Content-Type: text/html

THE EFFECTS OF THREE different COMPUTER TEXTS
ON READERS' RECALL

(BASED ON WORKING MEMORY, RISK-TAKING TENDENCIES, AND HYPERTEXT FAMILIARITY AND
KNOWLEDGE)




BY
Moon J. Lee, Ph.D. student
Mary Ann Ferguson, Ph.D.
Matthew C. Tedder, Software Developer



at the College of Journalism and Communications
The University of Florida













Moon Jeong Lee
 425 SW 2nd St. Apt. #2, Gainesville, FL 32601
Phone: (352) 375-3321 / E-Mail: <a href="/cgi-bin/wa?LOGON=A3%3Dind9909B%26L%3DAEJMC%26E%3D7BIT%26P%3D3351034%26B%3D--%26T%3DTEXT%252FPLAIN%3B%2520charset%3DUS-ASCII" >[log in to unmask]</a>






Abstract of the Paper Presented to AEJMC








THE EFFECTS OF THREE DIFFERENT COMPUTER TEXTS ON READERS' RECALL

(BASED ON WORKING MEMORY, RISK-TAKING TENDENCIES, AND HYPERTEXT FAMILIARITY AND
KNOWLEDGE)




     This study investigated the effects of three different computer texts on
readers' recall based on working memory capacity, risk-taking tendencies, and
hypertext familiarity and knowledge.  The results varied by gender.
     There was a significant text format effect on the male subjects' recall but
not on the female subjects' recall.
     The subjects' risk-taking tendencies were shown as significant factors for
the males' recall while the working memory capacity (reading span) was a
significant indicator for the females' recall.
















introduction
     Unlike paper-based materials, modern information technologies enable the
adaptation of communications to individual needs.  To improve a reader's recall
and comprehension of written information, different text formats such as the
order and detail of information presented can be modified in a manner more
beneficial for readers, individually.  At the center of these new technologies
are various forms of hypertext and hypermedia.  Hypertext is a computer-mediated
text in which highlighted words or titles serve as links to other excerpts or
documents of supporting information.  In this way, a reader can choose his/her
own order and level of detail.  Traditional text, on the other hand, is defined
as a text simply formatted from beginning to end, with no diversionary links to
other excerpts or documents.
     For the large number of ways in which hypertext is being employed, little
systematic research has been done investigating how much better or worse
information is learned from hypertext formats versus traditional text.
     In the new hypertext environment, the reader may form his/her own cohesion
and sequence of information.  Traditional text contrasts in that the writer
dictates the cohesion and sequence of information presented.  Under this
constraint, the best learning formats for individual readers cannot be
accommodated by the text.
     Many scholars hypothesize that a reader takes an active role in finding
information and encountering different types of information in a hypertext
environment (Bourne, 1990; and Dee-Lucas & Larkin, 1995).  However, there is
considerable disagreement as to whether hypertext is beneficial for all
individuals' cognitive learning processes (Gordon, 1990; Jonessen & Wang, 1990;
Spiro& Jehng, 1990; Schroeder, 1994; and Dee-Lucas & Larkin, 1995).  For
example, do any characteristics of personality impact the reader's ability to
understand and recall what he/she has read?
     Investigating the advantages and disadvantages of different hypertext
formats in relation to relevant individual characteristics is an important step
in exploring the new computer-based learning environment.  It may in fact prove
crucial in producing the best learning format for each individual.
     One of the key elements of learning or acquiring knowledge is working
memory; that is, the mind's temporary memory storage used for initial processing
of incoming information.  It is well known that different individuals have
different working memory capacities.  Therefore, it has been suggested that text
formats influence a reader's ability to recall what is read based on their
working memory capacity.  For example, researchers have hypothesized that some
computer text formats or styles may be processed more easily by readers with
different working memory capacities (Schroeder, 1994; Budd, Whitney, and Turley,
1995; Dee-Lucas and Larkin, 1995; Lorch, 1987).
     Personal behavioral characteristics such as risk-taking tendencies might
also influence a reader's recall and comprehension of different texts
differently.  For example, Donohew, Palmgreen, & Duncan (1980) postulate an
activation model of selective information processing based on their cognitive
and activation needs.  A basic assumption is that individuals have different
levels of arousal needs at which they are most comfortable.  In this aspect,
attention depends on an individual's need for stimulation by an information
source (Zillmann & Bryant, 1985).  Therefore, for high sensation seekers (high
risk-takers), information should be novel or sensational to be effective.  In
this aspect, individual risk-taking tendencies seem to have an effect on their
information processing.
     Research on individual differences may help uncover the effects of
hypertext and create more effective computer-mediated text based on an
individual reader's needs.  Considering the  different working memory capacity
of individuals and their risk-taking tendencies, relationships between text
formats and cognitive processing need to be identified in further detail.  This
research attempts to identify a relationship between working memory and
different computer text formats to support learning in a hypertext environment.
It also looks at how personal behavioral characteristics might influence
individuals' cognitive learning process.  The purpose of this study is to
investigate how text formats affect a reader's recall based on his or her
working memory span, risk-taking tendencies, and hypertext familiarity and
knowledge.  The central focus of the research is how these factors influence an
individual's recall.  Anticipated benefits and problems of hypertext will be
further addressed through literature review.
LITERATURE REVIEW
Hypertext and Cognition
     One of the challenges for instructional designers is how best to represent
information  (Schroeder, 1994).  Research on text structure, schema development,
learner control, and text enhancement techniques provide a good insight for
designing effective instructional materials to improve learning (Schroeder,
1994).
     Plude (1992) points out that selective attention can be conceived of as
executive control in the memory system.  In other words, selective attention
determines which information is processed or ignored.  Selection is one of the
necessary conditions for memory retrieval.
     Most studies have been done with traditional text, such as books, but more
recent studies address the acquisition of knowledge in terms of hypertext.  Many
scholars suggest that hypertext provides a high degree of interactivity.
Interactive learning is widely assumed to be effective because active learning
produces more effective learning outcomes.
     Traditional text presents a continuous linear flow of information.  When
readers read a text, they are required to build a mental representation of the
content in their own minds.  With traditional text, readers read from top to
bottom so that text processing is continuous.  However, hypertext seems to
interrupt continuous text processing where readers move back and forth between
the text units.  Dee-Lucas and Larkin (1995) speculate two possible effects of
this discontinuity.  First, "Interruptions in text study could interfere with
the development of an integrated representation of the text as a whole (p.
435)."  Each time when readers select a new peace of information, they have to
build a connection and incorporate with prior information.  It would be
difficult for readers to identify the main idea of an overall text.  Second,
"Interrupted text study may increase the depth of processing of content within
each unit by focusing attention on the individual unit (p. 435)."
     According to Bourne (1990), as a learner constructs a path to read the
material, he/she must make choices and become a more active learner.  This
active role of acquiring knowledge allows the learner to reconfigure his/her own
educational materials.
     To remember factual information successfully, retrieval must be in response
to some sort of cue (Farr, 1987).  Dee-Lucas and Larkin (1995) suggested that
unit titles would be better recalled than corresponding headings and subheadings
from traditional text.  They assumed that the way in which the text is
partitioned into units is highly salient to the reader because this information
guides him/her to the next choice.
     On the other hand, Jonessen and Wang (1990) argued that hypertext may
actually hinder factual recall.  Many scholars have suggested that hypertext
systems can cause disorientation in leaner and an additional cognitive burden of
remembering the links and comprehending information as a whole. (Heller, 1990;
Jonessen & Wang; Spiro & Jehng, 1990; Schroeder, 1994).  Phillips and his
colleagues (1992) examined different types of navigational devices in a
hypertext database and found that those provided with the most minimal
navigational tools achieved the highest recall.
     Researchers have shown that prior knowledge and existing schemata influence
learning.  For example, Schroeder (1994) suggested that users may have
difficulty tracking the overall structure of information in a hypertext and
relating it to their prior knowledge.  This problem may be worse for students
with low prior knowledge by causing disorientation or cognitive overload.  In
his study, those with high prior knowledge did better on most variables and
showed a greater increase in structural knowledge than those with low prior
knowledge.  One interesting finding is that treatments that provide more
structural knowledge support appear to be less influenced by the degree of prior
knowledge.  This study suggests that further research is required to clarify how
to help readers with low prior knowledge to get familiar with using hypertext.
     Schroeder (1994) pointed out that hypertext requires skills of navigating
and suggested that hypertext is not suit for highly structured learning tasks.
Hypertext readers must integrate specific content to the text as a whole.
Dee-Lucas and Larkin (1995) argued that readers may process the text units as
segment information rather than as interrelated information.  Thus, readers with
hypertext might have difficulty identifying the main points from the text as a
whole, compared with readers using traditional text.
     Maintaining coherence among information is a potential problem in using
hypertext, and some solutions have been proposed.  Dee-Lucas and Larkin's study
(1995) suggests minimally structured overviews that allow students to construct
their own organization of the information.
     Hypertext is normally integrated with other visual or auditory aids and
produces a better quality of learning.  Since hypertext is used in combination
with other media such as video, audio, and graphics, it is hard to consider
hyper-link systems themselves as hypertext applied in practice.  However, the
way hyper-link systems affect an individual's cognitive information processing
should be studied on a micro level.
     Individuals' relevant experience, previous knowledge of subjects, and
motivation are the important factors influencing recall and comprehension in
information processing.  However, how these factors actually work in the memory
process is unclear.
     Studies have explored the effects of different learning strategies in
hypertext learning environments by using learning strategies to predict simple
cognitive performance.  However, not much research has been conducted for
addressing the problems or the benefits of new hypertext systems which can adapt
to readers' individual preferences.

