users to move through a single site.
8 Spool (1999b) "never saw any user frustration with user scrolling" in his Web usability
experiments (p. 77).
At any rate, much of the general knowledge of scrolling behavior is based primarily
on Web usability research (Morkes & Nielsen, 1997; Spool, 1999b) rather than on research
on online news. The question of how to measure the amount of information a user reads
from a linear news article is still open. It is important to know how far users will scroll
down and continue reading linear articles--or if they scroll up and down, skipping among
sections (which would suggest that a nonlinear format could better accommodate their
behavior and/or needs). If users are not scrolling and reading the information in long, linear
articles, why continue to publish articles in that format?
A Stanford University and Poynter Institute study of online news readers (2000)
found that participants' eyes "systematically went over more than 75 percent of the length
of almost all those articles presented to them" (Introductory Highlights, 17). As the
researchers suggest, this finding is not surprising: Participants purposely selected most
articles that they looked at in the study because they were interested in the linked headline
or brief. This result is also conditioned by characteristics of their participants (e.g.,
participants read online news at least three times a week and were the kind of people who
would volunteer to participate in such an experiment).
The eyetracking study measured "the vertical length to which subjects perused an
article" (Introductory Highlights, 17).9 This measure does not appear to account for the
possibility that users might progress through a linear story nonlinearly, scrolling both down
and up--perhaps seeking sub-headings that would interest them. It is possible that some
9 "Reading" was defined technically as "flat right behavior": a sequence of eye fixations
(or pauses) moving from left to right across the screen in a rough horizontal line. More
detailed operational definitions of these terms are included in the online report at
(click on "Terminology").
users of a linear article skip to the ending and then return to the middle having never read
the first paragraph. A new technique is needed to measure the amount of a linear, scrolling
article read when users do not progress linearly through an article.
A technique developed for this experiment allows for comparison of amount read
between users of a linear article and users of a nonlinear hypertext article. It is relevant to
compare how much of an article users of competing formats read: Tremayne (2000) found
that the number of hypermedia story presentations is increasing on 10 major news Web
sites1 but acknowledged that whether users are using optional links is unknown. His
content analysis over three years found that the percentage of main stories presented
without any hyperlinks fell from nearly half in 1997 to less than 30 percent in 1999.11 Two
points should be noted: First, the study emphasizes contextual links to multimedia
components (e.g., a country mentioned in story text is linked to a map that is not
necessarily designed for that story). Second, the sample of sites is not random and contains
non-newspaper sites. The study, however, raises an important point: Increasing use of
hyperlinks in stories "means nothing if people do not use the optional links" (p. 22).
The key element of hypertextual news formats is the hyperlink, which offers users a
choice of accessible nodes. But presenting news in nonlinear hypertext formats is futile
10 Tremayne (2000) analyzed the Web sites of the following news outlets: The New York
Times ; USA Today ; The Washington Post
; Time Daily ; US News
; ABC ; CBS ; CNN
Interactive ; Fox News ; and MSNBC
11 It is interesting that, while the mean percentage of linear stories fell from 46 percent in
1997 to 26 percent in 1998, the mean percentage of linear stories increased to 29 percent
in 1999 (Tremayne, 2000).
unless users access different nodes and read the information presented in them. Thus, it is
important to study this aspect of user behavior. Knowing how much of nonlinear hypertexts
users read can give journalists a general idea of how much information and how many links
to include in hypertext stories. Naturally, the amounts of information and links to include
would vary from story to story. However, looking at how much of nonlinear hypertexts
users read--and comparing the amount read with users reading scrolling, linear articles--can
help determine if writing articles (or chunking articles written for print) and presenting
them as nonlinear hypertexts is worthwhile. If users read at least the same amount of text
and find the hypertext more satisfying and more credible, then online news outlets might
proceed with cost-benefit analysis to see how the price of delivering news in hypertext
formats relates to how much more satisfying and credible users find their hypertexts.
Effect of Format on Perceived Story Credibility
Credibility is critical if readers are going to continue to embrace and accept online
news sources; lack of credibility could keep the Web from becoming a major news source
in the immediate future (Johnson & Kaye, 1998). In terms of story content, perceived
credibility of online news stories is significantly enhanced by source attribution (Sundar,
1998). Deuze (1998) has suggested that online journalists can establish credibility
effectively by linking to information about how they got the story as well as to sources of
the statements, facts, and analysis included in the story. However, the question of how
nonlinear story formats that incorporate links (versus traditional linear formats that do not
employ links) affect perceived credibility has not been tested in the literature.
As Johnson and Kaye (1998) have argued, credibility of online news is crucial
"because past studies suggest that people are less likely to pay attention to media they do
not perceive as credible" (p. 325). Johnson and Kaye assessed user perceptions of
credibility of online political information and compared online newspapers, online news
magazines, online candidate literature, and online political issue-oriented sources to their
traditional print versions in terms of credibility. Two weeks before and two weeks after the
1996 presidential election, the researchers surveyed 308 politically interested Web users
online and found that online media sources tended to be rated more credible than their
traditional versions--a finding which confirmed a Pew Research Center (1996) finding that
online users surveyed judged the Internet as a more credible news source than traditional
print or broadcast media.12
Johnson and Kaye found that both online and traditional sources were judged as
only somewhat credible: Online newspapers, along with online news magazines,
specifically were judged "somewhat" credible by about 7 out of 10 respondents, compared
with 4 out of 10 for candidate literature and 6 out of 10 for issue-oriented sites. Yet, online
newspapers, along with candidate literature, were judged as significantly more credible
than their traditional print counterparts, compared with online news magazines and issue-
oriented Web sites, for which the mean credibility scores nearly matched those for their
12 In the Pew Research Center (1996) study, respondents were asked "Which of the
following statements comes closer to your opinion of the Internet?" and given three
answer choices. A majority (56 percent) agreed that "these days you are more likely to
find accurate information about what is going on from the Internet than in the daily
newspapers or on the network news"; only 22 percent agreed that "a lot of what you find
on the Internet cannot be believed"; 12 percent answered "neither" and 10 percent said
they did not know or refused to answer. A second Pew study that compared the
credibility of traditional and online news sources did not compare perceived credibility of
print parents/online counterparts; however, broadcast news Web sites were found to be
perceived as more credible than their parent organizations (Pew Research Center, 1998).
The researchers also found that reliance on online sources is significantly related to
credibility assessments, confirming the suggestion of"past studies ... that a medium's
credibility is strongly related to the degree to which people rely on it" (p. 334). Reliance
was found to be more strongly associated with credibility than Internet use.
Johnson and Kaye's study focused on comparative assessment of perceived
credibility of online versus traditional media sources. Based on the results, the outlook is
good for online newspapers, which received comparatively higher credibility ratings than
both their print counterparts and their online competitors (except online news magazines,
which received the same rating). However, because the findings are based on a small
convenience sample, they are not generalizable to either the Internet population or the
general population. Furthermore, the researchers did not concern themselves with why
users judged media sources the way they did. Hence, it is still unknown how the linearity or
nonlinearity of sources might affect perceived credibility. This thesis includes an
examination of just that.
Research Questions and Objectives
This thesis's general research questions include:
* Does presenting newswriting in a nonlinear hypertext format enable users to recall
news content more accurately than when news is presented in a linear electronic
* Do users find nonlinear news story formats satisfying?
* How much of hypertexts do users read?
* Do users find nonlinear news stories credible?
Through these questions, this study examines whether nonlinear story formats are
appropriate for delivering news in online newspapers. It also attempts to discover how
nonlinear story formats compare with current online story formats, which are grounded in
the linearity of the print tradition.
The hypotheses for this study include hypotheses that anticipate differences
between treatments (H1, H2, and H5), hypotheses that anticipate significant and positive
correlations between pairs of dependent variables regardless of treatment format (H3 and
H4), and a hypothesis that anticipates no difference between treatments (H6).
H1--Participants who use the linear treatment (scrolling) will be significantly more
accurate in their answers concerning content than participants who use the nonlinear
This hypothesis is based on an experiment by McKnight, Dillon, and Richardson
(1990, cited in McKnight, 1996). The study found that participants who read a text on
paper were significantly more accurate in their answers about the information presented
than those who read hypertext versions of the same text.
H2--Participants who use the nonlinear treatment (hypertext) will have significantly
higher amount read scores than participants who use the linear treatment (scrolling).
"Amount read" refers to the amount of the story participants read.
This hypothesis is based on the expectation that the range of choices in the
nonlinear treatment will provoke participants who are presented with the nonlinear
treatment to access and read more of the article than participants who are presented with
the linear treatment. This expectation is influenced by Nielsen's (2000) observation that
Web users are impatient and therefore "tend not to read streams of text fully. Instead, they
scan text and pick out keywords, sentences, or paragraphs of interest while skipping over
the parts of the text they care less about" (p. 104). It is expected that participants who use
the nonlinear treatment will access and read more of the article, even if they only scan the
contents of the nodes.
