Professional Documents
Culture Documents
net/publication/220465640
CITATIONS READS
6 67
2 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Tamara Sumner on 01 September 2015.
Kirsten R. Butcher
Department of Educational Psychology
1705 E Campus Center Drive
University of Utah
Salt Lake City, UT 84112 USA
kirsten.butcher@utah.edu
Tamara Sumner
Institute of Cognitive Science
Center for Innovation and Creativity
1777 Exposition Drive, Campus Box 594
University of Colorado at Boulder
Boulder, CO 80309 USA
sumner@colorado.edu
Author Note
The authors wish to thank Nicole Turnidge-Halvorson and Shaw Ketels for their
contributions to this research. This article is based upon work supported in part by the National
Psychology, 1705 E Campus Center Drive, University of Utah, Salt Lake City, UT 84112.
PRIOR KNOWLEDGE AND ONLINE LEARNING 2
Abstract
This study explored the impact of prior domain knowledge on students’ strategies and use of
digital resources during a Web-based learning task. Domain knowledge was measured using pre-
and posttests of factual knowledge and knowledge application. Students utilized an age- and
topic-relevant collection of 796 Web resources drawn from an existing educational digital library
to revise essays that they had written prior to the online learning task. Following essay revision,
participants self-reported their strategies for improving their essays. Screen-capture software was
used to record all student interactions with Web-based resources and all modifications to their
essays. Analyses examined the relationship between different levels of students’ prior knowledge
and online learning behaviors, self-reported strategies, and learning outcomes. Findings
demonstrated that higher levels of factual prior knowledge were associated with deeper learning
and stronger use of digital resources, but that higher levels of deep prior knowledge were
associated with less frequent use of online content and fewer deep revisions. These results
suggest that factual knowledge can serve as a useful knowledge base during self-directed, online
learning tasks, but deeper prior knowledge may lead novice learners to adopt suboptimal
Students frequently use Web-based materials as their primary, and often only, informational
sources for educational tasks (e.g., Graham & Metaxas, 2003). Although students are
enthusiastic about using online materials, research evidence suggests that even college students
perceive themselves as more capable of using technology than they actually are (Stone &
Madigan, 2007). When using the Internet to find answers to questions, Graham and Metaxas
(2003) found that students had difficulty identifying trustworthy sources, were not consistently
able to differentiate between advertising and fact, and failed to confirm online information by
The problem may be especially serious when one considers not just finding information
online, but learning from such information. For individuals engaged in self-directed learning
tasks (e.g., using online resources to write a scientific essay), effective use of online resources
requires students to deploy a series of complex behaviors. Students must decide on search terms
relevant to their information needs, conduct searches using their search terms and options from
online search engines, evaluate lists of potential resources to decide what to explore, choose how
long to explore an online resource, and decide when enough information has been gathered.
From a cognitive perspective, learners must identify and encode relevant information from
potentially large amounts of online content and continuously integrate the new information that
they find into an accurate, overall understanding of domain content. Overall, these challenges are
central to complex cognitive processes such as sensemaking (Russell, Jeffries, & Irani, 2008) and
self-regulated learning (Azevedo, Guthrie, & Seibert, 2004; Azevedo, Moos, Greene, Winters, &
Cromley, 2008). They also highlight the interplay between cognitive processes and online
PRIOR KNOWLEDGE AND ONLINE LEARNING 4
learning behaviors. Understanding more about how known cognitive variables impact students’
online learning behaviors can support the design and development of educational technologies
and interventions designed to improve the quality of students’ Web-based learning experiences.
In this research, we explored how prior knowledge influenced students’ online learning
behaviors and analyzed the relationship between these behaviors and students’ eventual
behaviors because prior knowledge repeatedly has been shown to be a key factor in predicting
learning with text (e.g., Adams, Bell, & Perfetti, 1995; McNamara, Kintsch, Songer, & Kintsch,
1996; McNamara & Kintsch, 1996; Recht & Leslie, 1988; Schneider, Körkel, & Weinert, 1989)
and multimedia materials (e.g., Kalyuga, 2005; Kalyuga, Ayres, Chandler, & Sweller, 2003).
The rationale for why prior knowledge plays a strong role in learning can be drawn from
research and theory in cognitive psychology. Relevant prior knowledge forms a framework for
incoming information (Chi, Glaser, & Rees, 1982), allowing new materials to be integrated into a
flexible knowledge representation that can be transferred to new situations (Kintsch, 1988,
1998). Without a conceptual framework of prior knowledge into which incoming information
can be integrated, learners typically focus on memorizing isolated facts that can be recalled but
cannot be applied outside of the context in which it was learned. This type of knowledge long
Prior knowledge not only can facilitate the development of more integrated knowledge
during learning tasks, but it also plays a strong role in determining the types of learning materials
that will be best suited to an individual learner. Research has found that creating instructional
materials or scenarios that challenge high-knowledge readers can improve their ultimate learning
PRIOR KNOWLEDGE AND ONLINE LEARNING 5
outcomes. For example, high-knowledge readers can benefit when they are provided with text
materials that have low coherence, likely because high-knowledge readers are able to engage the
inferential processes necessary to connect disparate materials (McNamara & Kintsch, 1996).
Existing knowledge in a domain can even, to some degree, compensate for poor reading skill
(Adams et al., 1995) and increase the functional capacity of working memory when readers make
connections between their prior knowledge and the to-be-remembered information (Ricks &
Wiley, 2009).