Working Memory and Information Processing
                One emerging concept in the information-processing literature is working
memory, a system responsible for the temporary storage and processing of
information (Baddeley, 1986).
Working memory plays a critical role in integrating information during recall
and comprehension.  Several models show a dual role of working memory:  First,
it grasps recent information from a text and connects it to related information
in long-term memory.  Second, it temporarily maintains main information for
constructing an overall understanding of the passage (Kintsch & Van Dijk, 1978;
Baddely, 1986; Lee-Sammons & Whitney, 1991).
     Individuals have different capacities for working memory.  Some researchers
have suggested that individual differences in working memory capacity predict
readers' level of text integration ability.  For example, Yuill, Oakhill, and
Parkin's study (1989) showed that readers with low working memory span have
poorer text integration abilities than readers with high working memory span.
Readers with low working memory capacity show poorer comprehension of texts than
readers with high working memory capacity because they seem less able to
maintain necessary information in an active state (Lee-Sammons & Whitney, 1991).
On the other hand, some data suggest that high-span readers are better able to
maintain multiple tentative interpretations and use text elements to test those
interpretations (Baddeley, Logie, Nimmo-Smith, & Brereton, 1985; Lee-Sammons &
Whitney, 1991).
     Lee-Sammons and Whitney (1991) examined the effect of working memory span
and readers' perspectives on comprehension of a narrative text.  In their study,
subjects were told to read the story from different perspectives (e.g. either
from the perspective of a potential home buyer or a potential burglar) and to
think about the relevance of each sentence from that perspective.  One
interesting finding is that subjects recalled more information relevant to the
new perspective than information relevant to the original encoding perspective.
For low and medium-span readers, shifting perspectives resulted in less recall
of new information than if the perspective was held constant while high-span
readers recalled information independent from shifting perspective.
Furthermore, low-span readers were poor at recalling information not relevant to
the original encoding perspective while high-span readers recalled similar
amounts of perspective-relevant and irrelevant information.
        The studies for measuring memory span were mostly done with a test of storage
capacities for unfamiliar and unrelated information.  These kind of a test
didn't seem to indicate the capacity of working memory available during reading.
Daneman and Carpenter (1980) designed a task to measure the capacity of working
memory during reading.  They presented subjects with a series of unrelated
sentences that they needed to comprehend to answer following text questions. At
the end of the presentation, subjects were asked to recall as many of the last
words of the sentences as possible.  The number of words a subject correctly
recalled became the subjects' reading span.  They found that the subjects'
active efforts to encode associations between the last words of the sentences
resulted in better recall.
        A well designed computer text requires less working memory by the way in which
the text is presented (Schroeder, 1994; Budd, Whitney, and Turley, 1995;
Dee-Lucas and Larkin, 1995; Lorch and Lorch; 1996).  Unfortunately, few computer
materials have been designed in consideration with this individual different
capacity.

Working Memory Capacity and Text Format
     In most theories of cognitive tasks, it is agreed that construction of
referential representation in memory for the interpretation of the text is
necessary.  The 'structural hypothesis' has been developed and tested using
different types of representations (e.g. different text formats) and cognitive
processes (e.g. recall or comprehension) (Nakamura, Kleiber, & Kim, 1992).  The
structural hypothesis suggests that recall depends on the memory strength (e.g.
high and low) of the interconnections and the number of types of
interconnections (e.g. vertical and horizontal) within a representation.
Readers who use a structural strategy to understand a text recall more top-level
information than do readers who process text in a linear manner.
     According to Budd, Whitney, and Turley (1995), in order to establish
relations between closed clauses or sentences, readers have to maintain the
information most recently processed from the text in working memory.  Thus,
working memory facilitates the comprehension of text by building coherence at a
local and general level.  In this aspect, if texts are coherent at a more
general or thematic level, readers will require less working memory capacity to
process new information.
     Individual differences in working memory capacity seem to be related to
differences in what kind of information, or how much information, is retained
when one is reading texts (Baddeley, 1986; Budd, Whitney, & Turley, 1995, and
Lee-Sammon & Whitney, 1995).  Reading is much easier if there is a clear concept
stated at the beginning of the passage because readers need only adopt and test
it against the remainder of the passage.  Otherwise, readers have to construct
one while they are reading.  Furthermore, readers are required to modify it when
they confront new information inconsistent with their initial interpretation
(Kieras, 1981).  These activities require more working memory to process
information.
     Budd, Whitney, and Turley (1995) investigated whether individuals'
different working memory capacity is related to different working memory
management strategies in reading an expository text.  One interesting finding
was that readers consider information displayed first as more important than
information displayed later.  In sum, when materials were easy, the performance
differences of readers with different working memory spans were small or
insignificant.  However, these differences got larger when materials were
difficult to understand.   Lower span subjects performed more poorly on a
detailed question in the topic-absent (where there is no heading) condition.
These data imply that certain text or passage formats can detriment or help
readers' recall of more detail information from a text.

Theory of Information Processing and Risk-taking

        Scholars have suggested that an individual's information processing depends
upon needs for novelty and sensation (Bardo, Donohew, & Harrington, 1996;
Zuckerman, 1994).  Needs for novelty and sensation seem to affect the initial
process of information.  Zuckerman, Persky, Hopkins, Murtaugh, Basu, & Schilling
(1966) suggested that risk takers may exhibit a higher need for arousal than
non-risk takers.
     Donohew et al. (1998) suggested that messages high in sensation value
should be more attractive to high sensation seekers.  Basically, it was
suggested that the effectiveness of messages depends on the characteristics of
target audiences such as whether they are high sensational seekers or not.
     According to the laboratory experiments of Donohew et al. (1991) and
Palmgreen et al. (1991), high sensation seekers (HSS) responded more positively
to the messages with the high level of suspense and tension than did the low
sensation seekers (LSS).   Also, HSSs tend to pay more attention to "high
sensation value programming" than to "low sensation value programming."
Therefore, the sensation value of information is an important determinant of a
message effect, particularly for HSSs even though the results indicate that many
LSSs are also affected by the same messages as much as HSSs.  Certainly, this
shows evidence that the characteristics of target audiences such as risk-taking
tendencies may produce different results in readers' learning when they read
information through different computer text formats. The results of these kinds
of studies will provide important direction for the design of effective
instructional materials with regard to the critical issues of effective computer
text design and placement.
        Being predisposed to risk-taking was a concept largely researched by Marvin
Zuckerman.  A basic assumption regarding different risk-taking tendencies is
that individuals have different degrees of predisposition toward sense-arousing
stimuli (Zuckerman, Kolin, Price, & Zoob, 1964).
     It was known that high sensation seekers tend to prefer visual complexity
(Zuckerman, Bone, Neary, Mangelsdorff, & Brustman, 1972) and spend more time
listening to music, attending movies (X-rated movies, in many cases), and
reading fictional novels (Ferguson, Valenti, & Melwani, 1991).
        Ferguson, Valenti, & Melwani (1991) postulate three types of risk-taking
tendencies:  impulsiveness, adventurousness, and rebelliousness.  Impulsiveness
represents reports of behaving without thought and getting "carried away."
Adventurousness represents self-reports of enjoyment of risk and new and
exciting experiences while rebelliousness represents self-reports of enjoyment
of wild parties, drinking, sex, and drug use (Ferguson, Valenti, & Melwani,
1991).
Impulsiveness
     Impulsiveness is associated with a dislike of thinking.  Those predisposed
to impulsive risk-taking score low on cognitive involvement with health, have
negative feelings about health, do not feel in control of their health, and have
little concern about their health.  Impulsiveness is a construct associated with
many types of risk takers (Eysenck & Eysenck, 1977).  Impulsive risk takers also
tend to be "young, single, female, or smokers" (Ferguson, Valenti, & Melwani,
1991, p. 219).
Adventurousness
        Adventurousness can be defined as "the desire to try new, exciting activities."
(Moore & Rosenthal, 1993, p. 98)  This is a risk-taking predisposition that
correlates positively with good attitudes about health, a feeling of control
over one's health, and strong health values. This would follow that adventurous
types would be less likely to take up and more likely to quit smoking than those
classified as impulsive.  Ferguson, Valenti, & Melwani (1991) also describe
adventurous types as "those who are young, single, male, or who never or only
occasionally attend church" (p. 218).  Additionally, both of the above studies
found driving fast was associated with these types of risk taking.
Rebelliousness
        From previous research (Ferguson, Valenti, & Melwani, 1991), rebellious
personality types were shown to be largely young, single, and male.  Rebellious
risk-taking is associated with high radio use and reliance on radio, but a low
reliance on all media for health information.  Those who are smokers and
ex-smokers have also tended to score high in this area (Jex & Lombard, 1998;
Lipkus, Barefoot, Williams, and Siegler, 1994).  Rebels were shown to not be
particularly religious.  Rebellious personality types also have been shown to
take risks not for perceived benefits, but rather for notoriety among others for
being rebellious or daring.  Rebellious adolescents also tend to respond to the
sensational aspects of a message rather than its perceived risks (Moore, 1996;
Donohew, Lorch, & Palmgreen, 1998).
     Based on these risk-taking constructs, target audience and style of text
presentations themselves are expected to interact with the subjects' risk-taking
tendencies.
     Based on the literature review, the following research questions were
investigated in this study.

RESEARCH QUESTIONS
     RQ1 - How do the different computer text formats influence the subjects'
recall when subjects' working memory capacity, risk-taking tendencies, hypertext
familiarity and knowledge, and total reading time are statically controlled?
     RQ2 - How does individual working memory capacity have an effect on the
subjects' recall?
     RQ3 - How do individual risk-taking tendencies influence readers' recall?
If significant effects are found, which text formats are better or worse for
which risk-takers?
     RQ4 - How does hypertext familiarity and knowledge influence subjects'
recall?
All these four research questions were evaluated based on gender.

Experimental Methodology
     The experimental design was a post-test only model with three variations of
computer text formats.  The manipulation involved reading an article in one of
the three formats: traditional text, structured hypertext, and networked
hypertext.  After subjects finished reading, they were asked to take the Reading
Span test, then the risk-taking tendency tests, and last the factual recall
tests[1].  The subjects also answered demographic questions.
     In the experiment, each subject was randomly assigned to one of three
groups, corresponding to the three text formats.  The subjects' working memory
span was measured by the Reading Span tests.

Developing the Stimulus Material and Measuring the Dependent Variable
     After seven initial articles were tested, a topic was selected based on the
level of interest and gender neutrality in the article.  The article selected
was, "The battle against segregation in America,"[2] being determined the most
interesting for both males and females.
     The development of the recall test for this article involved the following
procedure.  Twenty-two multiple-choice questions were written for a factual
recall test with approximately six questions for each paragraph of the article.
Sixteen graduate students at the University of Florida, participated in the
pretest to determine which items to use.  T-tests for the subjects' interest
level and item difficulty scores were computed.  Sixteen multiple-choice
questions were adopted and 6 questions were dropped from the recall test[3].

Study Design
     Subjects were randomly assigned to one of the three conditions:
traditional-text, structured hypertext, or networked hypertext.  The researcher
explained the test procedures at the beginning of the study and subjects were
told to read the text completely and not to go back to the readings after they
began the recall test.  Computers were preloaded with a letter from the
researcher explaining the study and a choice, which randomly led subjects to one
of three different conditions.  How much time each subject spent on each screen
of the article as well as which links he/she clicked was collected by the
computer program presenting the article.  The experiment was conducted in a
computer lab at the University of Florida.
     In condition 1, the article was presented in traditional text format with a
link captioned, "turn to the next page", to simulate the pages in a paper-based
book. The subjects were to read the article in traditional text format, from top
to bottom.  After the subjects finished reading, they clicked on the button
"Finished."  After clicking on this button, subjects were automatically taken to
the working memory tests, followed by the risk-taking tendency tests, followed
by the recall tests, and ending with the demographic questions.  A "Back" button
was available enabling the reader to return to the previous page while reading.
      In condition 2, the article was presented in a structured-hypertext
format.  The structured-hypertext consisted of each sub-topic on a different
screen, linking hierarchically by highlighted, underlined key terms in the text.
The reader began reading at the top level and was able to click on links for
several levels of sub-topical information.  As in condition 1, a "Back" button
was available enabling the reader to return to the higher levels in the
hierarchy.
     In condition 3, the article was presented in a networked-hypertext format.
The text was presented with each excerpt on its own screen.  The user was able
to navigate between excerpts by clicking on highlighted, underlined terms,
serving as links.  No structure, such as hierarchical, was used between links.
Instead, wherever a term existed in the article for which an elaborating excerpt
existed, a link was presented.  This created a web-like, user navigated
architecture of links where looping was possible.  As in conditions 1 and 2, a
"Back" button was available for the reader to return to previous screens, as
would be found in a web-browser or Windows Help file.
     After subjects finished reading, each subject was given three reading span
tests, the risk-taking tendency tests and the recall tests.  Fourteen questions
were given to subjects regarding demographic information, computer usage and
familiarity, affinity for history, attitude toward computers and the degree of
interest in and experience with hypertext.  Each subject took from 30 to 45
minutes for the whole process.
     Ninety-six undergraduates at the University of Florida were tested.
Twenty-two sessions were held.  Subjects were compensated with extra credit in
their college courses.




ANALYSIS AND FINDINGS
     The purpose of the analysis was to find out whether presentation of the
same information in different text formats affected subjects' recall scores and
if these scores varied by the subjects' working memory capacity, risk-taking
tendencies, and gender.  In addition, data were collected to describe subjects'
race, age, computer ownership and use, Internet access, World Wide Web site
viewing time, and familiarity with hypertext.  GLM analysis and linear
regression were used to test the effects of the different text formats with
gender, working memory span, risk taking tendencies, hypertext familiarity and
knowledge, and subjects' total reading time.

Recall Score
     Of the 16 recall measuring questions, the minimum number correct was 3 and
the maximum correct was 15 with a mean of 8.4 and a standard deviation of
2.9.[4]

Working Memory
      The Reading Span test was given three times.  Each test used a black
computer screen with unrelated sentences centered in large yellow type.  The
sentences were presented for approximately four seconds each, followed by
prompts for the subject to remember and enter the last word of each sentence.
Each test began in this manner with one sentence, then two, three, four, and
five sentences.  Each series of sentence displays and questions were followed by
a question about content to assure the subject read the complete sentence.  The
results of the first test (first sentence) were discarded due to exceptionally
low scores.




Measuring Risk-Taking Tendencies
     The risk-taking scale created by Ferguson & Valenti (1989, 1990) and
Ferguson, Valenti, & Melwani (1991) was used to measure subjects' risk-taking
tendencies.  In their research program, they classify risk takers by three
different attributes:
1.      Rebelliousness:
     This risk taker often takes risks as a way of breaking social norms, rules,
     and laws and prides himself/herself on being known as a rebel.
2.      Impulsiveness:
     This risk taker takes risks on the spur-of-the-moment and he/she enjoys the
     excitement associated with being spontaneously risky.
3.      Adventurousness:
     While this risk taker takes risks of a physical nature that may put his/her
     body in danger of physical harm or even death, he or she plans carefully
and
     trains actively to be prepared for the dangers.
     Based on a factor analysis, a Principle Component Analysis with a Varimax
Rotation, nine items were used to measure adventurousness[5], seven items
measured rebelliousness[6], and seven items measured impulsiveness[7].
Computer Use and Familiarity
     Seventy-seven percent of the subjects owned a computer and 99% (N = 95)
said they were able to use computers either at home or at work/school or both
while 1% (N = 1) responded "I don't use computers at all."  Also, 98% of
subjects have access to the Internet.  Fifty percent (N = 48) of the subjects
answered that the average time they spend on the Internet per day was less than
30 minutes, 24% (N = 23) more than 30 minutes to one hour, 12% (N = 11) between
1 and 2 hours, 6% (N = 6) over 2 hours while 8% (N=8) of the subjects said that
they didn't look at the Internet at all.
     For the question "What do you use a computer for?," 36% (N = 35) answered
for doing homework, 25% (N=24) for getting information, and 29% (N=28) for
chatting or e-mail.
     In terms of familiarity with the term "hypertext," 10% (N = 10) of the
subjects said they were 'very familiar' while 28% (N = 27) answered ' very
unfamiliar.'  Some 26% (N = 25) were 'somewhat familiar,' 9% (N = 9) were
'familiar,' and 26% of subjects answered 'somewhat unfamiliar' (N = 25).
     A question about their highest level of hypertext knowledge was asked.
Table 1 shows the proportion of subjects who selected each level.  Only about
14% said they could create hypertext documents.
Table 1 Level of Hypertext Use
" What is your highest level of hypertext knowledge?"
% of answering 'Yes'
I can create advanced hypertext documents.
      2%
I can create basic hypertext documents.
    12%
I know how to use hypertext, but I cannot create hypertext documents.
    20%
I have heard the term hypertext, but I don't know how to use it.
      7%
I don't know of hypertext at all.
     58%
    Total:    100%  (N=96)

     The subjects also were asked to rate their ease of reading from a computer
screen.  'Very comfortable' was the response of 31% (N = 30), 'somewhat
comfortable' was indicated by 31% (N = 30), 'neither comfortable nor
uncomfortable' by 15 % (N = 15), 'uncomfortable' by 4% (N = 4), and 'very
uncomfortable' was answered by 17% (N = 17).  In sum, 62% (N = 60) of the
subjects were fairly comfortable with reading text from a computer screen while
21% (N = 21) answered 'uncomfortable' or 'very uncomfortable.'
     When the question was asked, "What was the basis upon which you normally
decided to click on hyperlinks in the text?," 41% (N = 39) answered "to find
more information.  19% (N=18) answered "because a link was there" and 19% (N=18)
of the subjects said "pure curiosity."

Hypertext Familiarity and Knowledge
     To create an index to measure the use of computers and hypertext, a
Principal Axis, Varimax Rotation, factor analysis of subjects' level of
hypertext knowledge, familiarity with the term "hypertext," ease of using
hypertext, access to the Internet, time spent viewing Web sites, ease of reading
hypertext documents, comfort with a computer screen, computer use, and ownership
of a computer (9 total items) was conducted.[8]   Knowledge level and
familiarity with hypertext, ease of using hypertext, and the daily average time
on the Internet, loaded on factor 1.  The ease of reading hypertext and comfort
reading from a computer screen loaded on factor 2.[9]  The standardized item
alpha of the three Hypertext variables from on factor 1, was .73.  A hypertext
familiarity and knowledge index was created from a summed-index of factor
scores.

Demographic Information
     Of the 96 participants, 68% (N = 65) of subjects were white, 19% (N = 18)
were Hispanic American, 4% (N = 4) were Asian or Pacific Islander, 2% (N = 2)
were African American.  Ninety two percent of the subjects were between 19 and
22 years old.  66% were female (N = 63) and 34% were male (N =33).
HYPOTHESIS TESTS
The Effects of Text Formats: Condition in Interaction with Gender
     A GLM model was used to test whether computer text formats (traditional
text, structured hypertext, and networked hypertext) interacted with gender to
affect the number of correct answers on the recall test.  Other variables were
submitted as covariates[10].
     Text format had a significant effect on recall scores [F (2,95) = 3.4.6, P
< .04].  Also, there was a significant gender effect on recall scores [F (1,95)
= 2.9, P < .09].  Also, there was a near significant gender by condition
interaction [F (2,95) = 2.6, P < .08].
     For males, there was a significant text format effect on recall scores [F
(2,31) = 4.8, P < .018] and further analysis revealed that there is a
significant difference between the recall scores of the structured hypertext and
the networked hypertext with .011 significant level.  The mean score of the male
subjects who read the networked hypertext is 10.7 while the mean score of the
male subjects who read the structured hypertext is 7.8.  However, for the female
subjects, the condition effect was not significant [F (2,59) = .69, P < .51]
(Figure 1).

Figure 1.  Means for Recall by Condition and Gender

















Risk-Taking Tendencies, Working Memory Capacity, and Other

Factors

        A multiregression model was used to test whether the reading span, the
risk-taking tendencies, hypertext familiarity and knowledge, and the total
reading time had an effect on the total recall.
        For the male subjects, the model was shown as significant [F (6, 32) = 4.9, P <
.01] with R2 = .53.  The results showed that the more the subjects are
adventurous (Beta = .42, P < .01), the higher their recall scores are.  On the
other hand, the more the subjects are rebellious (Beta = -.34, P < .02), the
lower their recall scores are.  Also, the hypertext familiarity and knowledge as
well as the total reading time were the significant indicators on the male
subjects' total recall.  In other words, the more time they had to read (Beta =
.46, P < .01), the higher recall scores were, while the higher they scored in
hypertext familiarity and knowledge (Beta = -.3, P < .05), the lower the recall
scores were.  For the female subjects, the model was shown as significant [F(6,
56) = 5.74, P < .01] with R2 = .38.  The total reading time, the reading span,
and hypertext familiarity and knowledge were the only significant indicators for
the female subjects' total recall.  The more time they had to read (Beta = .48,
P < .01), the higher recall scores were.  Also, the higher their reading span
scores are, the higher recall scores were (Beta = 2.7, P < .01) while the higher
they scored in hypertext familiarity and knowledge (Beta = -.17, P < .10), the
lower the recall scores were.
          The subjects' impulsiveness (r=-.27, P < .008) and rebelliousness (r=-.24, P
< .017) were negatively correlated with the working memory accuracy test.
Gender had a significant effect on the hypertext familiarity and knowledge [t(1,
55) = 1.64, P < .05], the adventurousness [t(1, 94) = 2.86, P < .005], and the
rebelliousness [t(1, 94) = 3.14, P < .005].  The male subjects (M = .23) had
higher hypertext familiarity and knowledge than the female subjects (M = -.12),
were more adventurous (M = .39) than the female subjects were (M = -.20), and
were more rebellious (M  = .42) than the female subjects were (M = -.22).
        The subjects' total recall scores had a moderate correlation with total reading
time (r=.47, P < .001) and a weaker one with the Reading Span (r=.24, P < .01).
The higher the subjects' reading span was the better the subjects' recall scores
were.  Also, the total reading time was affected by impulsiveness [F(1, 91) =
3.11, P < .08] and text format conditions [F(2, 91) = 2.89, P < .06].  The more
the subjects were impulsive, the less amount of time they read.  The subjects
who read the networked hypertext spent more time reading than those who read the
traditional text.
SUMMARY
     In this study, the effects of reading from three different computer text
formats varied by gender.  For males, the subjects' recall scores varied by
reading the different computer text formats.  The difference was found between
the recall scores of the structured hypertext and the networked hypertext.  The
male subjects who read the networked hypertext had much higher recall scores
than those reading identical information from the structured hypertext
documents.  However, the text format effect was not shown as significant for the
female subjects when other variables are controlled statistically.
        The subjects' working memory span affected their recall scores.  Again, it
varied by gender.  For the male subjects, no working memory effect was found
while the female subjects' working memory capacity had a significant effect on
recall scores.
        Furthermore, when the risk-taking tendencies, hypertext familiarity and
knowledge, and the total reading time were tested in a multiregression model,
the subjects' adventurousness as well as rebelliousness were shown as the
significant predictors for the male subjects' recall.  The more the subjects are
adventurous, the higher their recall scores are.  On the other hand, the more
the subjects are rebellious, the lower their recall scores are.  Also, the
hypertext familiarity and knowledge as well as the total reading time were shown
as significant on the male subjects' total recall.  The more time they spend
reading, the better recall scores are, while the higher they score in hypertext
familiarity and knowledge, the lower the recall scores are.  For the female
subjects, the total reading time and working memory capacity are the only
significant indicators for predicting their recall scores.      In addition, the
total reading time was affected by impulsiveness and text formats.  The more the
subjects were impulsive, the less time they spent reading.  The subjects who
read the networked hypertext spent more time reading than those who read the
traditional text, when other variables were statistically controlled.
        The subjects' impulsiveness and rebelliousness were negatively correlated with
the working memory accuracy test.  Gender had a significant effect on the
hypertext familiarity and knowledge, the adventurousness, and the
rebelliousness.  The male subjects had higher scores in the hypertext
familiarity and knowledge than the female subjects.  The male subjects were more
adventurous than the female subjects were and the male subjects were more
rebellious than the female subjects were.
     Recall seems to be affected by the structure in which the information was
presented and the subjects' different working memory capacities as well as their
risk-taking tendencies.  Also, these effects varied by their gender.
LIMITATIONS AND DISCUSSION
     This study had several limitations that should be considered for
interpretation of the results and suggestions of further studies.  One
limitation of this study was the sample size.  Due to the small male sample size
(N=33), even though other variables were tested as covariates, a careful
interpretation should be applied to the results.  The generalization of the
results should be cautiously presumed.  Further research is required with more
subjects for more accurate results.
     The findings of this study reveal that the structure and presentation of
text influence how well information is recalled.  The study results indicate
that different kinds of hypertext systems may have different effects on an
individual's information processing.  Many scholars have suggested that
hypertext systems can overload cognitive processing capabilities and encourage
over-simplification of the information read.  However, these conclusions are
premature since individuals' characteristics such as gender, working memory
capacity, and risk-taking tendencies, seem to influence subjects' information
processing.
     One of the interesting findings was that the hypertext familiarity and
knowledge had a negative influence on the subjects' recall.  This might be
because readers with high familiarity and knowledge of hypertext don't pay as
much attention to the detail information as do readers who are new in the
hypertext environments.  Further research is necessary to find out whether this
is the case and what is really going on in terms of hypertext familiarity and
knowledge on readers' information processing.
     Some researchers assert the positive role of hypertext in learning, based
on the assumption that hypertext provides selective attention or reader control.
Many educators believe that selective attention determines which information is
processed or ignored; therefore, it accelerates information processing.  It is
worth examining "selective attention," in terms of hypertext use, more closely
to evaluate its implications for the future regarding factual recall of all
presented information and recall of information that interests to subjects.
     Considering individuals' working memory differences that exist in hypertext
learning materials, the findings in this research may give clues for further
research.  However, because the study contains several limitations regarding
sample size, more careful examination of the working memory effect on readers'
recall is highly recommendable.  This avenue, too, is worth exploring.
     Also, if these different approaches to learning are subject to differential
responses based on working memory capacity, risk-taking tendencies, and gender,
isolated and simplistic conclusions from the studies of individual differences,
which assert that high working memory people are simply physiologically or
biologically superior, hinder our understanding of human information processing.
For this reason, a careful examination of individuals' memory capacity and
computer text learning patterns will help researchers understand the complex
relationships among individuals' abilities, needs, and experiences as well as
the nature of learning in hypertext environments.





























REFERENCES


Baddeley, A. D. (1986). Working memory. New York. Oxford University Press.

Baddeley, A. D., Logie, R., Nimmo-Smith, I., & Brereton, N. (1985).  Components
   of fluent reading.  Journal of Memory and Language, 24, 119-131.

Baillet, S. D., & Keenan, J. W. (1986).  The role of encoding and retrieval
   processes in the recall of text.  Discourse Processes, 9, 247-268.

Bardo, M.T., Donohew, R.L. & Harrington, N.G. (1996).  Psychobiology
   of novelty seeking and drug seeking behavior.  Behavioral Brain Research, 77
   (1), 23-43.

Bourne, D. E. (1990).  Computer-assisted instruction, learning theory, and
   hypermedia: An associative linkage.  Research Strategies, 8 (4), 160-171.

Budd, D., Whitney, P., & Turley, K. J.  (1995).  Individual  differences in
   working memory strategies for reading expository text.   Memory and
Cognition, 23
    (6),
        735-748.

Dee-Lucas, D., & Larkin, J. H. (1995).  Learning from electronic texts:  Effects
   of interactive overviews for information access.  Cognition And Instruction,
13
   (3),  431-468.

Donohew, L., Lorch, E.P., & Palmgreen, P. (1998).  Applications of a theoretic
     model of   information exposure to health interventions.  Human
Communication
     Research, 24 (3), 454-468.

Donohew, L., Lorch, E.P., & Palmgreen, P. (1991).  Sensation seeking and
targeting of televised anti-
     drug PSAs.  In L. Donohew, H.E. Sypher, & W.J. Bukoski (Eds.), Persuasive
     communication and drug use prevention, (pp. 209-226).  Hillsdale, NJ:
Lawrence
     Erlbaum Associates.

Donohew, L., Palmgreen, P., & Duncan, J. (1980).  An activation model
     of information exposure. Communication Monographs, 47, 295-303.

Eysenck, S. B. G., & Eysenck, H. J. (1977).  The place of
     impulsiveness in a dimensional system of personality description.  British
     Journal of Social and Clinical Psychology, 16, 57-68.

Ferguson, M.A., Valenti, J.M., & Melwani, G. (1991).  Communicating
   with risk takers: a public relations  perspective.  Public Relations Research
   Annual, 3, 195-224.

Farr, M. J. (1987).  The long-term retention of knowledge and skills:  A
   cognitive and instructional perspective.  New York:  Springer-Verlag.

Heller, R.S. (1990).  The role of hypermedia in education: A look at the
   research issues.  Journal of Research on Computing in Education, 22, 421-441.

Herrman, D. J., & Searleman, A. (1992).  Memory improvement and memory theory in
   historical perspective.  In Memory improvement:  Implications for memory
theory
   .  (pp. 8-20).  New York:  Springer-Verlag.


Jex, T.T. & Lombard, T. (1998). Psychosocial Factors Associated with
     Smoking in Air Force Recruits. Military Medicine, 163 (4), 222-     225.
Jonassen, D. H., & Wang, S. (1990, October).  Acquiring structural knowledge
   from semantically structured hypertext.  Paper presented at the annual
meeting
   of the Association for the Development of Computer- Based Instructional
Systems,
   San Diego, CA.

Jonassen, D. H., & Wang, S. (1992, February).  Acquiring structural knowledge
   from semantically structured hypertext.  Paper presented at the annual
meeting
   of the Association for Educational Communications and Technology, Washington.
   DC.

Jonassen, D. H., & Wang, S. (1993).  Acquiring structural knowledge from
   semantically structured hypertext.  Journal of Computer-Based Instruction, 20
   (1), 1-8.

Kieras, D. E. (1981).  Component processes in the comprehension of simple prose.
   Journal of Verbal Learning and Verbal Behavior, 20, 1-23.

Kintsch, W., & Van Dijk, T. A. (1978).  Toward a model of text comprehension and
   production.  Psychologyl Review, 85, 363-394.