H2a--Participants who use the nonlinear treatment (hypertext) will spend
significantly more time reading the article than participants who use the linear treatment
This hypothesis is based on the expectation that the impatience of Web users noted
by Nielsen (2000) will lead participants in the linear treatment to progress more quickly
through the article because its linear narrative form is more familiar. It is expected that
participants in the nonlinear treatment will spend more time reading the article because the
narrative form is presumably less familiar than the traditional linear narrative.
H3--Regardless of treatment format, participants' hypertext comfort scores will be
significantly and positively related to their scores on all dependent measures.
The sub-hypotheses are as follows:
H3a--Regardless of treatment format, participants' hypertext comfort scores will be
significantly and positively related to their recall scores.
H3b--Regardless of treatment format, participants' hypertext comfort scores will be
significantly and positively related to their accuracy scores.
H3c--Regardless of treatment format, participants' hypertext comfort scores will be
significantly and positively related to their amount read scores.
H3d--Regardless of treatment format, participants' hypertext comfort scores will be
significantly and positively related to their perceived story credibility scores.
H3e--Regardless of treatment format, participants' hypertext comfort scores will be
significantly and positively related to their user satisfaction scores.
The sub-hypotheses are based on the expectation that participants who feel
comfortable with hypertext understand how this kind of system usually works. These sub-
hypotheses assume that participants who are comfortable with hypertext will know how to
navigate or scroll down to read the article. It is further assumed that these participants will
be less distracted mentally by scrolling or navigating than participants who are less
comfortable with hypertext; thus, hypertext comfort would increase participants' capacity
to recall more information accurately. It is also expected that their familiarity with the
medium will cause them to read more, to attribute more credibility to the story, and to feel
H4--Regardless of treatment format, participants' perceived story credibility scores
will be significantly and positively related to their user satisfaction scores.
This hypothesis is based in part on the findings of past research, cited in Johnson
and Kaye (1998), that people are less likely to pay attention to media they do not perceive
as credible. The researcher assumes that a high credibility rating will mean that the
participants will have paid a high level of attention to the article and will have felt satisfied
with their experience.
H5--Participants who use the nonlinear treatment (hypertext) will have significantly
higher satisfaction scores than participants who use the linear treatment (scrolling).
This hypothesis is based on Fredin's (1997) and Murray's (1997) suggestion that
nonlinear hypertext is an emotional medium that, if constructed and used well, can provide
a strong sense of emotional satisfaction. The researcher assumes thatthis suggested strong
sense of emotional satisfaction will play out in higher user satisfaction scores for
participants who receive the nonlinear treatment.
H6--Recall scores for participants who use the nonlinear treatment (hypertext) will
not differ significantly from recall scores for participants who use the linear treatment
This hypothesis is based on the results of Mensing, Greer, Gubman, and Louis's
(1998) experiment that tested whether information recall is affected by online news
presentation style. The researchers found no differences in recall between participants
reading linear news articles that required scrolling down multiple screens and participants
reading the same articles in a nonlinear hypertext format. Though the results of their study
contradict results of the often cited recall study by Gordon, Gustavel, Moore, and Hankey
(1988, cited in Charney, 1994; McKnight, 1996; and Nielsen, 1990) comparing hypertext
with linear electronic text, this hypothesis is based on Mensing et al. (1998) because the
design and objectives of this study are more like those in the Mensing et al. study. Gordon
et al. asked participants to read two articles, one in each format, with half the participants
reading general interest articles and half the participants reading technical articles. Mensing
et al. asked participants to read two news articles from a simulated version of an online
newspaper, with each participant receiving either a linear treatment or a nonlinear
treatment. Furthermore, due to the popularization of the Internet by the World Wide Web in
the 1990s, the researcher assumes that participants in general will be more familiar with
hypertext systems and less prone to any disorientation (the feeling of being lost within a
hypertext) possibly experienced by those who received a nonlinear treatment in the 1988
This study was aimed at determining the effects of online newspaper story format
on users' recall, accuracy, satisfaction, the amount of a story they read, and perceptions of
story credibility. An experiment was conducted to compare a nonlinear hypertext (chunked)
version of an online newspaper story to a conventional linear (scrolling) version of the
same story. In the experiment, participants were randomly assigned to one of two groups
that received either the linear treatment or the nonlinear treatment.
The 135 participants in this study were enrolled in an undergraduate advertising
course in the University of Florida's College of Journalism and Communications. The
researcher e-mailed the instructor the experiment schedule, and the instructor invited
students to sign up for a session that would take place in the following two weeks.
Ten sessions were held between January 29, 2001 and February 7, 2001. Seven
sessions were held on 5 consecutive days during the week of January 29: 2 sessions on
Monday; 1 session on Tuesday; 2 sessions on Wednesday; 1 session on Thursday; and 1
session on Friday. Three sessions were held the following week: 1 on Monday, 1 on
Tuesday, and 1 on Wednesday. Students who showed up for their sessions and participated
in the experiment received the same amount of extra credit points for their advertising
Two treatments--a linear treatment and a nonlinear treatment--were developed on an
IBM-compatible P.C. using Notepad and hypertext markup language (HTML) (see
Appendix A for examples of treatment screens). Both contained an identical news article
from the online edition of the daily newspaper The News-Press, Fort Myers, Florida. The
article, headlined "To the roof of Africa" and retrieved in October 2000 from the Web
address http://www.news-press.com/kilimanjaro/kilimain.html, ran in both the print and
online editions in September 2000. Permission to use the story was given by News-
Press.com Online Editor Jeff Roslow.
The researcher chose this article for four reasons. First, the researcher thought that
the storyline (18 individuals from southwest Florida attempt to climb Mount Kilimanjaro to
raise money for a children's hospital) would be appealing to the participants (college
students in Florida). Second, the researcher believed that the news story was not one that
the participants were likely to have been exposed to already because "Kilimanjaro for
Kids" was not a national news event.
Third, although the narrative was sequential, the story was written in parts. The
structure of the 4,231-word story could be broken down into 20 "components" as defined in
McAdams and Berger (2001). The components decided on by the researcher ranged in size
from 101 to 370 words (see Appendix B for a breakdown of the treatment-design process).
In addition, the story included brief profiles of its 18 main "characters" (the hikers from
southwest Florida). The profiles could be formatted in an HTML table and counted as a
21st component. The researcher thought that chunks of these lengths would be manageable
to read on a 17- or 21-inch computer monitor because they would not require an
overwhelming amount of scrolling.
The fourth reason the researcher chose this article was that it has qualities of a
multiform story as described by Murray (1997). Murray uses the term multiform story to
describe "a written or dramatic narrative that presents a single situation or plotline in
multiple versions, versions that would be mutually exclusive in ordinary experience" (p.
30). Such stories "often [but not always] reflect different points of view of the same event"
(p.37, researcher's emphasis). Though the narrative is told sequentially in one voice in
keeping with the linear tradition, the researcher believes that these different points of view
emerge in the journalist's telling of the story through quotations and through the brief
profiles of the 18 hikers (each hiker describes a personal memory from the trip). Based on
Murray's idea of a multiform story, the researcher reasoned that this story represents one of
the story types that would be especially appropriate for the online medium.
The format definitions used in the experiment are consistent with the definitions
used in the Huesca et al. (1999) study of readers' responses to competing narrative forms
for online news stories. The linear treatment used in the experiment presented the story in a
fixed order. Essentially, the story was an electronic copy of its printed version, which,
"when accessed by a computer, must be read by scrolling down the screen from beginning
to end" (Huesca et al., 1999, 14). The nonlinear treatment used in the experiment
contained no new material; only transitional phrases were altered slightly; and some design
reformatting was done to the character profiles. Essentially the same story text was "broken
into thematic parts that had to be activated by clicking on links along the left side of the
computer screen" (Huesca et al., 1999, 14). Mensing et al. (1998) used treatments with
similar features to test recall and enjoyment of competing linear and nonlinear online news
In the design of both treatments, the researcher attempted to follow Web design
guidelines in Web Usability: The Practice of Simplicity (Nielsen, 2000).
The experimental sessions were held in computer labs in the University of Florida's
College of Journalism and Communications. Sessions were held in three different labs due
to constraints built into the lab schedule; however, each lab had the same basic layout.
Each lab had 21 computers situated along three walls of the room in a U-shape.
Monitor sizes differed in the labs. The researcher made note of which participants used the
treatments on 17-inch monitors and which participants used the treatments on 21-inch
monitors. Linear treatment participants were directed to the computer terminals on and
facing the windowed side of the room (opposite the room entrance) and along and facing
the wall of the room represented by the bottom of the U. Linear treatment participants were
seated on these sides because upon entering the lab participants would face the windowed
side of the room and could notice the participants reading a treatment. The researcher
thought that--if any participants entering the room happened to notice what was on the
linear treatment screens from across the room--seeing the linear treatment (designed like
most online news articles they would have been exposed to before) would not influence the
participants' expectations concerning what they were about to read. Nonlinear treatment
participants were directed to sit on the opposite side of the room (facing the entrance). The
lab layout made it so that the participants in different treatment groups had their backs to
one another. A U-shaped arrangement of desks in the center of the room (behind the
computer terminals) provided a place to set the questionnaires until the participants were
ready for them.