Research in information search and retrieval has found that prior knowledge plays a
significant role in how individuals spend their time during Web-based searches. Although
experts in a domain spend much of their available search time scanning and reading materials in
electronic environments, novices often focus on formatting and modifying queries (Marchionini,
1995). This may indicate that novices have a hard time finding relevant results within a database
– likely due to the use of poor search terms – or that novices have a harder time recognizing
relevant resources in the results returned following a search. More recently, Marchionini (2006)
has emphasized the transition of Web-based tasks from traditional lookup (e.g., fact retrieval or
verification) to exploratory search that involves complex forms of learning and discovery (e.g.,
synthesis, integration, comparison) as students search for and analyze materials on the Web.
Since research in text comprehension has demonstrated that prior knowledge plays a
strong role in determining the processes in which learners will engage while studying
(Moravcsik & Kintsch, 1993), prior knowledge also is likely to have strong effect on students’
behaviors and strategies when engaging in online learning tasks. Once a text is selected, many of
the same cognitive processes are involved in comprehension regardless of whether that text is
PRIOR KNOWLEDGE AND ONLINE LEARNING 6
encountered online or in print (Butcher & Kintsch, in press). But the Web also poses unique
challenges to learners in the ways that they select and move between texts – the representation
formed by integrating information across documents has been called the intertext model (Perfetti,
Rouet, & Britt, 1999). Some research has found that prior knowledge plays a key role in
predicting the self-regulated processes by which students navigate and process the information
present in hypertext resources. For example, students with high prior knowledge have been
shown to more frequently plan and monitor their learning across hypertext resources whereas
low knowledge learners have been found to be more likely to implement specific strategies (e.g.,
summarizing, note taking) during hypertext learning (Moos & Azevedo, 2008). When choosing a
path through multiple online resources, high-knowledge learners tend to select hypertext links in
a highly-coherent order but lower-knowledge participants tend to follow links according to the
order driven by their interest rather than by textual coherence (Salmerón, Cañas, Kintsch, &
Fajardo, 2005).When students have adequate prior knowledge, they are better able to identify
relevant portions of complex multimedia presentations (Lowe, 1996, 2004) and to allocate their
online study time to more difficult materials (Reader & Payne, 2007).
Although the aforementioned studies show that prior knowledge may have substantial
influences on learners’ approaches to and strategies for learning in online environments, they
also tend to use tightly-controlled versions of online environments that limit the number and
types of digital resources that learners encounter. These constrained systems are, in many ways,
atypical of the large number and variety of materials that are found on the Web. For example,
system focused on atmospheric pollution. Moos and Azevedo (2008) used a commercially-
available digital encyclopedia, but limited participants’ access to three articles relevant to the
PRIOR KNOWLEDGE AND ONLINE LEARNING 7
heart and circulatory system. These limited systems may minimize the degree to which
participants engage in the browsing strategies (e.g., selection, navigation, and trial-and-error
discovery) that have become typical of Web-based search when faced with a large variety of
In more naturalistic environments, prior knowledge may help learners make better
decisions about when and how to engage with digital content, especially when they are free to
decide what digital materials they will use for learning. However, it is not entirely clear that prior
knowledge always will support deeper or more meaningful engagement with online content. In a
with high prior knowledge chose to make use of compact explanations rather than utilizing more
elaborated instructional materials (Scheiter et al., 2009). Although Scheiter et al. (2009)
interpreted this finding as indicative of high knowledge learners tendencies to generate their own
explanations rather than to seek pre-existing explanations, it also may indicate that higher
knowledge learners fail to seek or make use of all the online information available to them.
Whereas true experts may not need additional resources to construct deep and meaningful
explanations of domain content, novice students with higher prior knowledge may still benefit
from greater engagement with online materials if they process them in deep and meaningful
ways. Thus, a key question is how varying prior knowledge among student learners (who still
have much to understand about a domain) will influence their online behaviors.
In the remainder of this paper, we explore the relationship between prior knowledge,
students’ observable and self-reported behaviors and strategies during a self-directed, online
used to characterize the behaviors and processes students employ during learning – as well as
their resulting knowledge – as either more deep or more shallow. Construction-Integration (CI)
(Kintsch, 1988; van Dijk & Kintsch, 1983) is a well-known model of comprehension that
textbase level, and a flexible situation model level. The surface level representation is formed
when learners seek to encode the instructional materials verbatim, as might be the case when
memorizing a poem. The textbase level of representation is more typically the level of
knowledge that learners achieve when trying to memorize typical educational materials. The
textbase is formed from basic ideas in the resource, but is not a verbatim representation of the
original materials. Paraphrases or restatements of the learning materials typify the textbase level
of representation. The most flexible and long-lasting representation is the situation model, which
occurs when learners integrate new information with prior knowledge. The situation model level
materials – the situation model facilitates logical inferences and the application of the learner’s
knowledge to new situations or scenarios. In contrast, the textbase – which remains strongly tied
The characteristics of deep and shallow knowledge also can be used to characterize
assessments that purport to measure domain knowledge and student learning. These assessments
can be categorized by the degree to which they require rote (shallow) recall versus meaningful
(deep) application and synthesis. Whereas shallow assessments require students to recall learned
facts, deep assessments require them to apply learned information to new scenarios and contexts.