Lipkus, I.M., Barefoot, J.C., Williams, R.B., & Siegler, I.C. (1994).
     Personality measures as predictors of smoking initiation and cessation in
     the UNC alumni heart study.  Health Psychology, 13 (2), 149-155.

Lee-Sammons, W., & Whitney, P. (1991).  Reading perspectives and memory for
   text:  An individual differences analysis.  Journal of Experimental
Psychology:
   Learning, Memory, and Cognition,  17, 1074-1081.

Lorch, R. F. Jr. (1989).  Text-signaling devices and their effects on reading
   and memory processes.  Educational Psychology Review, 1, 209-234.

Mayer, R. E. (1984).  Aids to text comprehension.  Educational Psychologist, 19
   (1), 30-42.

Moore, S., & Gullone, E. (1996).  Predicting adolescent risk behavior
     using a personalized cost-benefit analysis.  Journal of Youth and
     Adolescence, 25 (3), 343-359.

Nakamura, G. V. (1994).  Scene schemata in memory for spatial relations.
   American Journal of Psychology, 107 (4), 481-497.


Palmgreen, P., Donohew, L., Lorch, E.P., & Rogus, M. (1991).
   Sensation seeking, message sensation value,  and drug use as mediators of PSA
   effectiveness.  Health Communication, 3 (4), 217-227.

Plude, D. J. (1992).  Attention and memory improvement.  In D. J. Herrman., H.
   Weingartner., A. Searleman., & C. McEvoy (Eds.), Memory improvement:
   implications for memory theory (pp. 150-168).  New York:  Springer-Verlag.

Schroeder, E. E. (1994, February 16-20).  Navigating through hypertext:
   Navigational technique, individual differences, and learning.  Paper
presented
   at the 1994 National Convention of the Association for Educational
   Communications and Technology Sponsored  by the Research and Theory Division,
   Nashville, TN.

Spiro, R. J., & Jehng, J. C. (1990).  Cognitive flexibility and hypertext:
   Theory and technology for the nonlinear and multidimensional traversal of
   compels subject matter. In D. Nix, & R. J. Spiro (Eds.), Cognition,
education,
   and multimedia:  Exploring ideas in high technology (pp.).  Hillsdale, NJ:
   Lawrence Erlbaum.

Wyer, R. S., &  Srull, T. K. (1986).  Human cognition in its social context.
   Psychological Review, 93, 322-359.

Yuill, N., Oakhill, J., & Parkin, A. (1989).  Working memory, comprehension
   ability and the resolution of text anomaly.  British Journal of Psychology,
80,
   351- 361.

Zillmann, D. & Bryant, J. (1985).  Selective Exposure to
     Communication.  Hillsdale, NJ:  Lawrence Erlbaum Associates.

Zuckerman, M. (1994). Behavioral Expression and Biosocial Bases of
     Sensation Seeking.  New York: Cambridge University Press.

Zuckerman, M. (1988).  Behavior and biology: Research on sensation
     seeking and reactions to the media. In L. Donohew, H.E. Sypher, & E.T.
     Higgins (Eds.), Communication, social cognition, and affect (pp. 173-194).
     Hillsdale, NJ:  Lawrence Erlbaum Associates.

Zuckerman, M., Kolin, E. A., Price, L., & Zoob, I. (1964). Development
     of a sensation seeking scale. Journal of Consulting Psychology, 28,
477-482.

Zuckerman, M., Persky, H., Hopkins, T.R., Murtaugh, T., Basu, G.K., &
     Schilling, M. (1966). Comparison of stress effects of perceptual and social
     isolation.  Archives of General Psychiatry, 14, 356-365.

[1]  In this experiment, the reading span tests and the risk-taking tendency
tests were given after the stimulus because it was considered that the stimulus
manipulation didn't affect the results of the reading span tests and the
risk-taking tendency test tests.
[2]  Topic 4:  "The battle against segregation in America" from How to prepare
for the GRE, P521

[3]  This is based upon the result of an analysis.  Items with Item
Discrimination scores greater than .50 were kept.  Also, T-tests for males vs.
females and item difficulty scores were computed.
[4]   The reliability of these 16 items was tested.  The Cronbach Alpha score
was .61.
[5]  Items are the following: "I welcome new and exciting experiences, even if
they are a little frightening," "I like adventure," " I like challenges," "I do
or would enjoy the sensation of skiing very fast down a high mountain slope," "
I enjoy or would enjoy skydiving," "I enjoy or would enjoy scuba diving," " Life
with no danger in it would be too dull for me," "I would like to drive or ride
on a motorcycle," and " I sometimes take different routes to a place I often go,
just for variety's sake."  The Cronbach Alpha score of these nine items was .87.
[6]  Items are the following: "I like wild and uninhibited parties," "I am
likely to drive after I have had several drinks, "Having lots of alcohol is the
key to a really good party," I like people who are partiers," I'm likely to do
drugs when I want to party," "I am at a party with other people.  Someone lights
a joint of marijuana and begins to pass it around the party.  A lot of people
are trying it.  I will try it," and the recoded item "I would make quite sure I
had another job before giving up my old one." The Cronbach Alpha score was .82.
[7]  Items are the following: "I often do things on the spur of the moment," "I
often get into a jam because I do thins without thinking," "I mostly speak
before thinking things out," "I get so "carried away" by new and exciting ideas,
I never think of possible snags," "I generally do and say things without
stopping to think," " I sometimes take different routes to a place I often go,
just for variety's sake," and " I like to explore a strange city or section of
town by myself, even if it means getting lost."  The Cronbach Alpha score was
.84
[8]  The Eigenvalue and the Scree Plot indicated a two-factor solution.  Factor
1 explained 34% of the variance and factor 2 explained 24% of the variance.
[9]  For the factor2, the Cronbach's alpha was too low to create an index.
[10]  Due to the small sample size, other variables such as working memory span,
risk-taking tendencies, total reading time and hypertext familiarity and
knowledge were tested as covariates.