When participants sat at their assigned terminal for the experiment, the computers
displayed identical first screens that asked the participants to press Control + Alt + Delete
to log on. Separate network accounts were created for each treatment to minimize the
possibility that participants would receive the wrong treatment. Participants in the linear
treatment were instructed on the instruction sheet to log on under the username
"JOU5000a"; participants in the nonlinear treatment were instructed to log on under the
username "JOU5000b" (Appendix F shows the linear treatment instructions; the only
difference between the two instruction sheets was the username). The researcher set the
treatment as a read-only file that appeared as the Internet Explorer browser default page. A
complicated password was designed to minimize the chance that participants would
remember it and use it to log on to the network via the account while the experiment was
not in session.
Instruments and Measures
Two questionnaires were prepared to measure the dependent variables recall,
accuracy, amount read, user satisfaction, and perceived story credibility (see Appendices D
and E). Coding instructions were employed to ensure proper coding of data (see
Appendices H and I).
Questionnaire 1 pertained to recall measurement (see Appendix D). The first and
only question on this questionnaire asked participants to list all the things they could
remember from the article they had just read. The sheet offered 30 numbered, horizontal
lines for responses; the question instructed participants to list one item per line. The
question was presented on its own sheet so that participants could not use information from
questions designed to measure accuracy and amount read in their recall responses.
Two judges--the researcher and a colleague who did not know the study's final
objectives--counted the lines containing items remembered by each participant; this number
was considered the recall measure.1 In the case where one participant drew extra lines and
numbered them 31 and 32, the coders disregarded the additions and assigned the participant
a recall score of 30 in keeping with the rule established prior to the start of coding (see
Appendix H, item 35). The judges reported 100% agreement. The scores given to
participants by one of the judges was considered the recall measure.
Questionnaire 2 pertained to the measurement of accuracy, amount read,
satisfaction, and perceived story credibility. Questions 1 through 5 were designed to
measure accuracy. The researcher asked four questions concerning basic content (number
of hikers, hikers' names and occupations, phrase used by hikers to mean "go slowly," and
official name of the event). The answers to these questions appeared at least twice in the
article and in both treatments. The questions were designed for answers that could be easily
marked as accurate or inaccurate.
The two judges agreed on possible accurate answers for the more open-ended
questions about names and occupations to ensure that a participant would get credit for
variations on accurate answers (e.g., "Dave" was acceptable for hiker David Webb). They
designed a list on which they based their judgments (see Appendix I). The two judges
coded participants' answers to questions 1 through 5 to rate participants' accuracy.
Questions 3 and 4 prompted participants to recall the names of two hikers and the
occupations of two hikers. Therefore, participants could receive a maximum score of 7 for
1 These answers were not checked for accuracy, which was measured separately using
questions 1 through 5 on Questionnaire 2 (see Appendix E).
accuracy. Correct answers were coded as "1"; incorrect answers were coded as "O";
missing answers were coded as "9" (the standard code for missing answers) and later
converted to zero. The judges reported 100% agreement. The scores given to participants
by one of the judges was considered the accuracy measure.
Questions 6 through 10 were designed to measure the amount of the story
participants read. Questions were based on content from five sections in the article. The
sections were determined by dividing the number of words in the article (4,234) by five
(846.8). Choices A and B offered one correct answer and one incorrect answer, although
not necessarily in that order. The third and final choice gave participants the option to
check "Didn't read that part." Participants were given 1 point for answering either A or B
and zero points for C. In other words, it was assumed that an attempt to answer a question
by choosing A or B meant the participants had read that part of the story. This method was
chosen so that the measure would not discriminate against participants with relatively
limited abilities to remember what they had read.
Questions 11 through 16 measured user satisfaction. The questions were adapted
from the scale used to measure user satisfaction in a published study that explored user
satisfaction with an interactive encyclopedia (see Kamerer and Wilcox, 1993). The same
scale was adapted in Mueller and Kamerer's (1995) study of reader preference for
electronic newspapers. Participants read an item and circled the number that corresponded
to their agreement (Strongly Agree, Agree, Neutral, Disagree, Strongly Disagree) with the
item. The items were: "I wish all news material were designed like this article"; "The
layout of the article was attractive"; "The article was extremely detailed"; "It was difficult
to read from the screen"; "It was difficult to understand the point of the article"; and "It
was easy to read the story."
Question 17 functioned as a manipulation check. It asked participants if they had to
click blue and/or purple hyperlinks to read different parts of the article.
Questions 18 through 23 were adapted from the 5-point bipolar-statement news
credibility scale in Rubin (1994). The items were: Is factual/Is opinionated; Is biased/Is
unbiased; Tells the whole story/Does not tell the whole story; Is accurate/Is inaccurate;
Does separate fact and opinion/Does not separate fact and opinion; Can be trusted/Cannot
The rest of the questionnaire consisted of demographic questions (gender and age)
and questions concerning comfort with hypertext. Participants also were asked if they had
known previously about the news event and if they had read the article previously. The goal
of these questions was to verify whether such circumstances had any effect on recall and
accuracy. In the two cases where participants reported that they had known previously
about the event, the effects were not statistically significant.
The time participants spent with the story was measured using the instruction sheet
(see Appendix F). Participants recorded the time they started and finished using the
treatment in the blanks provided on the instruction sheet using the Microsoft Windows
system clock available on the computer screen. Time spent with the treatment (in minutes)
was calculated by each coder, who wrote the number representing time spent by each
participant in the bottom left-hand corer of each instruction sheet and circled it. The
judges reported 100% agreement.
The treatments and questionnaire were pre-tested during an electronic publishing
class taught by a faculty member from the College of Journalism and Communications. The
pre-test took place in a lab similar to the labs used in the experiment but outside of the
College of Journalism and Communications. Seven participants used the linear treatment
and five used the nonlinear treatment. The unequal size of the treatment groups resulted
from the layout of the lab (divided by an aisle down the center) and from where participants
were sitting when the researcher entered the lab to administer the pre-test. To minimize
participants' awareness of manipulation in the treatments, participants were asked to
remain in their seats. The researcher stood at the front of the room to explain the procedure.
Participants on the right-hand side of the room from the researcher's perspective received
the nonlinear treatment; participants on the left-hand side of the room from the researcher's
perspective received the linear treatment. Pre-test participants were asked to make a check
mark beside the number of any question and/or instruction that gave them trouble.
Due to the fact that the pre-test was conducted in the last 20 minutes of the students'
class, time constraints allowed for only 10 minutes of reading time. To confirm the
researcher's assessment (based on informal pre-tests with family and friends) that the story
would take at least 15 minute to read, pre-test participants were asked to answer the
multiple choice question: "Did you need more reading time? (a) Yes, I needed about 2
more minutes. (b) Yes, I needed about 5 more minutes. (c) Yes, I needed about 10 more
minutes. (d) No, I had enough time." Five participants circled "b"; two participants circled
"c"; three participants circled "d"; and two answers were missing. This question confirmed
the researcher's assessment, resulting in the final instruction for the experiment: "It will
take most people AT LEAST 15 minutes to read the entire story. Read the entire story,
scrolling and/or clicking as necessary to access the text." (see Appendix F for Instruction
Emphasis was placed on the actions of scrolling and/or clicking because one pre-
test participant in the nonlinear treatment did not notice the story links. This prompted the
researcher to label the link choices with the heading "Story Links" and to change the
anchor link at the bottom of each component from "Back to the top of this page" to "Back
to the top of this page: Story Links." The underscore in the previous sentence represents
linked text that, when clicked, takes the user to the top of the present page, eliminating the
need for upward scrolling. This link was included on every page in the nonlinear treatment
consistently to increase usability.
The pre-test also proved useful in the design of the manipulation check used in the
final experiment. In the pre-test, the manipulation check asked: "Did the article contain
blue and/or purple hyperlinks in the left-hand column that led to different parts of the
article? Circle one: Yes No." Four participants (one in the linear treatment; three in the
nonlinear treatment) missed the question; one participant in the linear treatment failed to
answer the question and drew a question mark beside the answer choices. The researcher
decided that the words "in the left-hand column" complicated the question unnecessarily by
making the participants think about "left and right," which was not directly relevant to the
manipulation. Possibly the participants, students in a beginning electronic publishing class,
did not understand the word "hyperlinks." Altogether, this reasoning helped produce the
manipulation check ultimately used: "Did you click blue and/or purple hyperlinks to read
different parts of the article? Circle one: Yes No."
The pre-test also helped the researcher refine the measure of amount read in the
story. The researcher changed the number of choices available for Questions 6 through 10
from four choices (2 incorrect answers; 1 correct answer; 1 "Didn't read that part") to three
choices (1 incorrect answer; 1 correct answer; 1 "Didn't read that part"). The researcher
originally planned to count only correct answers toward the amount read score; however,
this method would have discriminated against participants with relatively less strong
abilities to remember what they had read. Because it is possible for a participant to have
read a part of the story but remember the details incorrectly, the researcher decided to
assume that an attempt to answer the question by circling a choice other than "Didn't read
that part" meant the participant had read that part of the story.