PRIOR KNOWLEDGE AND ONLINE LEARNING 9
Rote learning (at the textbase level) is often tested via traditional assessments that require recall
of information drawn from instructional materials, such as fill-in-the-blank, multiple choice, and
true/false items (Kintsch, 1998). In contrast, methods to assess deeper learning at the situation
model go beyond recall, requiring the learner to apply, generalize, or organize their knowledge in
some way. These methods can include simple assessments such as short-answer questions that
require inference and transfer, or more specialized methods such as keyword sorting tasks (e.g.,
McNamara & Kintsch, 1996; Wolfe et al., 1998), student-generated knowledge maps or concept
maps (e.g., Ferstl & Kintsch, 1999), and student-generated drawings of mental models (e.g.,
Butcher, 2006). By combining assessments that target these different levels of understanding,
researchers can determine the extent to which participants have encoded or integrated knowledge
during a learning episode. In this research, we use assessments targeting shallow and deep
knowledge to assess the impact of different levels of prior knowledge on students’ Web
Deep and shallow learning processes. CI characterizes not only the knowledge that
students can learn from instructional materials, but also has been used to characterize the types of
processes in which students engage as they learn. Wiley and Voss (Voss & Silfies, 1996; Wiley
& Voss, 1999) have used the CI model to categorize deep or shallow uses of informational
resources during writing: shallow uses involve borrowing or adding information from resources;
deep uses involve transforming information from this resources. As in previous work (Butcher &
Sumner, 2011), we drew upon Wiley and Voss (Wiley & Voss, 1996, 1999) and the CI model to
characterize shallow and deep learning behaviors and strategies during online learning. We
defined shallow processes as observable or self-reported behaviors that deleted ideas from a
that changed or transformed information in ways that created new knowledge. Typically, this
occurred when learners integrated multiple sources of information or generated new, domain-
relevant inferences.
In the current research, we were interested in how prior knowledge would influence
students’ online learning behaviors when using digital library resources to complete a common
educational task: writing an essay about a scientific topic (in this case, earthquakes and plate
tectonics). This research also explored the degree to which deeper and more shallow forms of
prior knowledge had different effects on students’ Web behaviors and learning outcomes.
Based upon our theoretical framework and prior research, we made four specific
Deep prior knowledge will be positively correlated with more frequent use of
Deep prior knowledge will be negatively correlated with the number of digital
resources used
Deep prior knowledge will be negatively correlated with the duration of digital
resource use
Deep prior knowledge will be positively correlated with gains in deep learning but
Method
PRIOR KNOWLEDGE AND ONLINE LEARNING 11
The data analyzed in this research were collected as part of a prior study of online learning and
conceptual personalization technologies. Additional details regarding the study methodology are
available elsewhere (Butcher & de la Chica, 2010; Butcher & Sumner, 2011). Although these
previous publications (Butcher & de la Chica, 2010; Butcher & Sumner, 2011) statistically
controlled for students’ prior knowledge to examine the impact of digital library tools and
regardless of prior knowledge, this work specifically examines the relationship between students’
prior knowledge, their specific increases in knowledge at the factual and deep levels, and their
behaviors during web-based, self-directed learning tasks. The current study also goes beyond
Butcher and Sumner (2011) by using a finer-grained analysis of students’ allocations of time
during self-directed use of digital content. These measures include the amount of time that
students spent selecting digital resources and the amount of time that they spent viewing or
Participants
Participants were 30 undergraduate students (20 females, 10 males) from the University of
Colorado at Boulder. All participants completed the study for course credit in an introductory
psychology course.
Materials
targeting factual knowledge using true/false items. The true/false test consisted of 40 statements
organized under 10 basic questions. Each statement targeted factual knowledge about the domain
that did not require knowledge application or transfer and could be learned from the online
PRIOR KNOWLEDGE AND ONLINE LEARNING 12
materials provided in the study. For each statement, participants indicated whether it was true or
For each question, multiple true statements were possible. One point was awarded for
each correct answer; thus, 40 points was the maximum possible score.
Deep knowledge assessment. Deep knowledge of the domain (earthquakes and plate
tectonics) was assessed using a domain application test. This test was administered at pre- and
posttest; the test consisted of five short-answer items that asked students to apply their
knowledge about earthquakes and plate tectonics to new contexts and scenarios. For example:
Alyssa says that plate boundaries are visible on maps as the boundaries between
continents and oceans. Thus, Alyssa reasons that earthquakes are most likely to occur at
these boundaries, along the coastal areas of major continents. Is Alyssa correct about the
location of plate boundaries and their relationship to earthquakes? Use what you know
Participants did not receive feedback on their answers during the study. Tests were scored by a
research assistant blind to the purpose of the study; scores were tallied using a rubric that
awarded points for relevant idea units. Because multiple points were possible for each item, a
Digital library materials. The online materials utilized in this study consisted of a
specialized database of 796 learning resources drawn from the Digital Library for Earth System
Education (www.DLESE.org) that had been developed in previous research (Butcher & Sumner,
2011). These Web-based learning resources included general informational resources, scientific
visualizations, animations, imagery, scientific data, simulations, and other forms of interactive
online content. All resources were targeted to high-school age learners and addressed topics in
earthquakes and plate tectonics. Although the number and variety of the digital resources in this
collection approximates the varied materials available on the Web, we limited the relevance and
behaviors could be examined in more depth. That is, we did not want to determine how prior
knowledge would influence the success of students’ search queries or their ability to generate
relevant search terms. Rather, we were interested in determining how different levels of prior
knowledge would influence the ways in which students approached and managed their online
Revision questionnaire. After students finalized their essays, they completed a questionnaire
about the strategies that they used during essay revision. All students had received computer-
generated feedback on their essays which highlighted up to five specific statements from the
essays as potentially needing feedback (cf., Butcher & Sumner, 2011). On the revision
questionnaire, students responded to three reflection prompts regarding each issue targeted by
feedback. For each set of reflection prompts, students were free to answer as they chose (e.g.,
1. Why do you think the system identified this statement for revision?
2. Describe how your revised statement is different from your original one. Please be as
specific as you can.
