HYPERTEXT LEARNING ENVIRONMENTS:
THEORY, DESIGN, AND RESEARCH 
Michael J. Jacobson
Learning Technology Center, Vanderbilt University
Nashville, TN 37240, USA
Chrystalla Maouri and Punyashloke Mishra
University of Illinois at Urbana-Champaign, 51 Gerty Drive
Champaign, IL 61020, USA
Aurora University, 347 South Gladstone Avenue
Aurora, IL 60506, USA
Jacobson, M. J., Maouri, C. , Mishra, P., & Kolar, C. (1996). Learning with hypertext learning environments: Theory, design, and research. Journal of Educational Multimedia and Hypermedia, 5(3/4), 239-281.
Permission to distribute electronically granted by the publisher, Association for the Advancement of Computing in Education (AACE).
THEORETICAL AND RESEARCH OVERVIEW
Cognitive Flexibility Theory
Situated Cognition Theory
Cognitive Flexibility, Situated Cognition, and the Design of Hypertext
The Instructional Utilization of Conceptually-Indexed Hypertext Learning Environments
Epistemic Beliefs and Learning With Hypertext
Hypertext Time-On Task
Short Answer Factual Acquisition
Scoring Methodology for Essays
Knowledge Synthesis Essays
Knowledge Representation, Hypertext Design Features, and Knowledge Transfer
Learner Control and Attitudes
Metaphors for Learning: Conceptual Landscapes, Hypertext, and Three Dimensional Criss-Crossings
This paper reports on a study into the acquisition and transfer of complex knowledge after using a conceptually-indexed hypertext learning environment based on recent cognitive theory and research. The experiment employed two differently structured hypertext "thematic criss-crossing" (TCC) treatments, Guided TCC and Learner Selected TCC, that demonstrated interrelationships between abstract conceptual and case-specific knowledge components in the hypertext materials. These two hypertext treatments differed primarily in terms of the modeling and scaffolding they provided and the degree of learner control. The control condition allowed the free exploration of the hypertext materials but provided no modeling or scaffolding support. Epistemic beliefs held by students concerning the nature of learning and the structure of knowledge were also assessed. Students who were assigned to the Guided TCC hypertext group and who regarded learning as an active process of constructing meaning were found to perform at a significantly higher level on a knowledge synthesis task of near transfer than students in the other treatment groups or students with a simpler set of epistemic beliefs. Overall, the results of this study suggest that students need explicit modeling and scaffolding support in order to learn complex knowledge from a case-based, conceptually-indexed hypertext learning environment. Implications of these findings for future hypertext and hypermedia research are also considered.
The capabilities of hypertext learning environments to store, interconnect, and provide access to a wide range of knowledge represented as text, graphics, audio, and video would seem to provide significant affordances to enrich student learning (Beeman et al., 1987; Jonassen, 1986; Kearsley, 1988). Although research has begun to explore educational applications of hypertext technologies, many of these early projects have primarily described features of hypertext systems or usage patterns of these systems by students. While some studies have demonstrated improved learning outcomes associated with the use of hypertext programs ( Baker, Niemi, & Herl, 1994; Beeman et al., 1988; Jacobson & Spiro, 1995; Jonassen & Wang, 1993; Lehrer, 1993; Shapiro, 1994), overall there has been limited empirical documentation of the educational efficacy of such systems. Factors contributing to the provisional nature of research on learning with hypertext include: (a) preoccupation with technological functionality, (b) methodological problems, and (c) lack of attention to relevant cognitive learning theory and research (Jacobson, 1994).
A principled basis for designing and using hypertextual learning environments is needed. We feel recent cognitive science theory and research into learning with technology systems is beginning to provide perspectives relevant to this task (Anderson, Conrad, & Corbett, 1989; Cognition and Technology Group at Vanderbilt, 1992; Collins & Brown, 1988; Mandl & Lesgold, 1988; Pea & Sheingold, 1987; Scardamalia & Bereiter, 1991; Spiro, Feltovich, Jacobson, & Coulson, 1992; Vosniadou, DeCorte, Glaser, & Mandl, in press; White, 1993; White & Frederiksen, 1987). While this work has involved a wide range of technological learning environments, there are two theoretical views we feel have particular relevance to the design and use of hypertext and hypermedia technologies: cognitive flexibility theory (Spiro, Coulson, Feltovich, & Anderson, 1988; Spiro et al., 1992; Spiro & Jehng, 1990) and situated cognition theory (Brown, Collins, & Duguid, 1989; Clancey, 1993; Greeno & Moore, 1993; Lave & Wenger, 1991; Norman, 1993; Suchman, 1987). As we discuss below, these two theories of learning share several common elements yet also compliment each other by focusing on different facets of the complexities associated with human learning processes, educational environments, and technological systems (Jacobson, 1994).
In this paper, we describe a study that explored learning a complex
domain, the social impact of technology, with a hypertext learning
environment based on recent theory and research. Before discussing
the study itself, we first provide an overview of the cognitive
flexibility and situated cognition theoretical perspectives that
are relevant to this research. We also discuss how beliefs learners
hold about the nature of learning and the structure of knowledge--which
we refer to as epistemic beliefs--have been found to influence
learning in general and learning with hypertext in particular.
Cognitive flexibility (CF) theory was initially formulated to address factors contributing to failures to learn complex knowledge at advanced instructional levels (Feltovich, Spiro, & Coulson, 1989; Spiro et al., 1988; Spiro, Vispoel, Schmitz, Samarapungavan, & Boerger, 1987). It proposes complex knowledge may be better learned for flexible application in new contexts by employing case-based learning environments that include features such as: (a) use of multiple knowledge representations, (b) link abstract concepts in cases to depict knowledge-in-use, (c) demonstrate the conceptual interconnectedness or web-like nature of complex knowledge, (d) emphasize knowledge assembly rather than reproductive memory, (e) introduce both conceptual complexity and domain complexity early, and (f) promote active student learning. These features have been used prescriptively to specify design elements for complex, multidimensional, and nonlinear environments such as hypertext and hypermedia (Spiro et al., 1992; Spiro & Jehng, 1990). The initial empirical study of this approach for structuring a hypertext learning environment found evidence that students learned a complex knowledge domain and applied their acquired knowledge in new contexts (i.e., knowledge transfer) (Jacobson, 1990; Jacobson & Spiro, 1995). A more detailed discussion and illustration of CF theory design elements used for the experimental hypertext program in this study is given below.
Recent work in the cognitive sciences has attempted a comprehensive socio-cognitivist theoretical treatment of human cognitive functioning. Central to much of this work is a view of human cognitive processing as being fundamentally situated in contexts of activity (Brown et al., 1989; Clancey, 1993; Edelman, 1992; Greeno & Moore, 1993; Lave & Wenger, 1991; Norman, 1993; Suchman, 1987). Clancey (1993) has observed that situated cognition (SC) theorists do not regard knowledge as being strictly inside the head, rather that knowledge is "a coupling between adapted neurological processes and how we have structured our environment" (p. 101). Indeed, SC theory holds that context is not only important, "but what constitutes the context, how you categorize the world, arise together with processes that are coordinating physical activity" (Clancey, 1993, p. 95).
An important issue for SC is the nature of activity, and how practitioners act in authentic situations in meaningful and purposeful ways (Brown et al., 1989). Some SC theorists have described products of authentic activity in terms of "indexicalized" representations (Brown et al., 1989; Suchman, 1987).  For example, language employs many expressions (e.g., I, you, here), that people can correctly interpret because of the context in which these expressions are used. Brown et al. (1989) further suggest that
knowledge...indexes the situation in which it arises and is used. The embedding circumstances efficiently provide essential parts of its structure and meaning. So knowledge, which comes coded by and connected to the activity and environment in which it is developed, is spread across its component parts, some of which are in the mind and some in the world much as the final picture on a jigsaw is spread across its component pieces. (pp. 36-37)
A corollary of indexicality is "learning methods that are embedded in authentic situations are not merely useful; they are essential" (p. 37).
Given a SC perspective, a central question is: When is learning "situated learning?" One limited view of SC learning is that it only involves learning in non-school settings with novices observing experienced practitioners (e.g., apprentices observing the master). While much of the early work in this area did involve research on learning in naturalistic settings (Lave, 1988; Suchman, 1987), SC learning theory is by no means intended to only refer to learning in non-school or "real world" contexts. A richer view of SC learning theory, more commonly held by researchers in this area, regards all learning--both school-based and non-school based-- as fundamentally situated (Brown et al., 1989; Cognition and Technology Group at Vanderbilt, 1990; Collins, Brown, & Holum, 1991; Greeno & Moore, 1993; Moore et al., 1994).