The procedures used in this study are summarized as follows:
1) Students from an undergraduate advertising class were invited by their instructor to
attend an experiment session to test an online newspaper article.
2) As the student participants came into the lab for the experiment, they were
systematically assigned to the linear treatment or to the nonlinear treatment. The
experiment was administered on a first-come, first-served basis. Treatments were
assigned in the order in which participants arrived as follows: The first participant to
enter the lab received the linear treatment; the second participant received the
nonlinear treatment; the third participant received the linear treatment, etc.
3) The researcher placed a packet containing all the experimental materials on a desk
behind the participant. Each page in the packet was marked with 1L, 3L, 5L, etc. for
participants in the linear treatment and 2N, 4N, 6N, etc. for participants in the
nonlinear treatment. The numeral in the code served as a unique identification number
for each participant and as a mechanism for keeping questionnaires completed by the
same participant together. The letter (L or N) in the code served no purpose other than
allowing the researcher to minimize error in questionnaire distribution. After the
participant signed the informed consent form, the researcher gave the participant the
instruction sheet (see Appendix F).
4) Participants were asked on the instruction sheet (see Appendix F) to record the times
they started and finished using the treatment in the blanks provided on that sheet using
the Microsoft Windows system clock available on the computer screen.
5) Participants were asked to raise their hands when they were finished using the
treatments. They then received Questionnaire 1 (see Appendix D). Participants were
asked to raise their hands when they had finished Questionnaire 1. They then received
Questionnaire 2 (see Appendix E).
6) After completing Questionnaire 2, participants were asked to sign a sheet unrelated to
data collection in order to verify their participation in the experiment for their
instructor. As they signed the sheet, the researcher reminded them to read the
debriefing note on the door clearly marked EXIT and thanked them for participating.
The note further explained that participants had been misled to believe that they all had
received the same treatment whereas there were two different treatments (linear and
nonlinear). It also urged them not to share any information about the experiment with
their peers who possibly would be attending subsequent sessions (see Appendix G).
Scale Reliability Analysis
Scale reliabilities were tested with Statistical Package for the Social Sciences
(SPSS). Questions 11 through 26 (except for question 17--the manipulation check; see
Appendix E) were divided into three categories: user satisfaction, perceived story
credibility, and comfort with hypertext.
Results showed that the scales had acceptable levels of reliability. The result for
user satisfaction showed a Chronbach's alpha score of .6795. The fourth item ("The article
was extremely detailed") was dropped to increase reliability.
Results showed a higher level of reliability for the items measuring credibility
(.7242) and a lower level of reliability for the items measuring comfort with hypertext
(.5975). However, the researcher found all reliability levels acceptable due to the small
numbers of items used to compute satisfaction (5), perceived story credibility (6), and
hypertext comfort (3) scores.
The data were analyzed using SPSS. To test the hypotheses, analysis of variance
and correlation were employed. ANOVA also was used to determine (1) whether knowing
about the event or reading the article before the session had any significant effects on recall
and accuracy and (2) whether monitor size had any significant effects on recall, accuracy,
and user satisfaction.
Summary of Findings
A majority of the hypotheses that anticipated differences between the treatments
(linear format vs. nonlinear format) were not supported. However, three of the seven
expected correlations were statistically significant: Significant correlations were found
between hypertext comfort and user satisfaction, hypertext comfort and perceived story
credibility, and perceived story credibility and user satisfaction regardless of treatment
format. These findings suggest a positive outlook for online news as long as the public
becomes increasingly comfortable with hypertext.
The hypothesis that predicted no difference between the treatments was supported.
Findings showed no significant difference between the recall scores of participants in the
nonlinear and linear treatments, suggesting that linear, screen-based formats and nonlinear
hypertext formats allow readers the same amount of recall power.
One hundred thirty-five participants took part in the experiment. One hundred
twenty-five produced usable questionnaires. Four questionnaires (3 nonlinear treatment; 1
linear treatment) were dropped because the participants failed to answer the manipulation-
check question correctly; all four participants' hypertext comfort scores fell below the
sample mean. Four more questionnaires were dropped because participants reported a
reading time less than 10 minutes--less than half the average amount of time spent reading
(20.31 minutes), suggesting that these participants did not follow the instructions to read
for at least 15 minutes. An additional two questionnaires were dropped because participants
reported having known about the event described in the article before the experimental
Most participants were between 18 and 22 years old (90 percent); 10 percent were
23, 24, or 25. Ninety-three participants (74 percent) were female and 32 participants (26
percent) were male.
Variables for Analysis
The independent variables for analysis included story format (linear or nonlinear).
The independent variables were:
* Story format (linear or nonlinear)
* Prior knowledge of the event
* Prior exposure to the story
Participants' answers to recall, accuracy, amount read, hypertext comfort, user
satisfaction, and perceived story credibility questions were considered in the measurement
of the dependent variables. The dependent variables were:
* Recall--the number of items a participant recalled from the story.
* Accuracy--the number of correct responses to five open-ended questions (two of the
five questions had two parts for a total of seven possible correct responses).
* Amount read--the number of five multiple-choice questions that the participant
attempted to answer.
* Time spent reading--the time at which the participant reported finishing reading the
story minus the time he or she reported starting to read the story.
* Comfort with hypertext
* User satisfaction
* Perceived story credibility
Comfort with hypertext was measured using a four-point scale (see Appendix E,
questions 24-26). Table 4-1 shows the mean and standard deviations for hypertext comfort.
A reliability analysis in SPSS showed a Chronbach's alpha score of .5975. Based on this
score, the three-item scale was judged sufficiently reliable to be used as a scale.
User satisfaction and perceived story credibility were measured using 5-point scales
(see Appendix E for questionnaire: questions 11-16 for satisfaction; questions 18-23 for
perceived story credibility). Table 4-1 shows the means and standard deviations for user
satisfaction and perceived story credibility. A reliability analysis in SPSS for the scale used
to measure user satisfaction showed a Chronbach's alpha score of .6795; a reliability
analysis for the scale used to measure perceived story credibility showed a Chronbach's
alpha score of .7242. Based on these scores, the scales were judged sufficiently reliable to
be used as scales.
Table 4-1: Means and standard deviations for hypertext comfort, user satisfaction, and
perceived story credibility
Dependent Variable Mean SD
Hypertext Comfort 2.22 0.51
User Satisfaction 3.50 0.61
Perceived Story Credibility 3.60 0.64
Tests of Hypotheses
H1 had proposed that participants who used the linear treatment would have
significantly higher accuracy scores than participants who used the nonlinear treatment.
The results did not support this hypothesis. The difference between treatments was not
significant. In other words, participants who used the nonlinear treatment were as accurate
as those who used the linear treatment in recalling particular information in the article.
H2 had said that participants who used the nonlinear treatment would read
significantly more of the story than participants who used the linear treatment. The results
did not support this hypothesis. Participants in the nonlinear treatment had a higher mean
score (4.42) than participants in the nonlinear treatment (4.34) but the difference was not
significant. In other words, participants who used the nonlinear treatment read as much of
the story as did participants in the linear treatment.
H2a had proposed that participants who used the nonlinear treatment would spend
significantly more time reading the article than participants who used the linear treatment.
The results did not support this hypothesis. There was no significant difference between
nonlinear and linear treatment groups when time spent reading was compared, although the
trend was in the predicted direction (F=2.731, df=l, p=. 101) (see Table 4-2).
Table 4-2: Effect of story format on time spent reading
Treatment group Mean SD N
Linear 19.59 4.26 66
Nonlinear 21.03 5.48 59
F (df=1) =2.731, p. =.101
H3 predicted that regardless of treatment format, participants' hypertext comfort
scores would be significantly and positively related to their scores on all dependent
measures (except for time spent reading, which the hypothesis did not include). Findings
partially supported this hypothesis. The dependent measures accuracy and amount read did
not correlate significantly with hypertext comfort; however, the dependent measures
perceived story credibility and user satisfaction did correlate significantly with hypertext
comfort (see Table 4-3). The correlation between recall scores and hypertext comfort
approached significance (p=.059).
Table 4-3: Hypertext comfort correlations
Perceived Story User Recall Accuracy Amount
Credibility Satisfaction Read
Comfort with .179 .185 .170 -.033 -.008
Hypertext p=.047 p=.035 p=.059 p=.714 p=.933
H4 had suggested that regardless of treatment format, participants' perceived story
credibility scores would be significantly and positively related to their user satisfaction
scores. The results supported this hypothesis. The correlation between credibility and
satisfaction was significant at the p=.01 level (see Table 4-4).
Table 4-4: Correlation between credibility and satisfaction
H5 had proposed that participants who used the nonlinear treatment would have
significantly higher satisfaction scores than participants who used the linear treatment. The
findings did not support this hypothesis. In other words, participants who used the
nonlinear treatment were as satisfied as those who used the linear treatment.
H6 predicted that recall scores would not differ significantly between participants
who used the linear treatment and participants who used the nonlinear treatment. The
findings supported this hypothesis. The difference between the treatments was not
significant (F=.373, df=l, p=.542). In other words, participants who used the nonlinear
treatment recalled as much as those who used the linear treatment (see Table 4-5).