PRIOR KNOWLEDGE AND ONLINE LEARNING 14
3. Explain what you did to revise your original statement and why. Please be as specific
as you can.
The revision questionnaire was segmented into idea units and coded, yielding the following data:
Depth of diagnosis. Participants’ responses to the first prompt were used to code
diagnosis of essay issues. All idea units written for the first prompt that focused on
diagnosis of the essay were coded as either deep (requiring transformation of essay
Idea preservation. Participants’ responses to the second prompt were used to code
the percent of issues for which students self-reported that they preserved (or attempted to
preserve) the basic concept or idea that they had written in their original essays.
Depth of revisions. Participants’ responses to the third prompt were used to code
self-reported strategies during the Web learning task. All idea units written in response to
the third prompt that focused on revision behaviors were coded as either deep (describing
Digital resource seeking. Participants’ responses to the third prompt also were
analyzed for idea units related to the use of digital resources during learning and revision
activities. This variable represents the percentage of issues (those statements identified as
in potential need of revision) for which students self-reported that they had used or
Final essays. Participants’ final essays were scored by a research assistant who was blind to
participant conditions and the purpose of the study. The research assistant gave each essay a
score for holistic quality between 1 and 7 (1 = lowest quality, 7 = highest quality), which
represented the overall quality of the essay (including accuracy and clarity of articulated ideas).
PRIOR KNOWLEDGE AND ONLINE LEARNING 15
The research assistant also calculated a content score for each essay. Content scores were
calculated by counting the number of unique (and correct) domain ideas present in each essay.
Screen recording software. Camtasia software was used to record participants’ on-screen
interactions as they revised their initial essays. These recordings enabled a detailed, in-depth
analysis of the amount of time that students devoted to different aspects of the Web-based task
(e.g., browsing links, digital resource exploration) as well as the ways in which students used
digital resources and worked with their essays. For example, copying and pasting text from a
digital resource to their essays or adding their own explanation of a graph viewed in a digital
resource.
Observable online behaviors. The screen capture videos of participants’ actions made it
possible to analyze three general stages during the Web-supported learning task: resource
selection, digital resource exploration, and essay revision. The on-screen videos of each
participants’ online actions were analyzed for the amount of time spent in each of these
Resource selection. Resource selection time is defined as the total amount of time
that participants spent retrieving and evaluating lists of digital resource links. Resource
selection time was measured from the second that a list of links or a search box first
appeared onscreen until the participant moved to a different page (either by clicking on a
link or switching back to essay writing). Because not all participants used the same
interface to identify digital resources and returned lists of links sometimes stay onscreen
while participants conduct another search, search time and link evaluation could not be
consistently distinguished. Thus, time spent generating search queries (when necessary)
participant reads, scrolls or otherwise interacts with the digital content during digital
resource exploration. Digital resource exploration time is defined as the total amount of
time that a participant spent in the digital resources that s/he selected (i.e., clicked on), as
measured from the second that the site/page loaded until the participant closed the
Essay revision. Essay revision time is defined as the total amount of time that
participants spent viewing, revising, or otherwise interacting (e.g., scrolling) with their
essays. Essay revision time was calculated from the second that the essay was brought
into view until participants closed the essay or switched to a different window (e.g., the
Screen capture videos also were used to examine the degree to which learners made use of
Number of unique digital resources: Screen captures were analyzed for the
number of unique digital resources that participants viewed during the online learning
task. The URL domain name was used to define “uniqueness” – navigation across pages
that shared a common domain name (e.g., nasa.org) was considered to be exploration of
the same overall site. The number of unique digital resources reflects the number of
Procedure
As reported in Butcher and Sumner (2011), data was collected during two sessions. In session
one, students first completed the prior knowledge assessments. Students had up to 10 minutes to
complete the factual knowledge assessment (the true/false items) and 20 minutes to complete the
PRIOR KNOWLEDGE AND ONLINE LEARNING 17
deep knowledge assessment (the domain application test). Students then explored five resources
about the theory of plate tectonics, earthquakes, volcanoes, and mountain formation for 15
minutes. The purpose of this initial exploration was to allow students to refresh their basic
domain knowledge without providing significant time for new learning; resources were chosen
by domain experts. Students then spent 30 minutes writing a 250+ word essay (the initial essay)
In session two, which took place approximately one week after session one, students
revised their initial essays for 35 minutes. Students were provided with access to the collection of
796 DLESE online resources to inform their revisions. Students’ use of online resources was
self-directed during revision; each student decided if, when, and how to make use of online
resources as they revised their essays. All students were allowed to move freely between their
essays, the search interface, and the online educational resources. After revising their essays,
students completed the revision questionnaire (20 min), the factual knowledge assessment (10
Results
Appendix A provides descriptive statistics for all self-reported and observed variables. It should
be noted that the maximum score on the deep knowledge assessment was nine points out of 30.
Thus, even students with higher levels of deep knowledge in this study should be considered
novice learners. This study cannot (and does not intend to) distinguish between the behaviors of
expert and novice learners. Rather, our focus is on how higher or lower levels of factual and deep
prior knowledge influence the online behaviors and knowledge outcomes of novice learners.
Interrater agreement was very good (Cohen’s Kappa = .83). Shallow behaviors/comments
elements of the essays such as grammar, spelling, style, or the addition of rote
coded as a shallow strategy. In the screen capture videos, shallow revision behaviors
were coded when students added information that was derived by copying text directly
revision behaviors were coded when students changed, interpreted, analyzed, predicted,
or otherwise went beyond the information in the digital resources or integrated multiple
sources of information when adding new content to their essays. Students’ self-reported
learning approaches and strategies were coded when they reflected these deep processes
or reported the use of deep behaviors. For example, on the revision questionnaire, “found
better information about the types of faults and corrected my definition of faults to reflect
students were very accurate in describing their revisions to essays and their use of online
resources. On average, 89% (min = 50%, max = 100%) of students’ self-reported actions
PRIOR KNOWLEDGE AND ONLINE LEARNING 19
matched observed behaviors seen in the screen captures. Both the median and mode
percent of match between self-reported and observed activities was 100%. Thus, only
A series of standard, multiple linear regressions were calculated, using levels of prior
knowledge as the independent variables and the frequency of an observed online learning
behavior as the dependent variable in each model. Table 1 shows the results of these models, and
Table 1
Linear regression models: Using levels of prior knowledge to predict observable behaviors
Dependent measure Β SE Β β
Independent variables (levels of prior knowledge)
Resource selection (F(2, 27) < 1)
Factual prior knowledge 0.17 5.73 < .01
Deep prior knowledge -3.26 14.06 -.04
Resource exploration (F(2, 27) = 4.02, p < .03; R2 = .23)
Factual prior knowledge* 21.77 8.46 .44
Deep prior knowledge -28.18 20.75 -.23
Essay revision (F(2, 27) = 1.18, p > .32; R2 = .08)
Factual prior knowledge -16.04 10.48 -.28
Deep prior knowledge 6.12 25.70 .04
Unique # of resources (F(2, 27) = 1.79, p > .18; R2 = .12)
Factual prior knowledge 0.16 .09 .33
Deep prior knowledge .11 .21 .09
* p < .05
Results demonstrate that levels of prior knowledge were not significant predictors of the
time that students spent seeking and evaluating links to digital resources (i.e., resource selection
PRIOR KNOWLEDGE AND ONLINE LEARNING 20
time), the time that students spent engaged with their essays (i.e., essay revision time), or the
number of unique digital resources that students utilized during learning. However, students’
prior knowledge did predict the amount of time that they spent engaged with digital resource
content during a Web-supported learning task. Higher levels of factual prior knowledge
predicted more time spent within the available digital resources, viewing or interacting with their
contents. Although deep prior knowledge did not reach statistical significance as a predictor of
digital resource exploration times, it had a negative relationship with digital resource
exploration: higher deep prior knowledge predicted less time spent viewing or interacting with
A series of standard, multiple linear regressions also were calculated to examine the
relationship between levels of prior knowledge and students’ self-reported learning strategies.
Table 2
Linear regression models (levels of prior knowledge and self-reported Web learning behaviors)
Dependent measure Β SE Β β
Prior knowledge type
Number of deep diagnoses (F < 1)
.04 .05 .15
Factual prior knowledge
.04 .13 .06
Deep prior knowledge
Number of deep revisions (F(2, 27) = 3.55, p < .05; R2 = .21)
.14 .09 .27
Factual prior knowledge
-.47 .21 -.38
Deep prior knowledge*
Idea preservation (F(2, 27) = 3.77, p < .04; R2 = .22)
-.01 .003 -.30
Factual prior knowledge†
.01 .007 .38
Deep prior knowledge*
PRIOR KNOWLEDGE AND ONLINE LEARNING 21
As can be seen in Table 2, across these self-reported variables, deep prior knowledge was
a consistent and significant predictor of behaviors. Higher levels of deep prior knowledge were
predictive of fewer deep revisions, less frequent use of digital resources during the revision task,
and more frequent attempts to preserve (rather than revise or transform) the ideas present in
students’ original essays. The statistical trend for higher levels of factual knowledge was
reversed: higher factual knowledge was associated with less frequent preservation of original
Overall, these results show that deep prior knowledge may reduce the depth of students’
self-directed behaviors during a Web learning task. Moreover, higher levels of deep prior
knowledge predicted lesser use of digital resources during learning, including reduced seeking of
digital information and less frequent deep revisions. Is it a problem that deep prior knowledge
influenced students’ behaviors in these ways during a Web-supported learning task? It might be
argued that higher levels of deep prior knowledge make it unnecessary for students to consult
available online resources because they have sufficient understanding to process information
deeply and to generate deep explanations on their own. It also is possible that higher levels of
deep prior knowledge are associated with preservation of original essay ideas because the
original ideas are useful and accurate. Thus, a key question is whether higher levels of deep prior
knowledge are more strongly correlated with deep learning outcomes and final essay quality than
factual prior knowledge, despite lower levels of engagement with digital resources and different
learning strategies.
PRIOR KNOWLEDGE AND ONLINE LEARNING 22
Table 3 shows bivariate correlations between students’ deep and factual prior knowledge,
Table 3
Correlations between levels of prior knowledge, learning outcomes, and Web behaviors
Interestingly, deep prior knowledge of the domain was not significantly correlated to any
measured learning outcome. In contrast, students’ prior factual knowledge about the domain was
PRIOR KNOWLEDGE AND ONLINE LEARNING 23
significantly and positively correlated with deep knowledge gain (as measured by proportion of
possible improvement on the knowledge application test) and negatively correlated with factual
knowledge gain (as measured by the proportion of possible improvement on the true/false items).
This is somewhat surprising, in that one might expect factual knowledge gains to be supported
by strong, existing factual knowledge. However, prior factual knowledge predicted longer
searches for digital content and greater use of digital resources provided in the Web collection
for longer periods of time. Moreover, as seen in Table 4, the time that students spent searching
for and using digital resources was significantly and positively related to the development of
deep understanding (as measured by the knowledge application assessment). Thus, strong prior
factual knowledge predicts greater use of digital content, and the greater use of digital content is
more closely associated with development of deep domain understanding than the accumulation
of additional domain facts. For these novice learners, higher levels of deep prior knowledge
appear to discourage strategies related to use of and engagement with digital content. However,
caution must be used in interpreting these data since causality cannot be determined.
Table 4
levels of deep prior knowledge hurt these novice learners. We have seen that higher deep prior
knowledge was negatively correlated deep essay revisions and positively correlated to the
preservation of original essay ideas (see Table 3), but were these self-reported strategies
correlated to learning outcomes? As seen in Table 5, the pattern of correlations between learning
outcomes and students’ deep and shallow learning strategies (i.e., the self-reported essay
diagnoses and revisions) were consistent with what would be predicted by the CI model
(Kintsch, 1988, 1998). Deep behaviors were significantly correlated with the development of
deep understanding. Shallow behaviors were negatively associated, though only at the level of a
statistical trend, with deep knowledge gain. The frequency with which students reported
preserving their original essay ideas during revision was significantly and negatively associated
with improvement in deep understanding: the more students worked with their original ideas
(rather than modifying them), the lower their gains on the deep knowledge assessment.
Table 5
As demonstrated by the correlations in Table 5, the strategy changes associated with deep prior
Web-based learning task. However, is there any evidence that deeper strategies themselves were
associated with stronger use of Web resources? Table 6 provides the bivariate correlations
between students’ self-reported strategies and their self-reported and observed use of digital
resources during learning. Results show that deep essay revisions were significantly and
positively correlated with the degree to which students reported seeking digital resources and the
observed length of time that they viewed or interacted with such resources.
Table 6
The overall pattern of results from the regressions and the correlational analyses
demonstrates a strong relationship between different levels of prior knowledge, students’ Web-
based learning strategies, and their observed online learning behaviors. Factual prior knowledge
predicted greater use of Web resources during learning whereas deep prior knowledge predicted
a greater tendency to avoid Web resources and to rely on one’s previous ideas. At the same time,
greater use of Web resources was positively correlated with deeper revision strategies and the
PRIOR KNOWLEDGE AND ONLINE LEARNING 26
development of deeper domain understanding. For these novice learners, higher levels of deep
prior knowledge may not have been an advantage. Rather, a strong textbase – in the form of
higher factual prior knowledge – predicted a greater willingness to use Web content and a greater
General Discussion
The current results demonstrate that, when examining the relationship between learning
behaviors and students’ prior knowledge during a Web-based educational task, it is important to
consider not only students’ overall amount of prior domain knowledge, but the depth of this prior
knowledge. Overall, only one of the predictions about deep prior knowledge was supported by
the results from this study. Deep prior knowledge was negatively correlated with the duration of
digital resource use – that is, higher deep prior knowledge predicted less overall use of Web
content. Surprisingly, the observed relationships between deep prior knowledge, learning
strategies, and deep learning outcomes were the opposite of what was predicted: deep prior
knowledge was negatively associated with deep essay revisions and deep learning outcomes;
further, there was no relationship between deep prior knowledge and the number of digital
resources used. In contrast to our predictions, factual prior knowledge demonstrated stronger,
positive relationships to the use of Web content than did deep prior knowledge. Higher factual
prior knowledge was positively correlated with the number of digital resources used or viewed
during learning.
Taken as a whole, deep prior knowledge mainly was associated with compromised
success in a Web-based learning task for these novice learners. Higher levels of deep prior
knowledge were associated with less frequent modification of domain ideas, less frequent
engagement with deep strategies during a Web-supported learning task, and a general reluctance
PRIOR KNOWLEDGE AND ONLINE LEARNING 27
to make use of the Web resources available. These results are consistent with prior research
findings showing that many students are unable to use the Web effectively for learning tasks
(Graham & Metaxas, 2003; Stone & Madigan, 2007), but the current results demonstrate that this
problem is not simply the result of students lacking sufficient domain knowledge to guide
effective online exploration. In this research, high-quality resources were returned for every Web
search. Moreover, this research found that students who were likely the most capable of learning
from high-quality online resources also were most reluctant to engage with them. This result
extends the previous finding that higher knowledge learners are less likely to engage with
elaborated instructional materials online (Scheiter et al., 2009). Our findings suggest that higher
deep prior knowledge may lead novice students to make poor choices about their online learning
behaviors during a self-directed task, although the directional nature of this relationship cannot
be conclusively determined from the analyses conducted here. Although all students in this
research had much to learn about the domain, deep prior knowledge predicted fewer deep
revisions and less frequent use of Web resources when students worked to improve their essays.