The challenge is to create environments in schools that embody some of the successful learning dynamics identified in the research on non-school contexts such as craft apprenticeship. Brown and colleagues (1989) discuss two examples of situated school-based learning involving mathematics that illustrate what they refer to as cognitive apprenticeship methods (Brown et al., 1989; Collins, 1989; Collins, Brown, & Newman, 1987). Aspects of cognitive apprenticeship include modeling the knowledge in an authentic activity, supporting the students doing the task through scaffolding or coaching, allowing students to articulate their knowledge and to confront ineffective strategies and misconceptions, and finally empowering students by gradually fading or withdrawing support. Collins (1989) has also discussed several ways in which a SC perspective may be applied to instructional technology systems ranging from microworlds and intelligent tutoring systems to computer networks and graphing packages. However, to date a SC or cognitive apprenticeship learning perspective has not been explicitly applied to hypertext and hypermedia learning environments.
One may regard multiple theories of learning and instruction, such as CF and SC theories, as different conceptual lenses from which to inform the design and the instructional use of hypertext learning environments (Jacobson, 1994). In this section, we utilize these conceptual lenses to first provide an overview of theory-based hypertext design features, followed by a consideration of general factors associated with the instructional utilization of such a hypertext program.
There are three hypertext design features central to the program developed for the present study: case-based materials, conceptual indexing, and case-theme commentaries. We refer to such a system as a conceptually-indexed hypertext learning environment. While these hypertext design elements were originally articulated from a CF theoretical perspective (Jacobson, 1990; Jacobson & Spiro, 1995; Spiro et al., 1992; Spiro & Jehng, 1990), here we briefly note how these elements are also consistent with SC theory.
Case-based hypertext materials. The recommendations of CF theory to use case-based materials and of SC theory to involve students in authentic activities are complimentary notions, both intended to address an unfortunate tendency in education to give students decontextualized activities divorced from the practices of the culture of real practitioners. Creating hypertext systems with multiple cases associated with a domain also allows multiple representations of knowledge and multiple contexts for learning (each a central element of these two theories) to occur as a natural part of the learning environment. Further, the use of case materials then forms the foundation for the next two hypertext design elements.
Conceptual indexing and variable hypertext links. Conceptual indexing involves coding the case-based materials with important abstract conceptual or structural knowledge (e.g., cognitive or mental models, schemas, themes) based upon understandings and representations held by domain experts. The use of hypertext conceptual indexing instantiates the CF theory element of "knowledge-in-use" and the SC theory element of "indexicality of learning and knowledge." In contrast to the fixed hypertext links found in most hypertext and hypermedia systems, conceptual indexing of cases provides the basis for variable hypertext links (Conklin, 1987; Jacobson, 1990; Jacobson, 1994; Jacobson & Levin, 1995; Jacobson & Spiro, 1995; Spiro et al., 1992; Spiro & Jehng, 1990).Variable hypertext links provide technical functionality for representing knowledge in multiple situations of use in a way that is difficult or impossible for traditional linear technologies such as text or video to do. Unfortunately, for use in hypertext learning environments, the more commonly employed fixed hypertext links suggest a non-situated view of knowledge--one that is rigid and prespecified by the given fixed link-node structure. In contrast, variable hypertext links instantiate CF and SC theoretical views by being flexible and adaptable to different conditions, suggestive of knowledge as an emergent property assembled or constructed in dynamically changing situated contexts.
Case-theme commentaries. The use of case-theme commentaries as a hypertext design feature is a corollary to theoretical rationales for conceptual indexing. This hypertext design element provides a short explanation of how a structural dimension of knowledge (e.g., a theme or concept) applies in different case-specific or situated contexts (see examples from the Technology and the Twentieth Century: Impact on Society and Culture, a hypertext program in the Method section). From a CF perspective, these commentaries describe knowledge-in-use (i.e., how the important thematic or abstract conceptual knowledge specifically applies across multiple case contexts). From a SC perspective, this design element helps make explicit to a learner nuances of a situated context in which knowledge is constructed and used. Thus, case-theme commentaries can provide indexicalized knowledge representations central to a SC view of learning and human cognitive functioning.
Given a hypertext learning environment with core design features such as those sketched above, how might such a system be used instructionally? In previous research (Jacobson, 1990; Jacobson & Spiro, 1995), a Thematic Criss-Crossing (TCC) learning activity was created that primarily employed these three hypertext design elements. Briefly, TCC was intended to use conceptual indexing to highlight different intellectual facets of the instructional material. The conceptual indexing created variable hypertext links across different case sections, with different theme combinations linking different case sections together.
Important aspects of CF and SC views of learning and cognition were implemented in the previous study by providing the learner with several different combinations of themes to guide the hypertextual TCC of case sections. From CF theory (which was the original inspiration for TCC), this activity was intended to "criss-cross the conceptual landscape," demonstrating the interconnected and web-like nature of knowledge and the process of knowledge assembly. The different conceptual indices structured the variable hypertext links for creating web-like connections between different case sections, which in turn permitted the program to rapidly assemble interconnected knowledge components based on different combinations of themes. From the SC theory perspective, the TCC activity modeled for the learner aspects of the situated and indexical nature of knowledge across multiple contexts. Furthermore, the experimental hypertext learning environment cognitively scaffolded the learner during the learning activity itself by providing support based on structural knowledge components (i.e., conceptual indices, multiple combinations of themes for different intellectual perspectives on domain, case-theme commentaries).
Hypertext TCC students have been found to perform at a higher level on a knowledge transfer task (a written problem-solving essay) than the control students who used a minimal hypertext/drill program (Jacobson, 1990; Jacobson & Spiro, 1995). General beliefs about the nature of learning have also been shown to influence the educational efficacy of TCC (Jacobson, 1990; Jacobson & Spiro, 1995), an issue we discuss next.
Research has begun to explore how the epistemic beliefs students hold about the nature of the learning process itself can influence learning (Feltovich et al., 1989Jehng, Johnson, & Anderson, 1993; Perry, 1968; Schoenfeld, 1983; Schommer, 1993; Schommer, 1994). Epistemic beliefs may be regarded as general assumptions held by the learner about the nature of learning and knowledge. Schoenfeld (1983) suggested that basic beliefs of this type may determine the cognitive resources a student allocates to a particular problem-solving activity. Research has documented how epistemic beliefs may contribute to learning difficulties such as integrating prior knowledge to new knowledge and poor reading comprehension (Schommer, 1993), mathematical problem-solving failure (Schoenfeld, 1983), and the formation of conceptual misunderstandings (Feltovich et al., 1989).
The influence of epistemic beliefs on learning with a hypertext system may be particularly important as characteristics of hypertextual learning environments (e.g., nonlinear access to interrelated knowledge components, emphasis on knowledge transfer and independent thinking instead of rote memorization of prespecified "facts") may conflict with epistemic beliefs held by the student. Preliminary evidence from the earlier learning and hypertext study (Jacobson, 1990; Jacobson & Spiro, 1995) suggests that students holding a set of simple epistemic beliefs and preferences about learning (i.e., those favoring rote memorization of hierarchically organized and prespecified material) were less able to learn and apply their knowledge after using a hypertext system than students who possessed a more complex set of epistemic beliefs (i.e., those favoring the more active personal construction of knowledge and regarded the conceptual structure of knowledge as having a more multidimensional and web-like form). This finding suggests that characteristics of a hypertext learning environment--such as conceptually-indexed material, studying the content in a nonlinear manner across multiple contexts--may be differentially effective for learners with different epistemic beliefs.
As this was a follow-up study, several of the research questions were intended to replicate findings or explore issues raised in the earlier learning and hypertext study (Jacobson, 1990; Jacobson & Spiro, 1995). What are aspects of a TCC learning activity with a conceptually-indexed hypertext that contribute to improved knowledge transfer compared to the free exploration of a hypertext without conceptual indexing? Is hypertext TCC really necessary, or is merely the nonlinear hypertextual exploration of the case materials (i.e., a non-theme-based approach based on fixed links) sufficient for learning complex knowledge from hypertext? Might there be a differential performance on knowledge acquisition, near transfer, and far transfer tasks by students learning with differently structured hypertext systems? How might epistemic beliefs about the nature of learning enhance or constrain learning with a hypertext learning environment? What are student attitudes about using such hypertext systems?
Two different thematic criss-crossing learning activities that
used a conceptually-indexed hypertext were developed for the present
study--Guided Thematic Criss-Crossing (Guided TCC) and Learner Selected Thematic Criss-Crossing (Learner Selected TCC). These two treatments and the control
condition (Free Hypertext Exploration), which are described in
detail below, varied in two main ways: (a) the amount of modeling and scaffolding provided to the students as they studied the hypertext knowledge-base,
and (b) the degree of learner control. We anticipated that the
Guided TCC and the Learner Selected TCC groups would perform at
comparable levels on tests of knowledge transfer because both
involved TCC learning activities that demonstrated conceptual
knowledge interrelationships across multiple contexts. However,
we expected that the Learner Selected TCC students would have
more positive attitudes about using the hypertext program because
of their moderate degree of control over using the hypertext compared
to the Guided TCC students who had a lower level of user control.
The Free Hypertext Exploration control condition provided no modeling
or scaffolding of conceptual interrelationships, but did allow
the greatest amount of learner control. Whereas these students
were expected to have the most positive attitudes towards using
the program, they were also expected to have the lowest scores
on the knowledge transfer tests. Another research hypothesis,
based on findings from the earlier study (Jacobson, 1990; Jacobson & Spiro, 1995), was that students with a more complex set of epistemic beliefs
about the nature of learning would perform at a higher level on
tests of knowledge transfer when using the main experimental hypertext
treatments (Guided TCC and Learner Selected TCC) than those students
with a simpler set of epistemic beliefs.
The main components of the experimental hypertext program used in this study, Technology and the Twentieth Century: Impact on Society and Culture, had been developed by the first author for use in the previous learning and hypertext study. This program had two main parts: "reading stage" and "study stage." For the current study, the hypertext materials in the reading stage were the same for all three treatment groups and allowed all students to receive the same initial exposure to the instructional content. The study stage presented contrasting hypertext conditions for each of the three treatment groups that are described below.
The Technology and the Twentieth Century: Impact on Society and Culture hypertext contained six cases describing a range of ways technology has influenced society (see Table 1). Six social impact of technology themes representative of expert knowledge in this area were also employed: Socially Constructed Nature of Technology, Actor Networks, Progress - Problems, Technological Efficiency, Freedom - Control, Community - Alienation). The cases consisted of one to eight sections of text, each of which was thematically-indexed with the applicable social impact of technology themes. Each case section had its own theme list (i.e., the conceptual indices) which linked to expert thematic commentary detailing how the theme applied to the particular case section. For example, in the case Aborted Introduction of the French Electric Car, the second section The EDF Plan was indexed with the themes of Social Technology, Actor Networks, Progress - Problems, Technological Efficiency, and Community - Alienation (see Figure 1), while in another case, The Dilemma of the Artificial Heart, the second section Features, Problems, and Costs was analyzed as having the themes of Actor Networks, Progress - Problems, and Technological Efficiency (see Figure 2). These two case sections had three of the themes in common (Actor Networks, Progress - Problems, and Technological Efficiency), while the second section of the Aborted Introduction of the French Electric Car had two themes not found in The Dilemma of the Artificial Heart (Social Technology, Community - Alienation). Thematic commentary was available that explained how the themes were instantiated in terms of the details in each case section (see the lower screens shown in Figures 1 and 2). That is, while the same general theme may be found on multiple case sections, the specific case-theme commentaries are distinct for each case section (as can be seen by comparing the case-theme commentaries for Technological Efficiency in Figures 1 and 2). All students went through the reading stage in the same manner to receive an initial exposure to surface components of knowledge (i.e., case-specific information), structural knowledge components (i.e., the themes), and the situated nature of this knowledge (i.e., the case-theme commentaries).
During the study stage, the three treatment groups were given the same on-line "knowledge synthesis" study questions for each of the four sessions of the study. As examples, during Session 2, the first knowledge synthesis study question was "Does the control over a group of people made possible by a particular technology, such as computers, always result in alienation?" (shown on screens for each of the three treatment conditions in Figures 3, 4, and 5), while during Session 3, the fourth knowledge synthesis study question was "Do social groups always strive to implement technologies to achieve efficiency?" These questions served three functions. First, the knowledge synthesis study questions were intended to provide an intellectual goal for the students during the study stage. Second, these questions were written to point out important issues and thematic interrelationships in the Technology and the Twentieth Century: Impact on Society and Culture hypertext materials and to help promote a deeper understanding of the content. Third, these questions were used for the assessment of near transfer. The students were told they would write short essays on two of the fifteen knowledge synthesis study questions, but they were not told ahead of time which questions would be selected. The two on-line knowledge synthesis study questions listed as examples above were the questions given to the students to write their knowledge synthesis essays.
Even though the students in the three treatment groups were given the same knowledge synthesis questions during the study stage, the hypertext design features of each treatment varied in terms of the degree of modeling and scaffolding and the degree of user control. Next we discuss the specific features of the three treatment groups.
Experimental condition 1: Guided Thematic Criss-Crossing. The Guided TCC condition employed essentially the same hypertext Thematic Criss-Crossing treatment that had been used in the previous study. The knowledge synthesis questions (from the study stage on-line activities, described above) were designed to provide a goal for the students as they engaged in thematic criss-crossing of the hypertext case materials. However, based on previous research into expert-novice differences (Chi, Glaser, & Farr, 1988), we felt novices in this domain might have difficulty identifying structural dimensions of knowledge (i.e., the social impact of technology themes) that were implicit in the knowledge synthesis study questions.
The Guided TCC treatment was therefore intended to model and scaffold the process of thematic criss-crossing. Modeling was provided by pre-selecting and listing themes of particular relevance to the knowledge synthesis study questions (Figure 3, top right of the screen) that were then used to structure the variable hypertext links for thematic criss-crossing. Cognitive scaffolding was available through the core design features in the program that supported TCC. For example, relevant themes for the knowledge synthesis study questions were displayed on the session Study Questions card (e.g., top right of Figure 3) and on the cards for the case sections (e.g., bottom screen, right side of Figure 2). Also, the session Study Questions card listed the variable hypertext links used to structure the hypertext navigation (lower right of Figure 3), and the case-theme commentaries made explicit how the abstract conceptual knowledge applied to the context of a particular case section (bottom screen, left side of Figure 2). Students thus had nonlinear access to sets of relevant case sections from which to study each question by clicking on a line in the list or clicking on the arrow key buttons to cycle through the case section list. It should also be noted that while the Guided TCC hypertext treatment provided cognitive scaffolding to the learner, it allowed a relatively low level of user control as both the themes and the case sections were pre-selected for each of the knowledge synthesis study questions.
The students in Guided TCC were expected to develop an interconnected understanding of the social impact of technology materials based on the case-specific and conceptual dimensions of knowledge. As this treatment was quite similar to the TCC treatment used in the previous study in which significant knowledge transfer was found, these students were expected to perform at a high level on the transfer tasks compared to the control students. However, we anticipated that the lower level of user control would lead to less positive attitudes towards using the program compared to the students in the other groups who had higher levels of user control.
Experimental condition 2: Learner Selected Thematic Criss Crossing. The Learner Selected TCC condition implemented features from earlier prototype cognitive flexibility hypertext systems (Spiro & Jehng, 1990). In order to investigate whether college students can readily engage in a hypertext TCC learning activity, this treatment allowed the learners to select their own theme combinations for the knowledge synthesis study questions. The program used these theme combinations to structure the variable hypertext links the student could access while studying the questions. It was assumed that selecting appropriate themes for different questions would be necessary for an effective (i.e., more expert-like) thematic criss-crossing of the case section materials while exploring issues raised in the knowledge synthesis study questions. In contrast to Guided TCC, however, there was no modeling of the process of selecting themes for particular questions for this group (although the students did receive on-line practice for the mechanics of doing this procedure during the first session). After the students selected the theme or themes they felt were applicable to the knowledge synthesis study question, the program's Search option (see Step 2 on the right side of the screen shown in Figure 4) would put the case sections indexed with the themes selected by the students into the variable hypertext link list (see bottom right side of the screen shown in Figure 4). As some theme searches might yield extremes in the number of case sections returned (e.g., 0 or 15), students had the opportunity to do multiple theme searches for each on-line knowledge synthesis study question and to see the case sections listed before actually studying the questions. To insure that the students in this condition did not spend an excessive amount of time on each question, the students were informed that once they started to read the case sections, they could not do additional theme searches for that particular question. After the students selected their themes and searched for the matching case sections, this condition provided the same type of cognitive scaffolding for thematic criss-crossing as the Guided TCC group received.
As the students in this treatment group were engaged in a thematic criss-crossing learning activity, they were expected to develop an interconnected and flexible understanding of the material. Furthermore, these students were hypothesized to do more active constructive processing as they needed to analyze each question and to select their own themes to structure the thematic criss-crossing. Consequently, these students were predicted to perform at the highest level on the knowledge transfer tests. Also, as this treatment provided a moderate degree of user control over the hypertext, these students were anticipated to have more positive attitudes towards the program than the low user control Guided TCC group (but not as positive as the high user control Free Hypertext Exploration group).
Control condition: Free Hypertext Exploration. The Free Hypertext Exploration condition allowed the students cumulative hypertextual access to all of the instructional materials they had read for each of the sessions. As with the other treatment conditions, the students read the case sections that were placed in a list on the right side of the screen (Figure 5). However, in Free Hypertext Exploration, all of the case sections read up through the current session were placed in the hypertext link list for the students to select from as they studied the on-line questions. These hypertext links, however, were fixed as the same links were available for all the questions (although the program did not provide hypertext links to case sections which had not yet been read by the student). The Free Hypertext Exploration control condition used fixed hypertext links in contrast to the variable sets of hypertext links based on different theme combinations employed in the Guided TCC and Learner Selected TCC experimental conditions. Of the three groups, the Free Hypertext Exploration permitted the highest degree of user control over accessing the hypertext materials, and were therefore expected to have most positive attitudes about using the program of the three groups.
Demographic information was collected at the first session (e.g., age, sex, year in school). A general cognitive profile of the students was obtained using: (a) a vocabulary test (French, Ekstrom, & Price, 1963) to provide a measure of verbal ability (Anderson & Freebody, 1979), and (b) an epistemic beliefs and preferences (EBP) instrument to assess beliefs about the nature of learning and the structure of knowledge. An earlier version of the EBP instrument had been administered in the previous learning and hypertext study (Jacobson, 1990; Jacobson & Spiro, 1995). In the present study, a new version of the EBP instrument was used in which low correlating items from the older EBP version were revised or eliminated. The new EBP instrument consisted of 18 pairs of matched statements, with one positively worded and the other negatively worded. Students rated the degree to which they agreed or disagreed with each statement on a seven point Likert scale. Representative items from the EBP instrument are listed in Table 2.
Thisstudy employed three main assessment measures: short answer questions, knowledge synthesis essays, and problem-solving essays. These measures were intended to be representative of activities that could be used in university classroom settings. The short answer questions were the same as those used in the earlier study and assessed the acquisition of factual case-specific information contained in the hypertext materials (e.g., What are the possible medical complications to the artificial heart? Historically, how did investigative agencies have to create a file on a suspected criminal or dissenter?). The knowledge synthesis essays required the students to write coherent and persuasive essays in order to effectively answer two selected knowledge synthesis study questions drawn from the pool of on-line questions used during the study stage (see above). The students were told they would write a short essay answering an on-line knowledge synthesis study question at each of the two test periods (at the end of Session 2 and Session 4), but they were not told which of the questions would be selected. This cognitively demanding task, which we regard as a test of near transfer, required the students to integrate relevant case-specific and abstract conceptual knowledge derived from several different case sections in the Technology and the Twentieth Century: Impact on Society and Culture hypertext.
The problem-solving essays were intended to assess far within-domain transfer. This task involved reading a technology and society problem scenario not been previously covered in the hypertext materials, and writing a short essay that analyzed the important issues in the scenario and proposed a solution to the problem. Two problem-solving essays from the previous learning and hypertext study were administered again (Jacobson, 1990; Jacobson & Spiro, 1995), and three new problem-solving essay scenarios were developed to investigate the reliability of an essay task of this type.
Subjects were 17 to 19 year old university students at the freshman and sophomore level who were paid for their participation. They were randomly assigned to one of the three experimental treatment groups. Of the 77 students who completed the study, eight had missing data and were eliminated from the final dataset. In the final pool of 69 students, 47 (68%) were female and 22 (32%) were male. The treatment group assignments were 25 to Free Hypertext Exploration (control group), 25 to Guided TCC Hypertext, and 19 to Learner Selected TCC Hypertext.
The study was conducted over a two week period, with four sessions (two per week) each lasting approximately two hours. At the first session, the students were given basic instructions on using a Macintosh computer and on using the Technology and the Twentieth Century: Impact on Society and Culture hypertext program. The main learning activities involved the "reading stage" which was common to all three groups and the "study stage" in which the actual contrasting experimental treatments were administered. Before the Session 1 and Session 2 reading stage activities, the students read general descriptions of the six social impact of technology themes (these were also available during the reading and the study stage activities). During the reading stage, the students were assigned specific cases at each of the four sessions of the study: Session 1 The Aborted Introduction of the French Electric Car; Session 2 The UTOPIA Project: Computers Unions, and Printing and Jehovah's Witnesses and Blood Transfusion; Session 3 Privacy and Computer Data Banks and High Technology, Automation, and the Worker; and Session 4 The Dilemma of the Artificial Heart. Students initially read each of the sections of the case in a linear manner similar to reading a book. They were also asked to look at the theme list and to read the case-theme commentaries on each case section. Unlike the previous hypertext study, this study did not require the students to read all of the case sections and case-theme commentaries.
A pretest was administered at Session 1, and the main evaluation instruments were given at the end of Session 2 and Session 4. The pretest consisted of the epistemic beliefs and preferences (EBP) instrument, and writing a social impact of technology problem-solving essay (to obtain measures of prior knowledge and writing ability). For the Session 2 and 4 tests, the format consisted of: short answer questions, knowledge synthesis essay, reading/recall task 1 (not discussed in detail in this paper, see note 10), problem-solving essay 1, reading/recall task 2, problem-solving essay 2, and attitude questions. Two versions of each test were prepared that counter-balanced the presentation of the problem-solving essays to control for possible sequencing effects; these were randomly distributed to the students.
The reverse worded items on the epistemic beliefs and preferences (EBP) instrument were recoded to positive (i.e., 1 = 7, 2 = 6, etc.), and a Pearson correlation matrix for these items was calculated. Low correlating items with the overall EBP instrument were removed, leaving 21 items that were used for subsequent analyses (Cronbach alpha reliability = .76). The group scores on the EBP (Mdn = 4.00, M= 4.00, SD = 0.63) were uncorrelated with verbal ability, age, or gender. As in the previous study, students at or above the mean were regarded as having a complex EBP (n= 35, M= 4.49, SD = 0.40) and students below the mean as having a simple EBP (n= 34, M= 3.48, SD = 0.33).
Over the four sessions of the study, the students for whom on-line data was collected spent 73.52 minutes in the reading stage (SD = 15.57 minutes) and 81.66 minutes in the study stage (SD = 13.44). For this subset of the students, there was no significant difference in the amount of time spent in either the reading stage or the study stage between the three treatment groups (reading stage: F2,43 = 0.17, p> .05; study stage: F2,43 = 0.57, p> .05).
The Session 2 and Session 4 short answer tests on factual information contained in the hypertext materials consisted of 10 and 12 items respectively. These two tests had Cronbach alpha reliabilities of 0.63 and 0.54 respectively (see means and standard deviations in Table 3). Vocabulary had a moderate correlation with Session 2 and Session 4 short answer scores (Pearson rof .48 and .35 respectively). With vocabulary as a covariate, a split-plot factorial analysis of covariance (SPFAC) 33·2 design (Kirk, 1982) was run on the short answer scores. There were two main factors of Group (three levels: Free Hypertext, Guided TCC, and Learner Selected TCC) and EBP (two levels: simple EBP and complex EBP), and a repeated measures factor of Session (two levels: Session 2 and Session 4). No significant main effect was found for EBP (F1,62 = 0.52, p> .05), but there was a significant within subjects interaction between the treatment Group factor and Session (F2,65 = 4.00, p< .05) (see Figure 6). Two treatment contrast interactions were then tested (Kirk, 1982). No significant difference was found for the mean short answer scores between the groups receiving the Free Hypertext Exploration and Learner Selected TCC treatments (F1,65 = 0.82, p> .05). However, the contrast comparing the average of the short answer scores for the Free Hypertext Exploration and Learner Selected TCC groups to the mean scores for the Guided TCC group revealed a significant difference for Session 4 (F1,65 = 6.82, p< .05). These results indicate that the Free Hypertext Exploration and the Learner Selected TCC groups performed at an equivalent level on the short answer tests on the Session 2 and 4 tests, and that students in these two groups also achieved significantly higher short answer scores by Session 4 than the Guided TCC group.
Each of the students wrote one problem-solving essay at the pretest, a knowledge synthesis essay during the Session 2 and Session 4 tests, and two problem-solving essays during each of the Session 2 and Session 4 tests. The knowledge synthesis essays were assigned scores ranging from 0 to 15 on four subscales (quality of question analysis, incorporation of themes or case information from computer readings, accuracy of response to the question, and writing quality) and on the overall quality and persuasiveness of the knowledge synthesis essay. Similarly, the problem-solving essays were scored 0 to 15 on each of five subscales (quality of analysis of key issues of problem, incorporation of themes or case information from computer readings, incorporation of relevant new ideas not previously presented in computer readings or in the scenario description, plausibility and creativity of solution to scenario issues, and writing quality) and on the overall holistic quality and persuasiveness of the problem-solving essay. Given the large number of knowledge synthesis and problem-solving essays collected, the essays were randomly divided into two piles and scored by the two trained raters (Garner, Alexander, Gillingham, Kulikowich, & Brown, 1991). The raters then randomly selected and scored approximately 20% of the essays from the other rater's set. With the exception of the pretest problem-solving essay subscale of writing quality, all subscales for both the knowledge synthesis and problem-solving essays were significantly correlated with the overall holistic score of the quality and persuasiveness of the written essays. The holistic scores were then used to determine the interrater reliability. These scores were collapsed from the 15 point scale to a 5 point scale, after which the percent agreement between the two raters (plus or minus one point) was calculated (Baker et al., 1994; Jacobson & Spiro, 1995). There was a 91% and a 90% agreement between the two raters on the jointly scored knowledge synthesis and knowledge transfer essays. The originally assigned holistic scores for the knowledge synthesis and problem-solving essays were used for subsequent statistical analyses.
Unlike the short answer items, there was only a low correlation between vocabulary and the knowledge synthesis essays. Consequently, the analysis of the knowledge synthesis essays used a split-plot factorial analysis (SPF) 32·2 design with Group and EBP as main factors, and Session as the repeated measures factor (see Table 3 for group means and standard deviations). No significant main effects were found, however, there was a significant within subjects interaction (Group x EBP x Session, F2,62 = 4.11, p< .05). (Figure 7 graphs the adjusted least square mean scores for the knowledge synthesis essays and displays the interaction.) A series of a priori contrast tests were conducted and two contrast tests were found to be significant. First, while no significant differences were found between the simple EBP and complex EBP students in either the Free Hypertext or the Learner Selected TCC conditions, the complex EBP students in the Guided TCC condition did achieve significantly higher scores by Session 4 than the simple EBP students in the same condition (F1,63 = 5.52, p< .05). The adjusted least square mean scores for the simple EBP students for Session 2 and Session 4 were 6.54 (SE = 0.79) and 5.62 (SE = 0.90) respectively, while for the complex EBP students the scores were 6.25 (SE = 0.82) and 8.33 (SE = 0.93) respectively. A second contrast compared the adjusted least square mean knowledge synthesis essay scores of the complex EBP students in the Free Hypertext group (Session 2 = 7.23, SE = 0.79, and Session 4 = 6.62, SE = 0.90) and in the Learner Selected TCC group (Session 2 = 6.82, SE = 0.83, and Session 4 = 5.82, SE = 0.97) to the adjusted least square mean knowledge synthesis essay scores achieved by the complex EBP students in the Guided TCC group (Session 2 = 6.25, SE = 0.82, and Session 4 =8.33, SE = 0.93). This contrast found that the complex EBP students in the Guided TCC group achieved significantly higher knowledge synthesis essays by Session 4 compared to the average of the scores for the complex EBP students in the Learner Selected TCC and Free Hypertext Exploration groups (F1,63 = 3.33, p< .05). Overall, these results indicate that students in the Free Hypertext Exploration and Learner Selected TCC groups performed in a statistically equivalent manner on the knowledge synthesis essays. Further, the students with complex EBP in the Guided TCC group scored significantly higher on the knowledge synthesis essays than all other students.
Table 4 shows the problem-solving essay means and standard deviations for the three treatment groups. As with the knowledge synthesis essays, there was no correlation between vocabulary and the problem-solving essay scores. An analysis of variance on the Session 1 pretest problem-solving essay found no significant differences between the three groups (F2,65 = 0.77, p> .05). As indicated above, the problem-solving essays were administered in a counter-balanced manner to determine if there were sequencing effects; no significant differences were found. The two essays written during each of the two sessions had moderately high correlations (rsession 2 = 0.57 and rsession 4 = 0.69), so the two problem-solving essays for each session were averaged together. As above, a SPF 32·3 design was run (Group and EBP factors), but with three levels of Session (Session 1 pretest, Session 2, and Session 4) for the repeated measures factor. (Missing data reduced the number of students for this analysis to 68.) The pretest, Session 2, and Session 4 problem-solving essay adjusted least square mean scores for the three groups were 6.58 (SE = 0.53), 8.96 (SE = 0.46), 8.88 (SE = 0.56) respectively for Free Hypertext Exploration; 5.69 (SE = 0.52), 9.18 (SE = 0.45), 8.50 (SE = 0.55) respectively for Guided TCC; and 6.04 (SE = 0.60), 8.37 (SE = 0.52), 8.19 (SE = 0.64) respectively for Learner Selected TCC. No significant main effect was found between the three groups on the transfer essay scores (F2,62 = 0.65, p> .05). However, a significant main effect difference was found with the EBP factor for the average of the essay scores over the three sessions (F1,62 = 4.22, p< .05). The adjusted least square mean scores for the two levels of EBP groups on the pretest, Session 2, and Session 4 problem-solving essays were 5.45 (SE = 0.46), 8.53 (SE = 0.40), and 8.17 (SE = 0.49) respectively for simple EBP, and 6.76 (SE = 0.44), 9.14 (SE = 0.38), 8.88 (SE = 0.47) respectively for complex EBP. There was also a within subjects interaction (F2,124 = 29.9, p< .05); overall pretest, Session 2, and Session 4 means were 6.10 (SD = 2.66), 8.91 (SD = 2.19), and 8.59 (SD = 2.72). Post hoc contrast tests of the interaction found that while there was no significant difference between the overall problem-solving essay mean scores of Session 2 and Session 4 (F6,62 = 0.29, p> .05), there was a significant difference between the pretest and the average of the problem-solving essay mean scores for Session 2 and Session 4 (F6,62 = 9.13, p< .05). In summary, all students wrote comparable Session 2 and Session 4 problem-solving essays, and all students significantly improved from the Session 1 pretest to the Session 2 and Session 4 problem-solving essays, but the complex EBP students wrote better problem-solving essays averaged across the three test sessions than the simple EBP students.
The students were asked to complete a ten item attitude instrument at the end of sessions two and four. This previously administered instrument (Jacobson, 1990) had four subscales: (a) perceived learning effectiveness of the instructional activity; (b) attitudes towards the hypertext; (c) reasonableness of the length of time involved in using the program; and (d) program ease of use. As there were no significant differences in the mean scores on the subscales in terms of either the Group or EBP factors or any change in the scores between sessions two and four, the subscale scores were used to create an overall attitude score. The Session 2 and Session 4 attitude scores were then averaged together. The Session 2 and Session 4 adjusted least square mean combined attitude scores for the three groups were 3.79 (SE = 0.08) and 3.85 (SE = 0.06) respectively for Free Hypertext Exploration; 3.90 (SE = 0.08) and 3.91 (SE = 0.06) respectively for Guided TCC; and 4.13 (SE = 0.09) and 3.99 (SE = 0.07) respectively for Learner Selected TCC. A two factor analysis of variance (Group and EBP) revealed significant main effects for both factors. The overall difference between the attitudes of the three groups was significant (F2,63 = 4.00, p< .05). Scheffe post hoc analysis of pairwise mean differences indicated that the attitudes of the Free Hypertext group (adjusted least square mean = 3.82, SE = .056) were significantly lower than the Learner Selected TCC group (adjusted least square mean = 4.06, SE = .056), but that the Guided TCC group attitudes (adjusted least square mean = 3.90, SE = .065) were not significantly different than the other two groups. For the second factor, the simple EBP group had more positive attitudes towards the program (adjusted least square mean = 4.00, SE = 0.05) than complex EBP students (adjusted least square mean = 3.86, SE = 0.05); this was a significant difference (F1,63 = 3.85, p= .05). There was no significant interaction between these two factors F2,63 = 2.26, p> .05).
One popular view of hypertext instructional systems is that substantive
learning will occur because students have great flexibility and
control over studying the material. The results of this study,
however, suggest that providing students with flexible access
and a high degree of user control to hypertextually interconnected
materials is not enough. In fact, specific hypertext design factors
were found in this study to significantly influence learning outcomes
on different knowledge acquisition and transfer tasks. We first
discuss the findings associated with the major research areas
of the study and then consider general issues associated with
learning from hypertext.
The EBP instrument used in this study included several revised items from the older version of the EBP employed in the Jacobson and Spiro (1995) study. Although a detailed comparison of the two versions of the instruments is beyond the scope of this paper, the overall Cronbach alpha reliability of the new instrument (with the low correlating items removed) was virtually the same as the alpha of the older version of the instrument (new EBPa = .76 versus old EBPa = .75). Also, the lack of a correlation between epistemic beliefs as assessed with the EBP and a general measure of verbal and academic ability (assessed with vocabulary, see Anderson and Freebody, 1979) replicates the finding from the study reported in Jacobson (1990) and Jacobson and Spiro (1995). This suggests that epistemic beliefs constitute an independent factor from verbal and general academic ability. Furthermore, as we discuss below, this study has found additional evidence that the types of beliefs students hold about the nature of learning and the structure of knowledge can influence how well they learn for near and far transfer from a hypertext.
There are three major knowledge acquisition and transfer findings in this study. First, the Free Hypertext Exploration and Learner Selected TCC groups scored higher on the short answer tests than the Guided TCC group by Session 4. Second, on the near transfer knowledge synthesis essays, the students with complex epistemic beliefs in the Guided TCC group scored higher than the students in the Free Hypertext Exploration and Learner Selected TCC groups. Third, there were no significant differences found between the three groups on the far transfer problem-solving essays, but complex EBP students in Learner Selected TCC did score significantly higher than simple EBP students in that same treatment group.
Our main hypotheses related to the performance of the Free Hypertext Exploration group on the short answer tests and the Guided TCC group on the knowledge synthesis essays were generally confirmed. However, we did not expect three findings: (a) the Free Hypertext Exploration and Learner Selected TCC students (rather than just the Free Hypertext Exploration students) scored at a higher level on the short answer tests than the Guided TCC students; (b) the complex EBP students in Learner Selected TCC scored lower than those in Guided TCC (and the same as both simple and complex EBP students in Free Hypertext Exploration) on the knowledge synthesis essays (rather than complex EBP students in both Guided TCC and Learner Selected TCC scoring higher); and (c) the Free Hypertext Exploration students performed at the same level as the Guided TCC and Learner Selected TCC students on the far transfer problem-solving essays (rather than Guided TCC and Learner Selected TCC scoring higher).
Why did we obtain this pattern of results? We suspect that three interacting factors contributed to these findings: the structure of knowledge represented in the hypertext, the nature of the learning activity, and the cognitive support provided by the three hypertext treatment conditions. With respect to the hypertext representation of knowledge, each of the groups were exposed to the case-based (i.e., surface) knowledge components and conceptual (i.e., deep structure) knowledge components associated with the technology and society domain during both the reading and the study stages. The nature of the learning activity during the study stage required studying a series of questions dealing with complex and multifaceted issues and thematic interrelationships that cut across all of the cases in the hypertext. However, the design features of the three study stage treatment groups varied in terms of the cognitive support provided to the students as they were engaged in this cognitively demanding learning activity. In particular, the treatments varied in terms of modeling how to use a conceptually-indexed hypertext to explore complex questions and the cognitive scaffolding provided by the treatments as the students traversed the case and thematic hypertextual "landscape." We suggest that the design variation associated with modeling and scaffolding provided during the study stage may explain the pattern of results in this study. To elucidate this thesis, we will consider the results on the knowledge synthesis essays, followed by the short answer scores, and ending with the problem-solving essays.
The results of this study suggest only students with more complex beliefs about the nature of learning and structure of knowledge were able to benefit from Guided TCC to prepare for the knowledge synthesis questions. This finding of an interaction between epistemic beliefs and learning with an instructional hypertext is consistent with the earlier hypertext study (Jacobson, 1990; Jacobson & Spiro, 1995) and with other research investigating learning and epistemological beliefs (Schoenfeld, 1983; Schommer, 1993; Schommer, 1994).
Those students in the Free Hypertext Exploration and Learner Selected TCC treatments lacked the same degree of modeling and scaffolding support of knowledge assembly provided to the Guided TCC group. The Free Hypertext Exploration treatment was closest to a "traditional" hypertext in that the technological opportunity was available to make nonlinear links across the hypertext materials, but no explicit cognitive modeling or scaffolding support was provided. The Learner Selected TCC, in contrast to Free Hypertext Exploration, did provide scaffolded support for a theme-based exploration of the materials. However, Learner Selected TCC did not explicitly model the process of theme selection for the knowledge synthesis study questions as did the Guided TCC. Despite these design differences, in most ways the students in the Free Hypertext Exploration and Learner Selected TCC groups performed the same on the dependent measures (e.g., similar short answer and knowledge synthesis scores).
It may be that the students in the Free Hypertext Exploration and Learner Selected TCC treatments were less successful at answering the complex knowledge synthesis questions because as non-experts they lacked an interconnected understanding of the deep structure of knowledge in this domain to help guide their hypertext traversals while studying. Furthermore, it is possible that because of this lack of attention to deeper, abstract conceptual knowledge interconnections, those students in Free Hypertext Exploration and Learner Selected TCC focused more on surface features of the hypertext materials and thus performed better on the factually oriented short answer questions.
In the earlier learning and hypertext study (Jacobson, 1990; Jacobson & Spiro, 1995), a significant difference on a far transfer task was found between the Thematic Criss-Crossing experimental group and the Hypertext/Drill control group. In the discussion of those results, it was argued that the Hypertext/Drill treatment inculcated inert or rigid representations of knowledge due to the decontextualized presentation of information and emphasis on rote recall associated with the drill treatment. In contrast, the Thematic Criss-Crossing Hypertext treatment used in that study (essentially the same as the Guided TCC treatment in the present study) helped the students develop more non-inert or flexible representations of knowledge which allowed these students to better utilize their acquired knowledge in the new contexts associated with the written problem-solving essay transfer tasks.
In the present study, however, there was no difference in the performance between the three treatment groups on the measures of far within-domain transfer (the problem-solving essays) with the exception of the complex EBP students in all three groups. One explanation for this finding is that there were common hypertext features available to all three treatment groups (e.g., multiple case-based materials, multiple abstract thematic or conceptual perspectives, thematic commentaries linking abstract conceptual knowledge to specific cases, nonlinear exploration of content). The hypertextual exploration of the instructional domain using these common hypertext features was evidently sufficient to help students develop flexible or "non-inert" understandings of the content and thus lead to the comparable levels of performance on the problem-solving essays. From the perspective of assessing differences in far transfer, the control condition selected for this study (Free Hypertext Exploration) was probably not sufficiently different from the experimental conditions, unlike the control condition used in the earlier knowledge transfer and hypertext study (Jacobson, 1990; Jacobson & Spiro, 1995) (although, as we discuss above, the present control treatment was adequate for highlighting factual knowledge acquisition and near transfer differences associated with these different hypertext designs). It is important to note that the core hypertext design features used in all three treatment groups in this study (e.g., case-based hypertext materials, multiple abstract thematic perspectives, thematic commentaries, nonlinear exploration of content) differ from those typically found in most instructional hypertext programs (e.g., fixed link access to nodes of information). Future research could specifically compare learning in terms of near and far knowledge transfer associated with case-based/conceptually-indexed hypertext designs and other hypertext design approaches.
This study also provides further evidence that epistemic beliefs are an important factor associated with learning for transfer in a hypertext environment. The students with complex EBP in the Guided TCC performed at the highest level on the knowledge synthesis essays. Also, in the tests of far within-domain transfer, the problem-solving essays, the complex EBP students in all groups on average scored higher than the simple EBP students. These results are generally consistent with the findings that EBP students in the Thematic Criss-Crossing Hypertext group achieved significantly higher scores on the transfer task than simple EBP students in the same group (Jacobson, 1990; Jacobson & Spiro, 1995), and with other research documenting the influence of epistemic or epistemological beliefs on learning (Feltovich et al., 1989; Schoenfeld, 1983; Schommer, 1993; Schommer, 1994). However, more work is needed to validate the different methods being used by researchers to assess epistemic or epistemological beliefs and to more systematically explore the influence of these belief structures on knowledge acquisition and transfer in technology-based and nontechnology-based learning environments.
Contrary to our expectations of a linear relationship about the degree of user control (from low to high) and attitudes towards using the program (negative to positive), the students in the Free Hypertext Exploration group, which permitted the greatest amount of user control over navigating through the hypertext, had the lowest average attitude scores of the three groups. The most positive attitude scores were those held by the Learner Selected TCC group, with the Guided TCC group having attitude scores in the middle. These findings indicate that there is not necessarily a direct correspondence between user control over a hypertext environment and positive attitudes towards using such a system. Compared to the students in the Free Hypertext Exploration procedure (the high learner control without modeling or scaffolding), the Learner Selected TCC procedure (with moderate amount of learner control and scaffolding) may have helped support these students in their use of the hypertext and therefore contributed to their more positive attitudes.
It is interesting to note that although the Free Hypertext Exploration and Learner Selected TCC groups held different attitudes towards using the program, the students in these groups performed in a similar manner on the learner outcome measures compared to the Guided TCC group (i.e., significantly higher scores on the short answer items and lower scores on the knowledge synthesis essays). Another surprising finding was that students with a complex set of epistemic beliefs had more negative attitudes about the program than those students with a simpler set of epistemic beliefs. The complex EBP students had been expected to have more positive attitudes about using the two experimental thematic criss-crossing treatments than the simple EBP students. Perhaps those students with complex epistemic beliefs, who overall had higher scores on the problem-solving essay tasks and who performed better on the knowledge synthesis task when in the Guided TCC group, were not necessarily cognizant that they were effectively learning for transfer. If so, this suggests that students may not have self-monitoring or metacognitive skills for ascertaining their success at learning for near and/or far transfer when using nonlinear environments such as hypertext. There may thus be a need for hypertext design elements or support in the social learning environment in which the hypertext is used to help students develop an appropriate metacognitive awareness of their learning when using such systems.
We have thus far discussed specific findings from this study. Now we consider more general issues related to metaphors for learning and hypertext. In CF theory, a central metaphor for learning complex knowledge, adapted from Wittgenstein, is that of "criss-crossing conceptual landscapes" (Spiro et al., 1992; Spiro & Jehng, 1990; Spiro et al., 1987; Wittgenstein, 1953). This metaphor suggests that advanced learning requires "criss-crossing" a complex domain from multiple intellectual perspectives, just as coming to know a rich topographical landscape requires criss-crossing that terrain using multiple paths and viewing it from multiple vantage points.
There is, of course, a "three-dimensional" aspect of a landscape. By extension, then, one may metaphorically regard a "three-dimensional" conceptual landscape as consisting of a two dimensional surface area (i.e., surface features of a case or a situated context) with a structural component along a third dimension (i.e., abstract cognitive or mental models, schemas, conceptual knowledge). We suggest that for substantive learning to result from using a hypertext environment, there must be three-dimensional criss-crossings of conceptual landscapes. Such criss-crossings must make salient to the learner the important structural components of knowledge in a domain and how those structural knowledge components exist in situated contexts with varying surface features. Our analysis of the data in this study is consistent with this perspective. The Guided TCC treatment helped promote learning along this "deeper" structural dimension as seen by the higher knowledge synthesis essay scores of the complex EBP students who used this learning activity. However, it is not enough to just hypertextually criss-cross different cases or situated contexts as students may only superficially learn about two dimensional or surface features of the conceptual landscapes. This is evidently what happened to students in the Free Hypertext Exploration and Learner Selected TCC groups. Unfortunately, these students primarily acquired low level, case-specific factual knowledge, and they did less well on the knowledge synthesis tests requiring a conceptual/structural level understanding of the instructional content. Merely "criss-crossing" using hypertext links is not enough to acquire an interconnected "situated structural" understanding of a complex topic.
Let us also consider a scenario where learners are in a domain for which they lacked the requisite conceptual knowledge or the mental models used to generate the conceptual indexing of the cases in the hypertext learning environment. We would expect that lacking the cognitive tools to "see" or understand the conceptual/indexical third dimension, the conceptual landscape would collapse to two dimensional "surface features" of the different cases. The learners could still engage in criss-crossing conceptual landscapes as the system automatically scaffolds the learners by providing the themes and re-editing the link/node structure for the case sections. But they would be cognitively "blind" to the underlying structural dimension of knowledge as they would only be cognizant of the surface features of the cases. Hence, it is important that the overall learning environment in which the hypertext and hypermedia systems are used be constructed to insure the salience of the three dimensional conceptual landscape to the learners.
This paper has considered theoretical and research issues associated with the design and use of hypertext learning environments. Overall, this study found evidence that a hypertext with conceptually-indexed case-based materials and modeling/scaffolding support can help students acquire and flexibly use complex knowledge. We view the results of this study related to knowledge transfer and learning from hypertext as encouraging, but provisional and in need of further research. However, we feel these results are educationally significant as there continues to be little rigorous empirical documentation of the educational efficacy of hypertext learning environments and the conditions under which significant learning outcomes with such systems are achieved. Also, we are hopeful that the methodology we are using--which attempts to systematically investigate different hypertext design features based on cognitive theory and research, to consider the influence of cognitive factors such as epistemic beliefs, and to assess learning using a range of knowledge acquisition and knowledge transfer measures--will continue to prove useful in identifying important factors associated with learning in hypertext learning environments. There is certainly much more work to be done.
Anderson, J. R., Conrad, F. G., & Corbett, A. T. (1989). Skill acquisition and the LISP tutor. Cognitive Science, 13, 467-505.
Anderson, R. C., & Freebody, P. (1979). Vocabulary knowledge (Technical Report No. 136). Champaign, IL: University of Illinois, Center for the Study of Reading.
Baker, E. L., Niemi, D., & Herl, H. (1994). Using Hypercard technology to measure understanding. In E. L. Baker, & H. F. O'Neil, Jr. (Eds.), Technology assessment in education and training (pp. 133-152). Hillsdale, NJ: Lawrence Erlbaum.
Beeman, W. O. , Anderson, K. T., Bader, G., Larkin, J., McClard, A. P., McQuillan, P. J., & Shields, M. (1987). Hypertext and pluralism: From lineal to non-lineal thinking. In Hypertext '87 Papers (pp. 67-88). Chapel Hill: The University of North Carolina.
Beeman, W. O., Anderson, K. T., Bader, G., Larkin, J., McClard, A. P., McQuillan, P. J., & Shields, M. (1988). Intermedia: A case study of innovation in higher education (Final report to the Annenberg/CPB Project). Providence, RI: Brown University, Office of Program Analysis, Institute for Research in Information and Scholarship.
Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learning. Educational Researcher, 18(1), 32-42.
Chi, M. T. H., Glaser, R., & Farr, M. J. (Eds.). (1988). The nature of expertise. Hillsdale, NJ: Lawrence Erlbaum.
Clancey, W. J. (1993). Situated action: A neuropsychological interpretation response to Vera and Simon. Cognitive Science, 17, 87-116.
Cognition and Technology Group at Vanderbilt. (1990). Anchored instruction and its relationship to situated cognition. Educational Researcher, 19(6), 2-10.
Cognition and Technology Group at Vanderbilt. (1992). Technology and the design of generative learning environments. In T. M. Duffy, & D. H. Jonassen (Eds.), Constructivism and the technology of instruction: A conversation (pp. 77-89). Hillsdale, NJ: Erlbaum.
Collins, A. (1989). Cognitive apprenticeship and instructional technology (Technical Report No. 474). University of Illinois at Urbana-Champaign, Center for the Study of Reading.
Collins, A., Brown, J., & Newman, S. (1987). Cognitive apprenticeship: Teaching the crafts of reading, writing, and mathematics. In L. Resnick (Ed.), Knowing, learning, and instruction. Hillsdale, NJ: Erlbaum.
Collins, A., & Brown, J. S. (1988). The computer as a tool for learning through reflection. In H. Mandl, & A. Lesgold (Eds.), Learning issues for intelligent tutoring systems. New York: Springer-Verlag.
Collins, A., Brown, J. S., & Holum, A. (1991). Cognitive apprenticeship: Making thinking visible. American Educator, 15(3), 6-11, 38-46.
Conklin, J. (1987). Hypertext: An introduction and survey. IEEE Computer, 20(9), 17-41.
Edelman, G. (1992). Bright air, brilliant fire: On the matter of the mind. New York: Basic Books.
Feltovich, P. J., Spiro, R. J., & Coulson, R. L. (1989). The nature of conceptual understanding in biomedicine: The deep structure of complex ideas and the development of misconceptions. In D. Evans, & V. Patel (Eds.), The cognitive sciences in medicine (pp. 113-172). Cambridge, MA: MIT Press (Bradford Books).
French, J. W., Ekstrom, R. B., & Price, L. A. (1963). Kit of reference tests for cognitive factors. Princeton, NJ: Educational Testing Service.
Garner, R., Alexander, P. A., Gillingham, M. G., Kulikowich, J. M., & Brown, R. (1991). Interest and learning from text. American Educational Research Journal, 28(3), 643-659.
Gentner, D. (1983). Structure mapping: A theoretical framework for analogy. Cognitive Science, 7, 155-170.
Gick, M. L., & Holyoak, K. J. (1987). The cognitive basis of knowledge transfer. In S. M. Cormier, & J. D. Hagman (Eds.), Transfer of learning: Contemporary research and applications (pp. 9-46). New York: Academic Press.
Greeno, J. G., & Moore, J. L. (1993). Situativity and symbols: Response to Vera and Simon. Cognitive Science, 17, 49-59.
Jacobson, M. J. (1990). Knowledge acquisition, cognitive flexibility, and the instructional applications of hypertext: A comparison of contrasting designs for computer-enhanced learning environments. Unpublished doctoral dissertation, University of Illinois at Urbana-Champaign.
Jacobson, M. J. (1994). Issues in hypertext and hypermedia research: Toward a framework for linking theory-to-design. Journal of Educational Multimedia and Hypermedia, 3(2), 141-154.
Jacobson, M. J., & Levin, J. A. (1995). Conceptual frameworks for network learning environments: Constructing personal and shared knowledge spaces. International Journal of Educational Telecommunications, 1(4), 367-388.
Jacobson, M. J., & Spiro, R. J. (1995). Hypertext learning environments, cognitive flexibility, and the transfer of complex knowledge: An empirical investigation. Journal of Educational Computing Research, 12(5), 301-333.
Jehng, J. C., Johnson, S. D., & Anderson, R. C. (1993). Schooling and students' epistemological beliefs about learning. Contemporary Educational Psychology, 18, 23-35.
Jonassen, D. H. (1986). Hypertext principles for text and courseware design. Educational Psychologist, 21(4), 269-292.
Jonassen, D. H., Grabowski, Barbara, & Yacci, Michael. (1993). Structural knowledge: Techniques for representing, conveying, and acquiring structural knowledge. Hillsdale, NJ: Erlbaum.
Jonassen, D. H., & Wang, S. (1993). Acquiring structural knowledge from semantically structured hypertext. Journal of Computer-based Instruction, 20(1), 1-8.
Kearsley, G. (1988). Authoring considerations for hypertext. Educational Technology, 28(11), 21-24.
Kirk, R. E. (1982). Experimental design: Procedures for the behavioral sciences (2nd ed.). Monterey, CA: Brooks/Cole.
Koladner, J. (1993). Case-based reasoning. San Mateo, CA: Morgan Kaufmann.
Lave, J. (1988). Cognition in practice: Mind, mathematics and culture in everyday life. Cambridge: Cambridge University Press.
Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation. Cambridge: Cambridge University Press.
Lehrer, R. (1993). Authors of knowledge: Patterns of hypermedia design. In S. P. Lajoie, & S. J. Derry (Eds.), Computers as cognitive tools (pp. 197-227). Hillsdale, NJ: Erlbaum.
Lepper, M. R. (1985). Microcomputers in education: Motivational and social issues. American Psychologist, 40(1), 1-18.
Mandl, H., & Lesgold, A. (Eds.). (1988). Learning issues for intelligent tutoring systems. New York: Springer-Verlag.
Moore, J. L., Lin, X., Schwartz, D. L., Petrosino, A., Hickey, D. T., Compbell, O., Hmelo, C., & Cognition and Technology Group at Vanderbilt. (1994). The relationship between situated cognition and anchored instruction: A response to Tripp. Educational Technology, 34(8), 28-32.
Norman, D. (1993). Cognition in the head and in the world: An introduction to the special issue on situated action. Cognitive Science, 17(1), 1-6.
Pea, R. D., & Sheingold, K. (Eds.). (1987). Mirrors of minds: Patterns of experience in educational computing. Norwood, NJ: Ablex.
Perry, W. G., Jr. (1968). Pattern of development in thought and values of students in liberal arts college: A validation of a scheme. Harvard University, Bureau of Study Counsel. [ERIC Document Reproduction Service No. ED 024315].
Scardamalia, M., & Bereiter, C. (1991). Higher levels of agency for children in knowledge building: A challenge for the design of new knowledge media. The Journal of the Learning Sciences, 1(1), 37-68.
Schank, R. C. (1982). Dynamic memory. Cambridge: Cambridge University Press.
Schoenfeld, A. H. (1983). Beyond the purely cognitive: Belief systems, social cognitions, and metacognitions as driving forces in intellectual performance. Cognitive Science, 7, 329-363.
Schommer, M. (1993). Epistemological development and academic performance among secondary students. Journal of Educational Psychology, 85(3), 406-411.
Schommer, M. (1994). An emerging conceptualization of epistemological beliefs and their role in learning. In R. Garner, & P. A. Alexander (Eds.), Beliefs about text and instruction with text (pp. 25-40). Hillsdale, NJ: Erlbaum.
Shapiro, A. M. (1994). Complex concept acquisition and the representation of knowledge: A study in hypermedia-aided instruction. Paper submitted for publication.
Spiro, R. J., Coulson, R. L., Feltovich, P. J., & Anderson, D. K. (1988). Cognitive flexibility theory: Advanced knowledge acquisition in ill-structured domains. In Tenth annual conference of the Cognitive Science Society (pp. 375-383). Hillsdale, NJ: Erlbaum.
Spiro, R. J., Feltovich, P. J., Jacobson, M. J., & Coulson, R. L. (1992). Cognitive flexibility, constructivism, and hypertext: Random access instruction for advanced knowledge acquisition in ill-structured domains. In T. M. Duffy, & D. H. Jonassen (Eds.), Constructivism and the technology of instruction: A conversation (pp. 57-75). Hillsdale, NJ: Erlbaum.
Spiro, R. J., & Jehng, J. C. (1990). Cognitive flexibility and hypertext: Theory and technology for the nonlinear and multidimensional traversal of complex subject matter. In D. Nix, & R. Spiro (Eds.), Cognition, education, and multimedia (pp. 163-205). Hillsdale, NJ: Erlbaum.
Spiro, R. J., Vispoel, W. P., Schmitz, J. G., Samarapungavan, A., & Boerger, A. E. (1987). Knowledge acquisition for application: Cognitive flexibility and transfer in complex content domains. In B. K. Britton, & S. M. Glynn (Eds.), Executive control processes in reading (pp. 177-199). Hillsdale, NJ: Erlbaum.
Suchman, L. (1987). Plans and situated actions: The problem of human-machine communication. Cambridge: Cambridge University Press.
Vosniadou, S., DeCorte, E., Glaser, R., & Mandl, H. (in press). International perspectives on the psychological foundations of technology-based learning environments. Hillsdale, NJ: Erlbaum.
Vosniadou, S., & Ortony, A. (Eds.). (1989). Similarity and analogical reasoning. New York: Cambridge University Press.
White, B. Y. (1993). ThinkerTools: Causal models, conceptual change, and science education. Cognition and Instruction, 10(1), 1-100.
White, B. Y, & Frederiksen, J. R. (1987). Qualitative models and intelligent learning environments. In R. Lawler, & M. Yazdani (Eds.), AI and education: Learning environments and intelligent tutoring systems (pp. 281-305). Norwood, NJ: Ablex Publishing.
Wittgenstein, L. (1953). Philosophical investigations. New York: Macmillan.
 This is an updated version of the paper that originally appeared in the Journal of Educational Multimedia and Hypermedia, 4(4), 321-364. Back
 There has also been important research in the field of artificial intelligence dealing with indexing as a key factor for understanding the dynamic nature of memory structures and how computational case-based reasoning systems can be developed (Koladner, 1993; Schank, 1982).Back
 The notion of indexicality in situated learning is similar to the conceptual indexing of case-based hypertext materials, see below.Back
 For other discussions of SC learning activities, see (Cognition and Technology Group at Vanderbilt, 1990; Collins et al., 1991; Moore et al., 1994).Back
 Fixed links are simply links that remain the same in a hypertext, whereas variable links refer to links that change or vary depending on specific conditions or contexts. We should also distinguish between "on-the-fly" links based on rapid computational algorithms searching through large hypertext databases and the variable links based on conceptual indices used in our system. Computed or "on-the-fly" link engines, such as those now available on the World Wide Web, create links that were not predetermined or stored in the system. In contrast, variable links, such as we used in this study, are determined from prespecified conceptual indices based on a detailed analysis of expert knowledge in a domain. Although links of this type are not as user adaptable as "on-the-fly" computed links, variable links do provide the learner with considerable flexibility in combining or assembling knowledge from the hypertext-base through numerous possible combinations of different thematic indices and cases. We feel the research on conceptually-indexed hypertext or hypermedia systems, such as this study, is suggestive of how these systems can create viable learning environments for students. However, learning activities using hypertextual environments with "on-the-fly" links which lack the modeling and scaffolding features of conceptually-indexed hypertext designs probably will impose greater cognitive and metacognitive demands on the learner and could in turn result in learning difficulties and frustrations for the users. Given "on-the-fly" computed links based on student interactions and queries are now becoming available, the conditions under which such hypertext linking functionality might be effectively used to promote significant learning (in contrast to simple information access) should be carefully explored in future research. Back
 However, fixed hypertext links are quite appropriate for certain purposes, such as providing nonlinear hypertext information access in an on-line help module for a word processing program. Back
 The term epistemology is frequently found in the literature. However, epistemology is strictly defined as a field of inquiry into the nature, limits, and validity of knowledge. We favor a broader term, epistemic, which refers to the general characteristics of knowledge and the experience of knowing as a descriptor for our work in this area. Back
 See Jacobson (1990) and Jacobson and Spiro (1995) for a more complete description of the hypertext program and the domain content. Back
 We regard the social impact of technology as an "ill-structured" knowledge domain as there is variable coverage of the wide scope themes with respect to the case sections. For a discussion of domain conceptual structure, see Jacobson and Spiro (1994). Back
 Guided Thematic Criss-Crossing did permit more user control than in the hypertext Thematic Criss-Crossing treatment in the previous study. In that study, the students were required to read all of the case sections and thematic commentaries for the selected themes to insure common coverage of the instructional content. The present study allowed the students to choose the case sections and case-theme commentaries they wanted to read. Back
 A free recall task was also administered in which the students read a new technology and society passage and then wrote down as much as they remembered in a three minute period. Given space limitations, a more detailed discussion of this dependent measure is omitted as no significant differences between the groups were found. Back
 One may have a concern that paying students is not appropriate in a study exploring a situated perspective on learning. We feel this is a suitable methodological procedure for a study of this type. As Brown et al. (1989) note, the research on situativity and learning is primarily intended to uncover the conditions under which successful learning occurs. Motivation is one such factor that has been identified as necessary for successful learning (Lepper, 1985). How can a researcher influence students to participate in a long experimental study that is not related to a class they are taking? In such circumstances, paying students for their time and effort is certainly one effective motivational technique. Furthermore, "learning for pay" is actually very common in the "real world." Most people find it necessary to learn in job situations for which they are being paid, things that they would probably otherwise have little reason to want to learn. Back
 Unfortunately, on-line data for 24 students was lost due to computer problems. Back
 A two-way analysis of variance revealed a significant interaction with the Ease subscale with Group and EBP (F2,63 = 3.197, p= .05), however a Scheffe post hoc test found no significant pairwise mean differences. This was probably due to the unequal number of students in the cells and because the Scheffe test is a conservative post hoc test. Back
 One might argue that the Learner Selected TCC students did not perform at a higher level on the knowledge synthesis essays because of experimental constraints on using the system. In this study, the Learner Selected TCC students were allowed to do multiple theme searches for the questions, but they could only read and study the final searches' list of case section cards. We would have preferred not to impose this constraint on theme searching, but the amount of time available in the computer laboratories was