Table 4-5: Effect of story format on recall
Treatment group Mean SD N
Linear 20.94 6.89 66
Nonlinear 21.69 6.91 59
F (df=) =.373, p. =.542
Results Regarding the Research Questions
The first research question asked whether nonlinear formats enable users to recall
news content more accurately than do linear formats. The findings showed that there were
no significant differences between the two treatment groups for both recall and accuracy
scores. Simply put, the nonlinear hypertext story format does not seem to affect users'
recall or accuracy. This suggests that nonlinear Web-based hypertext is neither inferior nor
superior to a linear Web-based format for increasing users' ability to recall information and
to recall it accurately.
The second research question asked whether users find nonlinear hypertexts
satisfying. The findings showed that there were no significant differences between the two
treatment groups when user satisfaction scores were compared. In other words, users who
read the article in the linear format felt as satisfied reading the article as did users who read
the article in the nonlinear format. This means that, at worst, users are no less satisfied by
The third question asked how much of nonlinear hypertexts users read. The findings
showed that there were no significant differences between the two treatment groups when
amount read scores were compared. In other words, participants read the same amount of
the story regardless of treatment format. This result may be a function of the instruction
that participants in both treatment groups received to "read the entire story." However, the
finding suggests that story format does not affect how much of a story users read when
instructed to read an entire story of this type for an experimental purpose.
The fourth and final question asked whether users find nonlinear hypertexts
credible. The results did not show a significant difference in perceptions of story credibility
between the two treatment groups, although the mean score for the nonlinear treatment
group was slightly higher (3.69) than the mean score for the linear treatment group (3.53).
In other words, participants who read the nonlinear version found the story as credible as
did participants who read the linear version. This finding means that, at worst, users find
nonlinear formats no less credible than linear formats.
Hypertext--the technology that enables readers to follow multiple paths through
electronically-stored information--has been recognized as a promising format for online
storytelling, including online journalism (Huesca et al., 1999; Mensing et al., 1998;
Murray, 1997; Vargo et al., 2000). Despite the opportunity to take full advantage of
hypertext, most online newspapers continue to publish articles in linear formats that present
information as if it were written for the printed newspaper. The findings of this study
suggest that an online news feature published in a linear format is as effectively delivered in
a nonlinear hypertext format.
Given that the next generation of users most likely will take the associative features
of hypertext for granted, the ultimate viability of online newspapers might depend on how
effectively journalists use hypertext to convey information. In this case, their viability also
would hinge on how satisfied users of nonlinear hypertext formats feel, how credible users
perceive the nonlinear articles to be, and how much of the nonlinear articles users are
willing to read. In this context, it is important to find out whether publishing news articles
nonlinear hypertext formats enhances the audience's experience of and perceptions of the
information being communicated.
The purpose of this study was to analyze the effects of competing online news story
formats (linear vs. nonlinear) on users' recall, accuracy, satisfaction, and perceptions of
story credibility; this study also examined the amount of a story read by users of a linear,
scrolling story versus by users of nonlinear hypertexts.
Effect of Format on Accuracy
Findings showed that participants who used the nonlinear treatment were as
accurate in their answers to general questions about story content as those who used the
linear treatment. It is interesting that, although the difference was not significant, the
nonlinear treatment mean for accuracy was higher (.76) than the linear treatment mean
(.72)--a pattern that runs in the opposite direction of the results of the study (McKnight,
Dillon, and Richardson, 1990, cited in McKnight, 1996) that led to the hypothesis that
participants who used the nonlinear treatment would be significantly less accurate than
participants who used the linear treatment.
McKnight, Dillon, and Richardson (1990, cited in McKnight, 1996) concluded that
their finding--that users of hypertext were significantly less accurate in their answers about
information presented to them than users of a paper or linear word-processing format--
suggested that "the familiar structure inherent in [linear versions] supported [users']
performance" (p. 231). The lack of a significant difference between the two treatment
groups in this experiment would imply that users are more familiar with hypertext as a
structure for information delivery in 2001 than they were in 1990. This suggested increase
in familiarity might be limited to this population (college students) and attributable to
participants' exposure to hypertext during the course of their educations. For example, at
the University of Florida, students have been encouraged since the late 1990s to use the
university's Web-based hypertext system for administrative tasks such as registering for
courses, obtaining class schedules, and paying fees. Such activities may have contributed to
familiarity with hypertext as a structure for information delivery within this demographic
group, possibly making the structural-familiarity effect suggested by the results of
McKnight, Dillon, and Richardson (1990, cited in McKnight, 1996) less relevant when
comparing users of nonlinear- and linear-based systems--but maybe only for groups
familiar with hypertext.
This result also could mean that participants, who had been assumed to be more
familiar with linear formats, are as comfortable with hypertext formats (i.e., not distracted
by clicking links for information), enabling them to recall information from both formats
with the same accuracy. The finding also could mean that participants found the nonlinear
treatment design highly usable and were able to focus on the story content; the accuracy of
participants in earlier experiments might have suffered due to disorientation (the feeling of
being lost in a hypertext) brought on by the design of the nonlinear treatments. In addition,
the result could mean that the fill-in-the-blank questions designed to measure accuracy
were too simple, enabling participants to achieve accuracy regardless of the treatment they
received. In sum, it could be that participants are generally comfortable with hypertext, that
the nonlinear hypertext was well-designed, that the questions were too easy--or it could be
some combination of these possibilities. It should be noted, given the rise of the World
Wide Web in the 1990s, that an increased comfort with hypertext probably would hold
across the board for regular Web users.
Effect of Format on Amount Read
Results showed that story format had no effect on how much of the article
participants in either treatment group read. In other words, participants read the same
amount of the story regardless of format. This result could have been caused by the
instructions: All participants were asked to "read the entire story" or to read for at least 15
minutes. Findings showed that time spent with the article did not differ significantly
(participants who used the linear treatment spent 19.59 minutes; participants who used the
nonlinear treatment spent 21.03 minutes). This suggests that participants followed the
instruction to read the entire article or to read for at least 15 minutes.
The researcher does not believe that the results regarding amount read are a blow to
the reasoning on which the hypothesis was based. Had the participants been instructed to
read until they felt like stopping and/or had they been offered choices of articles that they
found more appealing, it is possible that participants in one treatment would have read
more of the articles) and/or spent more time reading the articles) than participants in the
other treatment. Subsequent research might instruct participants to read until they feel like
stopping and provide them a choice of articles to increase the chances of their sticking with
any one article long enough to achieve more meaningful results regarding amount read.
There also might have been a problem with the questions designed to measure
amount read (see Appendix D, questions 6-10). The researcher assumed that an attempt to
answer the question (i.e., participants did not circle "Didn't read that part") meant that
participants had read that content as well as nearby content, whether or not they could
recall it correctly;1 therefore, inaccurate answers were counted toward the amount read
score. It is possible that participants, who had been asked to read the entire story, hesitated
to circle the answer choice "Didn't read that part" and guessed at answers even when they
had not read certain parts of the story--despite the instruction to answer all questions "as
honestly and as best" they could.
1 As explained in Chapter 3, the researcher chose this method so that the amount read
measure would not discriminate against participants with relatively limited abilities to
remember what they had read.
Future research might combine the amount read and accuracy measures used in this
study. It also might employ eyetracking equipment to measure amount read. Future
eyetracking would need to improve upon the Stanford and Poynter (2000) measure of how
far "down" an article users read. Improved eyetracking would account for the eye
movement that accompanies both downward and upward scrolling as well as for whether
users were reading or scanning. When eyetracking is not feasible, more questions covering
more individual sections of article content should be asked.
Hypertext Comfort and User Satisfaction, Perceived Story Credibility
Results showed a positive correlation between hypertext comfort and user
satisfaction, meaning that--regardless of format used--participants who reported high levels
of hypertext comfort (relative to other participants) found the story more satisfying than did
participants who reported low levels of hypertext comfort. This finding is not surprising. It
makes sense that users who feel comfortable using a medium feel more satisfied with their
experience of information presented in that medium than do users who feel uncomfortable
with the medium. In most cases, users who feel comfortable with hypertext are likely in
control of their experience--they have learned to scroll and click to get information and
probably feel satisfied when they have called up certain information. On the other hand,
users who feel uncomfortable with hypertext may feel less in control--they probably are
more prone to disorientation. Users with minimal hypertext comfort might feel frustrated at
the idea of possibly losing their place. Based on the argument that hypertext can provide
emotional satisfaction, it seems natural that low comfort levels, which are likely
accompanied by disorientation and feelings of frustration, would negatively affect
Findings also showed that comfort with hypertext correlated significantly and
positively with perceived story credibility. In other words, participants who reported high
levels of hypertext comfort (relative to other participants) found the story significantly
more credible than did participants who reported low levels of hypertext comfort. It could
be that, as reasoned in the hypothesis, comfort with scrolling and or clicking online
documents to access information increases one's perception of information as credible. It
also could be that, given the association between reliance and credibility found in Johnson
and Kaye (1998), users who are comfortable with hypertext are more reliant on online
media in general, and that reliance on online media leads to a higher perceived credibility
as well as comfort.