If higher deep prior knowledge was associated with deeper learning or higher quality essays at
posttest, one might consider these students to be effectively managing their approaches to online
learning without assistance from Web resources. However, in this research, deep prior
knowledge did not predict learning outcomes. Rather, higher levels of factual prior knowledge
were associated greater learning, as well as with the use of greater numbers of digital resources
for longer periods of time. So, why did higher factual prior knowledge predict deeper learning
One possibility is that higher factual prior knowledge may serve as an effective basis for
Higher factual prior knowledge was positively correlated both with the use of Web resources and
necessary (but not sufficient) for development of the situation model (Kintsch, 1998). In this
research, existing factual knowledge does appear to prepare students to learn more deeply
through the use of deep strategies and Web exploration. Higher factual prior knowledge may
have prompted students to seek resources that explained the connections or relationships between
the facts that they knew. In contrast, higher deep prior knowledge was negatively correlated both
with meaningful revisions and students’ perceived need for Web information. It appears that
higher deep prior knowledge changed both the types of revision that students perceived as useful
as well as the information resources they considered necessary to complete such revisions.
These results from this study are consistent with findings from text comprehension
research demonstrating that high-knowledge learners may actually learn less than their lower-
knowledge peers when instructional materials fail to provide sufficient challenge to spur higher
knowledge learners to engage deeply and to participate actively in the learning situation
(McNamara, Kintsch, Songer, & Kintsch, 1996; McNamara & Kintsch, 1996). It may be the case
that the Web environment or the Web-based learning task must be made more challenging to
engage learners with higher levels of deep prior knowledge; for example, by using deep
questioning during online tutoring dialogues (Craig, Sullins, Witherspoon, & Gholson, 2006),
McAlack, Menchana, & Stoddart, 2005), or making use of adaptive scaffolding appropriate to
the individual learner (Azevedo, Cromley, & Seibert, 2004). Although higher levels of deep prior
knowledge did not predict deeper strategies during online learning, current findings
demonstrated that students who do take a deep approach to online learning are likely to benefit
PRIOR KNOWLEDGE AND ONLINE LEARNING 29
from their use of high-quality digital resources. Further, the current results suggest that higher
levels of factual knowledge may provide students with enough background to make strong use of
Web materials at the same time that they are willing to extend and transform their original ideas.
knowledge. For example, recent research has demonstrated that the trustworthiness of a text
source plays a significant role in comprehension (Bråten, Strømsø, & Britt, 2009). Other research
has demonstrated that effective online learners must utilize a variety of self-regulated learning
activating prior knowledge (Azevedo, Guthrie, & Siebert, 2004). The current results demonstrate
that prior knowledge itself may not be sufficient to drive deep, strategic processes during online
learning. Indeed, for novice learners, stronger deep prior knowledge may compromise their
effective use of deep learning strategies with Web content. However, a strong factual knowledge
base appears to facilitate better use of Web content and, ultimately, support deeper
understanding.
materials and deep online learning strategies. Moreover, both deep approaches to learning and
longer overall engagement with digital resources were related to stronger domain understanding
as assessed by a knowledge application test. These data show that a combination of high-quality
online resources, such as those found in educational digital libraries, and deep learning strategies
can promote meaningful domain understanding even during short periods of study. However, the
directionality of these data cannot be determined. Do students apply deeper learning strategies
when interacting with high-quality digital resources? If students recognize the trustworthiness of
these resources (cf., Bråten, Strømsø, & Britt, 2009) and their domain relevance, it is plausible
PRIOR KNOWLEDGE AND ONLINE LEARNING 30
that they may employ more effective comprehension strategies. However, it is also possible that
students who gravitate toward using deeper learning strategies are simply more successful in
making use of most digital resources, especially high-quality resources that are clearly relevant
It is important to keep in mind that these data were collected during a constrained
experimental task in a laboratory environment. Although the study used a large collection of real,
online materials drawn from an educational digital library, the fact that these materials were
chosen to be age- and topic-relevant does ease some burdens associated with typical Web-
supported learning in naturalistic environments. For example, in this study, students’ searches
inevitably returned relevant and high-quality digital resources. Students were not faced with
weeding out irrelevant, erroneous, or useless materials. For self-directed Web learning tasks
outside of the lab, we expect that the usefulness of digital resource exploration in promoting deep
understanding is largely tied to the quality and relevance of the resources that students locate and
select. Outside of the lab, we also expect that students with higher deep prior knowledge may
experience different motivations (e.g., grades) that could influence the degree to which they
employ deep strategies and/or engage with available online materials. It also is possible that
learners with higher deep prior knowledge could be supported by personalized learning systems
(de la Chica, Ahmad, Sumner, Martin, & Butcher, 2008) that increase the depth or complexity of
Overall, the observed results provide an interesting foundation for future work assessing
the relationships between students’ levels of prior knowledge, online learning strategies,
References
Adams, B. C., Bell, L. C., & Perfetti, C. A. (1995). A trading relationship between reading skill
307-323.
Azevedo, R., Cromley, J. G., & Seibert, D. (2004). Does adaptive scaffolding facilitate students'
Azevedo, R., Guthrie, J. T., & Seibert, D. (2004). The role of self-regulated learning in fostering
Azevedo, R., Moos, D. C., Greene, J. A., Winters, F. I., & Cromley, J. G. (2008). Why is
Butcher, K. R. (2006). Learning from text with diagrams: Promoting mental model development
Butcher, K. R., & de la Chica, S. (2010). Supporting student learning with adaptive technology:
Butcher, K. R., & Kintsch, W. (in press). Text comprehension and discourse processing. In A. F.
Butcher, K. R., & Sumner, T. (2011). Self-directed learning and the sensemaking paradox.
http://dx.doi.org/10.1080/07370024.2011.556552
Chi, M. T. H., Glaser, R., & Rees, E. (1982). Expertise in problem solving. In R. J. Sternberg
(Ed.), Advances in the psychology of human intelligence (Vol. 1). Hillsdale, NJ: Erlbaum.
Craig, S. D., Sullins, J., Witherspoon, A., & Gholson, B. (2006). The deep-level-reasoning-
question effect: The role of dialogue and deep-level-reasoning questions during vicarious
de la Chica, S., Ahmad, F., Sumner, T., Martin, J. H., & Butcher, K. R. (2008). Computational
Ferstl, E. C., & Kintsch, W. (1999). Learning from text: Structural knowledge assessment in the
Erlbaum.
Graham, L., & Metaxas, P. T. (2003). Of course it's true; I saw it on the internet!
University Press.
Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect.
University Press.
University Press.
McNamara, D. S., Kintsch, E., Songer, N. B., & Kintsch, W. (1996). Are good texts always
McNamara, D. S., & Kintsch, W. (1996). Learning from texts: Effects of prior knowledge and
Moos, D. C., & Azevedo, R. (2008). Self-regulated learning with hypermedia: The role of prior
Moravcsik, J. E., & Kintsch, W. (1993). Writing quality, reading skills, and domain knowledge
360-374.
PRIOR KNOWLEDGE AND ONLINE LEARNING 34
Perfetti, C. A., Rouet, J.-F., & Britt, M. A. (1999). Toward a theory of documents representation.
Reader, W. R., & Payne, S. J. (2007). Allocating time across multiple texts: Sampling and
Recht, D. R., & Leslie, L. (1988). Effect of prior knowledge on good and poor readers' memory
Ricks, T. R., & Wiley, J. (2009). The influence of domain knowledge on the functional capacity
Russell, D. M., Jeffries, R., & Irani, L. (2008). Sensemaking for the rest of us. Paper presented at
Salmerón, L., Cañas, J. J., Kintsch, W., & Fajardo, I. (2005). Reading strategies and hypertext
Scheiter, K., Gerjets, P., Vollmann, B., & Catrambone, R. (2009). The impact of learner
Schneider, W., Körkel, J., & Weinert, F. E. (1989). Domain-specific knowledge and memory
Stone, J. A., & Madigan, E. (2007). Inconsistencies and disconnects. Communications of the
van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse comprehension. New York:
Academic Press.
PRIOR KNOWLEDGE AND ONLINE LEARNING 35
Voss, J. F., & Silfies, L. N. (1996). Learning from history text: The interaction of knowlede and
comprehension skill with text structure. Cognition and Instruction, 14(1), 45-68.
Whitehead, A. N. (1932). The aims of education and other essays. London: Ernest Benn Limited.
Wiley, J., & Voss, J. F. (1996). The effects of 'playing historian' on learning in history. Applied
Wiley, J., & Voss, J. F. (1999). Constructing arguments from multiple sources: Tasks that
promote understanding and not just memory for text. Journal of Educational Psychology,
91(2), 1-11.
Wolfe, M. B. W., Schreiner, M. E., Rehder, B., Laham, D., Foltz, P. W., Kintsch, W., &
Landauer, T. K. (1998). Learning from text: Matching readers and texts by latent
Zheng, R., Stucky, B., McAlack, M., Menchana, M., & Stoddart, S. (2005). WebQuest learning
Appendix A
Means and standard deviations for prior knowledge, self-reported, and observed variables
M SD Min Max
Factual prior knowledge 28.4 5.0 16 37
Deep prior knowledge 3.2 2.2 0 9
Improvement (% of possible): factual 6.9 26.4 -54 44
knowledge
Improvement (% of possible): deep 11.9 16.4 -20 52
knowledge
Shallow essay diagnosisa 1.1 1.3 0 4
Deep essay diagnosisa 3.1 1.4 0 5
Idea preservationa (%) 9.4 7.9 0 25
Shallow essay revisionsa 3.1 2.2 0 9
Deep essay revisionsa 4.0 2.6 0 10
Digital resource seekinga (%) 42.9 39.4 0 100
Posttest essay: holistic quality score 3.8 1.7 1 7
Posttest essay: content score 5.4 2.4 1 11
Resource selectionb (sec) 208.5 150.7 56 656
Digital resource explorationb (sec) 525.1 253.3 69 1140
b
Essay revision (sec) 1082.8 287.1 371 1595
b
Number of unique digital resources 6 2 2 10
a b
Note. N=30. Self-reported by participants. Coded from screen captures.