In any case, it appears that online news sites have a chance of their articles being
perceived as more credible by the next generation of users (who will be familiar with
hypertext). This finding could be important for online newspapers, which can offer plenty
of what users focus on: text. This perceived story credibility could be useful in building and
retaining readership. Moreover, increased perceptions of story credibility in an online
newspaper could boost perceptions of the news organization that produces the online
newspaper. This finding appears to be good news for online news in general. Yet, given the
newspaper industry's credibility concerns (ASNE, 1999), it might be of particular interest
to the newspaper industry when considering online newspapers because it suggests that
solid investment of time and resources in crafting effective online news stories now could
pay off significantly with higher perceptions of the news organization's overall credibility
in the future.
Hypertext Comfort and Recall, Accuracy, Amount Read
Results showed that comfort with hypertext does not correlate with users' recall or
users' accuracy in recalling story content, although the correlation with recall approached
significance. In other words, participants who reported feeling comfortable with hypertext
did not have significantly higher recall or accuracy scores than those who reported lower
comfort levels. It might be that participants found the article generally interesting, paid
attention to it, and had no trouble recalling information from the story and answering
general questions about the content accurately.
Perhaps it should not be overlooked that the correlation of hypertext comfort and
recall approached significance: This suggests the possibility that comfort with hypertext
might increase the quantity of information one is able to recall from an article, maybe
because less cognitive effort is spent on navigating through the article. It also could be a
sign that users with good memories (i.e., high capacity for recall) are naturally more
comfortable with hypertext. In any case, given that participants who reported higher
hypertext comfort actually did not have significantly higher recall scores, the results of this
study suggest that presenting an article in either format (linear or nonlinear) would not offer
users who are comfortable with a significant advantage: Users who are not as comfortable
with hypertext would recall as much information from either story format and recall it as
Findings also showed that comfort with hypertext does not correlate with how much
of a story one will read by scrolling and/or clicking. Again, this finding might reflect the
instructions all participants received to read the entire story or to read for at least 15
minutes. Another explanation for this finding might be that the questions designed to
measure amount read were too easy or that participants guessed the answers rather than
circling "Didn't read that part."
Perceived Story Credibility and User Satisfaction
Results showed a significant correlation between perceived story credibility and
user satisfaction. Participants who perceived the story as more credible also were more
satisfied with their experience of the article in general. As reasoned in the hypothesis, it is
possible that participants who felt the story was credible paid more attention to it and
enjoyed it more than participants who perceived the story as less credible. Similarly, it is
possible that participants who felt they could trust the information that the story offered felt
highly satisfied with their experience of the article. However, it is also possible that
participants who felt satisfied with their experience of the article felt that they could trust
the information more.
From the data analyzed in this study, it is impossible to tell if credibility is a
predictor of satisfaction or vice versa. It would be interesting to replicate this test using a
more controversial story (e.g., a story about political issue) or multiple stories that had been
prejudged to vary significantly in their credibility to see whether the correlation holds.
Deliberate use of such a story or stories, combined with questions about participants' views
on and interest in the issue and/or the article itself, may be helpful in figuring out which
variable predicts which or whether one variable is a predictor of the other at all.
In any case, the correlation warrants further inquiry. If perceived story credibility is
found to be a predictor of satisfaction, then it is crucial for online news to be credible (as
well as to avoid the perception that the content cannot be trusted) in order to satisfy users.
Such a finding would have implications for how advertising and special sections are
handled on online news sites in general. However, if satisfaction is found to be a predictor
of perceived credibility, it is vital that online publications find out which formats satisfy
their users. This is where the importance of audience analysis comes into play.
Web usage is completely discretionary, so users must be kept happy (Nielsen,
2000). Keeping users happy requires Web designers to understand the needs of their
audiences. For example, if senior citizens make up the majority of publication's readership
and it is known that most senior citizens are not comfortable with hypertext, it would be
best to present articles in a linear, scrolling format that uses large, bold headlines to
distinguish story sections. However, if a publication's audience consists mostly of younger,
career-oriented members who do not have much time to read the news and who are known
to prefer clicking links to specific information presented briefly, it would be important to
present articles in nonlinear hypertext formats so that users could get the information they
need quickly and feel satisfied. Of course, it might be somewhat expensive to conduct a
scientific audience analysis to find out which formats users prefer. But, if satisfaction
predicts credibility, the expense could pay off later in terms of higher perceived credibility.
This issue deserves the newspaper industry's attention because a 1999 study by the
American Society of Newspaper Editors revealed that "the public perceives that
newspapers do not consistently demonstrate respect for, and knowledge of, their readers
and communities" (ASNE, 1999, 13). The newspaper industry might turn to online
newspapers to improve this perception of newspapers. Offering online news in formats that
meet users' needs would be one way of demonstrating knowledge of readers. Given the
correlation between credibility and satisfaction, respecting readers' online needs (by
presenting online news in formats that meet those needs and that readers find satisfying)
ultimately could enhance perceived credibility of the organization and its products overall.
Effect of Format on User Satisfaction
Results indicated that format (linear vs. nonlinear) had no effect on user
satisfaction. In other words, users of the linear format felt as satisfied as did users of the
nonlinear format. It might be that no significant difference was found because of the lack of
precision in the user satisfaction measure: The scale measuring user satisfaction was
sufficiently reliable (Chronbach's alpha was .6795 for the five-item scale), but it was less
reliable than the researcher had hoped it would be. Another possibility is that user interest
in the content was an intervening variable. The study failed to consider how much interest
participants had in the article they were asked to read. Future research probably should
consider this point in order to make future results about the effect of format on user
satisfaction more meaningful.
Results regarding user satisfaction also would be more meaningful if users could
choose from stories about a range of topics as in Huesca et al. (1999). This would increase
the opportunity for participants to choose a story that interests them. It could be especially
useful to test stories that are not likely to be inherently interesting but are of significant
importance (i.e., stories on health care reform, government budgets, and zoning). Testing
both stories that users find interesting and stories that users find boring would enhance
general understanding of the effects of format on user satisfaction.
The lack of a significant difference between treatments also might be explained, at
least in part, by the artificiality of the lab environment: Results might have been different if
participants (1) read the articles in their own homes or on whatever computer they find
most comfortable to use and/or (2) selected the article at their leisure without being asked
to read an entire story that may or may not have interested them significantly. What is
more, the age and general computer familiarity of the users might have limited the variance
in user comfort with both formats.
The finding that format did not affect user satisfaction probably should not be seen
as a strike against Fredin's (1997) and Murray's (1997) arguments that hypertext can
enhance users' experience of information. It is a limitation of the research design that users
were not offered a choice of articles. The Internet is a user-driven medium; but in this
experiment, the user experience was very much driven by the researcher, who selected the
article participants read. This limit on user choice may have had a negative impact on
participants' user satisfaction scores. Future research probably should incorporate story
choices, as did Huesca et al. (1999), to improve the testing of these theories.
As a final point, user satisfaction scores for participants in both groups fell
generally in the middle of the spectrum. The mean user satisfaction score for participants in
the nonlinear treatment was slightly higher (3.53) than the mean score for participants in
the linear treatment (3.46). This result may be read as a "go ahead" for journalists to
experiment with nonlinear hypertext story formats--since findings showed that
manipulation of the story format had no effect on user satisfaction. It appears that, at worst,
putting stories in nonlinear hypertext format does no harm.
Effect of Format on Recall
Results indicated that story format had no effect on recall. In other words,
participants who used the nonlinear treatment were able to recall the same amount of
information about the story as participants who used the linear treatment. This finding
confirms the research results of Mensing et al. (1998), who measured recall of linear and
nonlinear news stories, as well as the findings of Lee (1998), who tested the effects of
hypertext recall based on gender. Both found that story format had no significant effect on
This study might have confirmed the results ofMensing et al. (1998) because of
similarities in research design (participants were assigned to read either linear or nonlinear
articles). However, research similarity might not be the cause because this finding also
confirmed the result of Lee (1998), whose design differed from both the study by Mensing
et al. (1998) and this study (participants in Lee's study were assigned to read both a linear
and a nonlinear article). It could be that, in the year 2001, users are familiar enough with
hypertext that they are as capable of recalling information from nonlinear articles as from
linear articles. The recall-related findings ofMensing et al. (1998), Lee (1998), and this
study contradict the findings of Gordon et al. (1988, cited in Charney, 1994; McKnight,
1996; Nielsen, 1990). This flip-flop in recall findings could be due to an increase in
exposure to hypertext systems (especially the World Wide Web) in the 1990s.
The results of this experiment indicate that no harm would come from journalistic
experimentation with nonlinear hypertext formats for online news features. Findings
showed that users of the nonlinear format recalled as much information, recalled
information as accurately, felt as satisfied, read as much of the news feature, and perceived
the story to be as credible as did users of the linear format.
Online news editors might want to experiment with nonlinear hypertext formats for
news because, based on this study, it appears that users who feel comfortable with
hypertext (relative to other users) would be significantly satisfied with their reading
experience and would share significantly high perceptions of story credibility, regardless of
story format. This suggests that online news editors should not be discouraged yet if they
cannot afford to experiment with new formats; however, online newspapers that can afford
to might as well because it appears that experimenting with new formats would not damage
user satisfaction and perceptions of credibility--both would remain high and could be even
higher in the future if more effective formats are developed with the next generation of
hypertext-savvy users in mind.
This study concludes that experimentation with nonlinear formats, which has been
encouraged in the trade press (Fitzgerald, 1996; Dube 2000; Outing, 2000; Scanlan, 2000),
would not do harm and might prove crucial for the future. This conclusion has important
implications for online newspapers. Aside from any benefits that could stem from user
focus on text, experimentation with nonlinear text-based formats would be necessary if, as
Outing (2000) has suggested, newspaper companies of the future will be publishing to
home-printed newspapers, the Web, e-mail, PDAs, mobile phones, e-readers, and pagers in
addition to print-delivered newspapers.
This study does not suggest that newspapers would benefit from delivering or that
users would benefit from consuming all kinds of news stories on various electronic devices.
For example, it is unlikely that many users would read the 4,231-word feature story used in
this experiment on their mobile phones, no matter how brief and tightly-written the story
components were. Yet it is possible that many users would want to read this story on the
Web or on an e-reader (once the usability of e-readers increases sufficiently).2 Moreover,
2 For a discussion of e-readers and how they might incorporate news content, see Steve
Outing's February 2001 online column "Can news content save e-books?" Retrieved
March 5, 2001, from the World Wide Web:
the skills that newspaper journalists will develop from experimenting with nonlinear
hypertext formats (i.e., writing brief self-sufficient texts and linking effectively) will be
highly transferable to and critical for the publishing scene in the new millennium.
Limitations of the Research
This study had several limitations that should be considered (along with the ideas
suggested in the discussion) in interpreting its findings and designing future research. This
section discusses four main limitations.
Limitation 1: This study did not measure participants' interest in and liking of the
story. Chapter 2 asserts that reader interest in the story probably should be considered as a
possible intervening variable that could be used to explain user satisfaction scores. This
point, however, was overlooked in the questionnaire design.
Limitation 2: The experiment lacked a method for precise measurement of the
amount of time each participant spent reading the story. Having participants record the
times just before they opened and just after they closed the Web browser is not as precise
as having the machine record the times the browser was opened and closed. This means
that, for a participant reporting a time of 21 minutes in this study, there was no way of
telling if the participant spent 21 minutes 1 second with the treatment or 21minutes 59
seconds with the treatment. This factor warns against basing any future hypotheses about
time spent with linear stories versus time spent with nonlinear stories on the results of this
To increase precision of measurement, future experiments could employ scripts that
have the computer time-stamp the treatments used by participants. Some scripts are also
capable of tracking time spent in each node as well as the path a user takes through a
nonlinear treatment. Such scripts yield data that could be analyzed for insight into how
users read nonlinear hypertexts.
Limitation 3: Results based on a study of 18- to 25-year-old advertising students
have limited generalizability. These students were all relatively the same age, but actual
newspaper readers would represent a far larger age range and thus likely broader variance
in hypertext comfort, user satisfaction, perceived story credibility, and recall. Thus, it is
imperative that future research investigate whether age affects the dependent variables
analyzed in this study. Results showed that age had no significant effects on the dependent
variables with the exception of accuracy (F=2.590, df=7, p=.016). The sample also lacked
balance in terms of the ratio of female participants to male participants (93:32). However,
results showed that gender had no significant effects on the dependent variables. The ratio
of females to males may be less problematic, though, as research indicates that women are
increasingly in the majority of Web users (Pew Research, 1998, 2000).
Limitation 4: The experiment was conducted in different labs. As a result, 65
participants who produced usable questionnaires used the treatments on 17-inch monitors
and 60 participants who produced usable questionnaires used the treatments on 21-inch
monitors. To the credit of the study, an analysis of variance in SPSS showed that monitor
size had no significant effects on the dependent variables. Still, consistency in setting and
equipment is preferable in a controlled experiment.
Suggestions for Future Research
Future research might replicate this study by correcting its weaknesses to see both if
the results would be confirmed and whether liking of or interest in the story intervenes with
user satisfaction and/or maybe credibility. This would require that participants be asked at
least one additional question about their liking of or interest in the story. Questionnaire 2
also might ask participants to report their hometown(s) because it was possible that the
credibility and user satisfaction scores of participants from southwest Florida might have
been affected by the story's favorable representation of southwest Floridians.
In addition, scripts that can record the time spent with treatments and track user
movements through the nonlinear treatment could be employed to increase precision of the
time data and to offer insight into how users read nonlinear hypertexts. Furthermore, more
reliable scales probably should be developed and/or the adaptation of existing scales
improved to measure comfort with hypertext, user satisfaction, and perceived story
Researchers also might improve the nonlinear treatment design (compared with the
design used in this study) before replicating the experiment, perhaps by arranging the
components in a different fashion or by finding a way to eliminate the "Story Links"
anchor link at the bottom of each component; Nielsen (2000) has reported finding some
users confused by same-page links. It could be that the nonlinear treatment design in this
experiment was too basic (i.e., the manipulation was not drastic enough) to have any
influence on the dependent variables.
Future research also might include variations on the experimental design used in
this study. This study tested the effects of one story two ways (linear and nonlinear).
Potential studies could compare the effects of format using two (or more) stories. It would
be important to do this for two reasons: First, offering users a choice of story would be a
better simulation of everyday life because when users go online for news, they are
presented with choices. Second, adapting more than one story would enable researchers to
hypothesize about what kinds of stories are effective when presented in nonlinear formats.
Findings based on such hypotheses would be especially useful to online journalists.
In this study, the researcher made an educated guess that a non-breaking news
feature incorporating various characters and a multiplicity of perspectives would be
suitable for telling via nonlinear, Web-based hypertext. Presenting a news brief or a 500-
word story in nonlinear hypertext format did not appear to make sense. As for breaking
news, nonlinear hypertexts also did not seem to make sense both methodologically and
logically: By the time the components would be crafted and linked, the news would no
longer be breaking. Users, however, might prefer reading some or all these kinds of stories
in nonlinear formats: Naturally, they would turn to the outlets that provide them. It is time
to investigate which online story formats users prefer for different kinds of news stories so
that online news sites may better serve and satisfy their users.
TREATMENT SCREEN EXAMPLES
Linear treatment screen
Kilimanjaro for Kids
23 South Floridians hike to the roof of Africa to raise money for a children's hospital
By Pat Booth, Florida Times
September 30, 2000
At 18,500 feet above sea level, all that is familiar disappears.
No trees. No flowers. No animals.
No birds. No insects.
No clouds, either. When the mist clears, they hover hundreds
of feet below.
There is no sound, save the wind rushing across wide open
At 18,500 feet above sea level, the desert meets the arctic.
Gently sloping sand hills roll to the horizon and look like they
should shimmer under a blazing afternoon sun.
They don't because to the left of the sand hills, two massive
glaciers -- as tall as five-story buildings -- thrust from the
landscape. An enormous crane seemingly plucked them off a
polar ice cap and dropped them there, on top of a mountain
in Africa where they clash with sun and sand.
When darkness fell on that lunar landscape, 18 Southwest
Florida hikers burrowed into tents -- cold, ill, nervous,
r/- -1 I
Nonlinear treatment: First screen
Kilimanjaro for Kids
23 South Floridians hike to the roof of Africa to raise money for a children's hospital
Near the Peak
Home Page A
By Pat Booth, Flonda Times
September 30, 2000
The dream began last winter.
In January, Todd Kendall, 44, a board member for The
Children's Hospital of Southwest Florida, noticed Collier
County sent thousands of patients but few donations to the
south Fort Myers hospital. Few of his Naples neighbors and
friends knew the hospital existed.
Kendall wanted to compete with the usual black tie and
charity golf affairs. He needed something to shake up people
and get them talking, get them excited.
Kilimanjaro for Kids was born.
Within weeks of publicizing the idea, Kendall found himself
with a group of 23 people, mostly from Naples, who
volunteered for the fund-raising adventure.
Group members hoped the image of sun-loving, sea-level
Floridians climbing a cold steep mountain would capture the
imagination of their neighbors and open their pocketbooks.
In the months before the climb, the hikers raised $250,000
for the hospital and succeeded in grabbing the community's
lE^Blrdi a A IrfI r 14-P
Nonlinear treatment: Subsidiary screen
Kilimanjaro for Kids
23 South Floridians hike to the roof of Africa to raise money for a children's hospital I
Rtnrv I inks:
Near the Peak:
On lunar turf
Some turn back
2 hikers too ill
At 18,500 feet
One hourto go
Man slips, rocks fall
At 18,500 feet above sea level, all that is familiar disappears.
No trees. No flowers. No animals.
No birds. No insects.
No clouds, either. When the mist clears, they hover hundreds
of feet below.
There is no sound, save the wind rushing across wide open
At 18,500 feet above sea level, the desert meets the arctic.
Gently sloping sand hills roll to the horizon and look like they
should shimmer under a blazing afternoon sun.
They don't because to the left of the sand hills, two massive
glaciers -- as tall as five-story buildings -- thrust from the
landscape. An enormous crane seemingly plucked them off a
polar ice cap and dropped them there, on top of a mountain
in Africa where they clash with sun and sand.
When darkness fell on that lunar landscape, 18 Southwest
Florida hikers burrowed into tents -- cold, ill, nervous,
Back to the top of this page: Story Links
DESIGN OF THE TREATMENTS
The researcher/designer made an effort to emphasize the article text as much as
possible in the design of the treatments. The treatments had identical headlines. White
background, black text, and a sans-serif font (Verdana) were used in both treatments to
optimize readability on the screen. The treatments shared the same basic layout: A color
navigation bar appeared down the left-hand side of the screen, where links are commonly
located in online newspapers (see Appendix A for screen shots of the treatments).
In the nonlinear treatment, the color navigation bar was set to 30 percent of browser
width and contained links to story components. The researcher/designer determined story
components by reading through a printed copy of the original story as it appeared on The
News-Press.com from beginning to end. When the researcher/designer noticed a shift in
scenes or change in topic, she drew a horizontal line across the page between the
paragraphs where the shift occurred. In the end, the researcher/designer drew 21 lines; the
paragraphs of text above, below, or in between lines became the individual components.
The researcher/designer then labeled each component with a short, catchy phrase
(or label) that summarized the content or theme of the component. Components then were
grouped under four larger themes that became main story sections; one exceptional
component (the list of hiker profiles) was left to stand alone as its own main section. The
researcher/designer selected one component (an overview of the fundraiser) to be the
content of the top page (see Appendix C for a diagram of the nonlinear treatment story
structure with component labels).
Each story section was given a different color background in the colorful navigation
bar in an attempt to use color to orient the user (see screen shots of the nonlinear treatment
in Appendix A). All five main story sections were linked off the top page. The links
appeared as a list in the navigation bar. On the top page, the unlinked words "Top page"
appeared at the bottom of the link list in the navigation bar; a small graphic (a .gif file) of a
mountain was placed next to the unlinked words in an attempt to orient the user further
using a visual. When the user pointed his or her mouse-controlled arrow over the small
graphic, the words "You are here" appeared as ALT text. The literal adaptation of a "You
are here" indicator (Nielsen 2000; Spool, 1999b) also was intended to help orient the user.
The page of each story component was clearly marked with the same mountain-graphic
indicator, which appeared consistently to the right of the unlinked component label in the
color navigation bar.
The story was designed so that when the user clicked on a link to a main story
section, he or she went to a predetermined component in the story section; and a list of
links to other components in that section appeared in the story-section color block in the
color navigation bar (see subsidiary screen example in Appendix A). This method enabled
the researcher to minimize the amount of editing done on the original text so that users in
both treatments would read as much of the same text as possible.
In the linear treatment, the single-color "fake" navigation bar was set to 20% of
browser width and contained no links. The story text sequence matched the text that
appeared at the Web address http://www.news-press.com/kilimanjaro/kilimain.html in
October 2000. To increase usability and minimize effects on participants' recall scores, the
researcher inserted subheadings. Subheadings matched the linked component labels used in
the nonlinear treatment and introduced the same components, with three exceptions. The
writer of the original story started the narrative in the middle and then employed flashback
techniques to narrate the entire story. Therefore, the researcher/designer felt that it would
be inappropriate to insert the same subheading-labels for the text that represented the first
two "components" because they appeared first in the original linear story order.
Conversely, the researcher/designer added one subheading ("Attracting attention")
in the linear treatment to introduce the component that gave information about the fund-
raiser. In the nonlinear treatment, this component appeared on the top page and therefore
was labeled "Top page" in the navigation bar.
Because the article came from a Florida newspaper and the experiment was
conducted at a university in Florida, the writer's name was changed to a gender-neutral
name, and the name of the newspaper was changed to a generic name to minimize effects
for any students who possibly knew the writer and/or were familiar with the newspaper.
NONLINEAR TREATMENT STRUCTURE
H Iv i r r.
JOijur lt' y
f, 4-, ....
' !' "
4 4 1
r 9 ..
P- T -p 4 k
A ** >
4 i. 4
< C *.!*Jf
" ...-* to. ,
Instructions: Please answer the following question as honestly and as best you can.
1) List as many things as you can remember from the article you just read. List one thing
per line. You may list them in any order.
***When you are done, raise your hand to receive Questionnaire 2.***
Note: The questionnaire received by participants was formatted to fit on one side of a
single-sheet handout. The font used was 10-point Courier New.
Instructions: Please answer the following questions as honestly and as best you can.
Please complete all four pages of this questionnaire.
1) Write the exact number of hikers from South Florida who raised funds for the
children's hospital by participating in the climb:
2) Write the first and/or last names of two different hikers in the blanks below:
3) Write the occupations of two different hikers in the blanks below:
4) What phrase did the hikers say to mean "go slowly"?
5) What was the official name of the charity event?
For questions 6 through 10, circle the LETTER of your answer.
6) Where were the hikers when they first spotted Kilimanjaro's summit?
(A) In the plane
(B) Arusha National Park
(C) Didn't read that part
7) What, according to the article, is the native language of the Kilimanjaro region?
(C) Didn't read that part
8) What did the man who planned the event say about the trip overall?
(A) "The children at the hospital are going to thank us."
(B) "It exceeded all of my expectations."
(C) Didn't read that part
9) Which winter song did the hikers sing at one point?
(A) "Let it Snow"
(B) "Go Tell it on the Mountain"
(C) Didn't read that part
10) What did the hikers see in the valley just below the summit?
(A) Snow-capped trees
(B) Blue tents popping out of sand
(C) Didn't read that part
For questions 11 through 16, circle the NUMBER that corresponds to your
feelings toward the online news article you read.
11) wish all news material
were designed like this article.
12) The layout of the article
13) The article was
14) It was difficult to read
from the screen.*
15) It was difficult to understand
understand the point of this article.*
16) It was easy to read
Neutral Agree Agree
3 4 5
3 4 5
3 4 5
3 4 5
3 4 5
3 4 5
17) Did you click blue and/or purple links to read different parts of the article?
Circle one: Yes No
In questions 18 through 23, circle the NUMBER that corresponds to
your feelings toward the article you just read.
18) Is accurate
19) Is factual
20) Is biased
21) Tells the whole
5 4 3 2 1 Is inaccurate
5 4 3 2 1
5 4 3 2 1
5 4 3 2 1 Does not tell the whole
22) Does separate
fact and opinion
23) Cannot be trusted
5 4 3 2 1
5 4 3 2 1
Does not separate
fact and opinion
Can be trusted*
24) Rate your comfort with reading news articles on the computer.
Circle one: Very comfortable Comfortable Uncomfortable Very Uncomfortable
25) How often do you read on the Web (other than e-mail) during a typical week?
Circle one: Frequently Sometimes Rarely Never
26) Rate your comfort with clicking links to get information.
Circle one: Very comfortable Comfortable Uncomfortable Very Uncomfortable
27) What is your gender?
Circle one: Male
28) What is your age (in years)?
29) What is your enrollment status at UF?
Circle one: Undergraduate Graduate
30) Had you known of the news event reported in this article before today's session?
Circle one: Yes No
31) Had you read this article before today's session?
Circle one: Yes No
*Denotes items that were reverse-coded.
Note: The questionnaire participants received was fit on a four-page (single-sided)
handout. The font used was 10-point Courier New.
Note: This appendix shows the linear treatment instructions; the only difference
between the two instruction sheets for the linear treatment and the nonlinear
treatment was the usemame. Participants in the linear treatment were instructed to
log on under the username "JOU5000a"; participants in the nonlinear treatment
were instructed to log on under the username "JOU5000b."
1. Log on to the network by pressing Control + Alt + Delete and typing in the
2. Using the clock available in the lower right-hand comer or the computer screen,
mark ON THIS SHEET the time you logged on in the blank below:
Logged on at:
3. Immediately open the Microsoft Internet Explorer browser by double-clicking
on the Internet Explorer Shortcut icon on the desktop. (The icon is a blue,
lowercase "e." The icon appears as the fourth icon down from the top in the row
of icons on the left-hand side of the screen).
4. Make sure you see a story titled "Kilimanjaro for Kids." It will take most people
at least 15 minutes to read the entire story.
Read the entire story, scrolling and/or clicking as necessary to access the text. If
you're not interested in the story after 15 minutes, you can stop. If you are
interested, you can read as long as you like.
5. When you are finished reading the story, CLOSE the Internet Explorer browser.
(Note: Once you close the browser, you may not reopen it so make sure that you
are finished or have read for at least 15 minutes.)
6. Using the clock in the lower right-hand corner of the computer screen, mark the
time you finish reading ON THIS SHEET in the blank below:
Finished reading at: