Knowledge construction and knowledge representation in high school students' design of hypermedia documents.

This study documented the processes of knowledge construction and knowledge representation in high school students' hypermedia design projects. Analysis of knowledge construction in linking and structural building yielded distinct types and subtypes of hypermedia documents, which were characterized by four features of knowledge representation: (a) typography, (b) associativity, (c) nonlinearity, and (d) abstraction. Students were found to have diverse knowledge construction processes and considerable variation in their ability to use hypermedia for knowledge representation. Their perceptions of the design experience suggested that difficulty in using hypermedia for conceptual building was related to their lack of experience in intentionally constructing a structural representation. Designing hypermedia documents for content area learning allowed students to be authors of knowledge and thus provided opportunities to develop effective cognitive and knowledge representational skills. This study provided initial evidence of the nature of complex learning in constructing hypermedia documents. Some processes involved in the transformation of knowledge were discussed. However, further research is needed to provide full understanding of the nature of this transformation.

**********

Knowledge is a structure with a purpose (Perkins, 1986). When learners "construct their own knowledge" they are constructing a structure that has meaning to them, and when they have accomplished this, it is believed that they understand something better than they did before. The student-as-designer or designing-to-learn point of view (Ehrmann & Balestri, 1992; Wilhelm, Friedemann, & Erickson, 1998) applies this idea of knowledge construction to student construction of artifacts that embody their understanding of certain concepts. For example, science students who are able to successfully support 50 pounds on a bridge constructed from popsicle sticks illustrate their understanding of certain principles of physics. Projects of this sort appear to be very motivating to learners, yet the difficulties in assessing exactly what is learned make many teachers hesitant to take on the promises and challenges of project-based learning.

A number of authors (e.g., Lebrer, 1993; Lehrer, Erickson, & Connell, 1994; Wilhelm, Friedemann, & Erickson, 1998) suggested that the design of hypermedia artifacts is another way for learners to represent their understanding. Jonassen and his colleagues (Jonassen, Myers, & McKillop, 1996; Jonassen & Carr, 2000) believed that the authoring of hypermedia artifacts is among the most engaging constructivist activities. This was certainly found to be the case, particularly at the high school level, in the Doing Science project (McGrath et al., 1996/1997) and in at-risk students' learning and motivation (Liu & Rutledge, 1997; McGrath, Sylvester, & Chen, 1999).

Making Thinking Explicit

Hypermedia authoring is one of a number of design processes that can help learners make their thinking and understanding visible (Carlson, 1990; Dede, 1992; McFarlane, Williams, & Bonnett, 2000), and thus better open to critical examination. When used as a tool to link together conceptual elements, hypermedia allows many ways of organizing information, looking at the content, and representing knowledge structures and conceptual relationships (Jonassen, 1988). In this way, hypermedia appears ideal for the constructive, creative, recursive, and analytical elements of design (Ehrmann & Balestri, 1992). In addition, hypermedia can support collaboration, metacognition, and the development of higher-level information processing strategies (Carlson, 1990).

Rumelhart and Norman's (1978) three modes of learning, accretion, restructuring, and tuning, provided a useful framework for identifying different levels of learning. Accretion refers to the accumulation of knowledge in one's knowledge base; restructuring refers to the reorganization of one's existing knowledge structure; and tuning refers to a continuing modification of the existing knowledge structure to improve the accuracy, generalizability, and specificity. While accumulation yields quantitative changes of one's knowledge, restructuring and tuning account for the emergent quality of complex learning.

McAleese (1985) described three problems of knowledge representation during authoring: exteriorization, anomalous state metacognition, and self-confrontation. Exteriorization requires making one's implicit knowledge structure explicit by communicating or explaining a knowledge domain. This process entails an anomalous state metacognition, which forces authors to become aware of their own state of knowledge and resolve inconsistencies in their underlying knowledge structures. This state may result in cognitive dissonance and authors' self-confrontation to accommodate "the internal self with the external realization" (p. 303). Through exteriorization, metacognition, and self-confrontation, authors are forced to modify or reorganize their knowledge structures.

Hypermedia artifacts also present a particularly interesting case of designing to learn because the document itself provides a trail or map of the learner's thought processes through the content, arrangements, and links among concepts. There is very little research on how to make use of the document itself to assess and understand student learning (McFarlane et al., 2000), although it would appear that the nodes and links should provide researchers and teachers with a good picture of the concepts and connections that the student has constructed.

Knowledge Design Skills

Carver, Lehrer, Connell, and Erickson (1992) addressed various skills needed for student designers: project management, research, organization and representation, presentation, and reflection skills. Subsequently, Leiner (1993) and colleagues (Lehrer et al., 1994) proposed four primary types of cognitive processes involved in the construction of hypermedia documents: planning, transforming, evaluating, and revising. Planning requires students to define the nature of the problem for their project, decompose the problem into sub-components, and assign roles and tasks for collaborative authorship. Next, students need to transform information into knowledge, which "is distinguished from information by its systematicity--a network of relationships that sustain one or more themes" (Lehrer et al., 1994, p. 229). In this phase, students need to collect information, develop new information, interpret and organize information (e.g., segment and link information), and represent their understandings. Finally, evaluation and revision require students to communicate effectively with others about their intentions and goals for the projects, elicit feedback from others, and take actions to improve their projects.

Problems with Student-Authored Hypermedia Artifacts

Farrow (1993) found that students' hypermedia documents generally were rather primitive and had limited educational value to other students. Analyzing children's hypermedia composition, Nicol (1989-90) concluded that students "need new and more engaging hypermedia models for presenting information" (p. 16). Following Nicol's pioneer work, Hay, Guzdial, Jackson, Boyle, and Soloway (1994) analyzed high school students' hypermedia composition across various disciplines. They observed two categories of writing styles: annotated text (AT) and integrated composition (IC). While AT documents are similar to text-only documents with links providing parallel information, IC documents distribute information among the text and the content of the links. Overall, they found students produced more AT documents than IC documents, and noted that "the lack of structure and/or models led to a number of students just putting media together without much thought" (p. 316). More recently, Erickson and Lebrer (2000) investigated stu dents' conceptions of links over time and suggested that student designers need to know certain functions of links to effectively transform the knowledge content.

Although hypermedia holds the promise of adapting itself to the individual's cognitive skills, Joyce (1988) cautioned that "the ready adaptability of these [hypermedia] tools to more traditional uses is especially compelling given the technological frosting they so easily spread upon stale cake" (pp. 10-1 1). Park (1991) also pointed out the difficulty of representing dynamic human cognition as simple and static structures of nodes and links. Dede (1988) concluded that more sophisticated skills were needed for representing knowledge based on hypermedia. Thus, to use hypermedia technology appropriately and effectively in the constructivist paradigm, it is important to examine how the characteristics of hypermedia affect the nature of students' knowledge construction and knowledge representation processes.

Problems with Early Research

There are serious problems with most early studies of student-created hypermedia artifacts. First, one-time samplings of the first time a student or group has constructed a hypermedia document tend to yield rather linear projects done in a traditional way but with the addition of other media (i.e., AT document, Hay et al., 1994). Usually students haven't worked with either the technologies, the construction of hypermedia documents, or project-based learning, and they must learn a number of new skills as well as "buy into" a new way of learning (Perkins, 1991). Thus, little idea of how learners create such documents after they become good at it is known. Although recent studies have emphasized repeated hypermedia design experiences and appropriate instructional contexts (Erickson & Lehrer, 1998; Erickson & Lehrer, 2000), still little is known about both the types and processes involved in creating hypermedia documents.

The purpose of this qualitative study, therefore, was descriptive (Marshall & Rossman, 1995): to document the constructive process of creating collaborative hypermedia artifacts through participant observation, in-depth interviewing, and document analysis. Hypermedia concepts as well as methods and conventions in hypermedia design were emphasized within the student-as-designer context, in which participants' design activities and their emerging hypermedia artifacts were recorded and analyzed. This allowed exploration of the nature of knowledge construction and knowledge representation within the hypermedia environment, and the expansion of current knowledge about the resulting types of hypermedia documents created by students. This study was gnided by two research questions:

(1.) What is the nature and individual process of knowledge construction in designing hypermedia documents?

(2.) What are the characteristics of knowledge representation in the types of hypermedia documents created by students?

METHOD

Participants

This study was conducted in a high school located in a midwestern city. Sixteen tenth graders (seven females and nine males, including nine Caucasians, six African Americans, and one Hispanic) enrolled in a sophomore level science course were the primary participants. Additionally, descriptive data obtained from 13 other students in a different section of the course (taught by the same teacher) were included to provide a rich account of student perspectives. These students (eight females and five males) included six Caucasians, six African Americans, and one Hispanic. In total, 29 students participated in this study.

Due to the teacher's effort to provide a constructivist learning environment, the students had been familiar with a constructivist-oriented, technologically-supported learning environment prior to the study. Most of the students in this group had had some experience of creating collaborative hypermedia artifacts such as Word Wide Web (WWW or Web) projects. Thus, this group of students provided a unique opportunity to observe their constructive use of hypermedia beyond their initial reactions to both the technology and a constructivist learning environment. In addition, the teacher, a doctoral student in a nearby university, was experienced in project-based learning, technology, and constructivist pedagogy.

Instructional Setting

Students were assigned by the teacher to work in pairs. They met in the classroom for 90 minutes two-three times a week (depending on the block schedule) for two months. The 90-minute class period was divided into two 45-minute sections. During each half of the class period, some design teams worked on the computer, while the other teams read the assigned materials, researched and synthesized information, engaged in group discussions, and planned/designed their projects.

Each team was asked to design a hypermedia document to represent their understanding of all the concepts/topics covered during the subunit on Water. The requirements of their projects included: (a) demonstrate understanding of the concepts studied in class, (b) show the interrelatedness of concepts and methods of linking ideas, and (c) understand that the knowledge base generated by each team will be used by future students as a basis for their learning and understanding. Students were asked to save their projects on a diskette at the end of each class to provide electronic records of their knowledge construction process. They had access to eight IBM-compatible computers, a color scanner, and a commercial database of graphics and photos. Students were encouraged to use the Internet access in the school's library to research information. Storyspace 1.2 (Bemstein, 1994) was used as the hypermedia authoring tool. Storyspace was chosen for this study because of its ease for linking and organizing ideas. This tool provides four types of views (map, chart, treemap, and outline) for alternative representation and visualization of ideas. Items contained in each view can be easily manipulated. In this study most students chose to work either in the map view or chart view. While the tree diagram in chart view (Figure 4) displays a strictly hierarchical format, the map view (Figure 7) presents a web structure of information connected by named or unnamed links.

Prior to the study, students were given an opportunity to complete a small project using Storyspace (1994). To further guide students in the hypermedia authoring environment, several "focus sessions" were held at the beginning of the study. These sessions included discussion of hypermedia concepts and demonstration of examples and techniques for designing hypermedia (1). To assist students with their roles as knowledge designers, each design team was asked to keep a designer's notebook, in which they described their major goals and responsibilities, planned their ideas, and designed outlines or concept maps for each class session. The major purpose was to support students' linking of ideas and reflecting upon the organization of the knowledge content. The notebook also provided data on students' perceptions of the learning environment as they were asked to report problems and difficulties they encountered. Finally, to facilitate cognitive processes of planning, transforming, evaluating, and revising (Lebrer, 1993; Lehrer, Erickson, & Connell, 1994) in designing hypermedia, the teacher provided feedback and suggestions, monitored group and individual performance, and identified areas of the student projects that needed improvement. At the end of the study, they presented their projects to their peers and a group of guests, including teachers, university students, and school district staff members.

Data Sources

Data sources included student projects, designer's notebook, observations, interviews, teacher journal, and student self-report. Two major steps of data analysis involved compiling a case file for each design team and exploring common themes, patterns, regularities, and irregularities in data through systematic comparisons and contrasts across case files and data sources. Eight projects were sampled throughout the eight-week period. At each point of observation, each project's structural representations, including map view and chart view in Storyspace (1994), were captured into image files. The printouts of these files were compiled into a notebook and used to inspect changes attempted by each design team, and to make comparisons among groups.

Various data sources were further analyzed from two angles: knowledge construction and knowledge representation. With respect to knowledge construction, for example, the designer's notebook and student projects provided both quantitative and qualitative information. Quantitative data (e.g., word count, node count, link count, and the number of levels in students' knowledge representation) were summarized and tabulated. This information represented quantitative changes of students' knowledge structures and gave clues as to the nature of their construction processes. Qualitative data included students' use of integration techniques (i.e., types of links), changes in structural representations over time, and problems/difficulties as perceived by students. The serial visual inspections of student projects were particularly useful to observe the process of changes during knowledge construction. With respect to knowledge representation, characteristics of nodes, links, and organization in students' design were exa mined. In particular, characteristics of organization were analyzed in terms of the patterns and the types of structural representations.

RESULTS

Hypermedia Authoring and Knowledge Construction

By examining how students constructed the basic elements of hypertext (i.e., nodes and links), we observed the processes of knowledge construction unfolding. Table 1 shows eight design teams' weekly production of words, nodes, and links throughout the study. Figures 1-3 show patterns of word, node, and link construction, respectively over the eight-week period.

Patterns of word, node, and link construction. While Figure i shows a similar trend of word production in all design teams, Figure 2 reveals specific tuning process in some design teams. This process involved an ongoing modification of the existing nodes. An example of the ways the nodes were arranged and modified in the chart view of Storyspace (1994) during the early phase of the project can be seen in Figure 4. This sort of tuning process was consistently observed throughout the study.

The greatest inter-group diversity was observed in the patterns of link construction (Figures 3a & 3b). While Figure 3a shows a similar increase in link construction in four design teams, Figure 3b displays a more distinctive linking process in each team. Three of these teams (1, 2, 5) concentrated on the accumulation of information initially and put off the task of linking, which we refer to as the strategy of suspending linking (Figure 5 for an example).

Types of hypermedia documents. To understand the types of projects created by students, we analyzed the types of links and the organization of nodes and links in each hypermedia document. In Storyspace (1994), links can be created between nodes or specific sections of text or images within a node. Storyspace distinguishes these two types of links as basic link and text link, which are analogous to the sequential links and relational links as described by Grabinger and Dunlap (1996). While sequential links provide a single path between related nodes, relational links permit multiple paths to related information. Examples of relational links include associative links, elaborative links, and hierarchical links (Grabinger & Dunlap, 1996). The organization of nodes and links determines whether the hypertext is unstructured or structured. While unstructured hypertext can be described as "random, node-link hypertext" (Jonassen, 1989, p. 51), structured hypertext "consists of sets of nodes, each set accessible from a ny other set. Each set is explicitly arranged to depict the structure of the information" (p. 52).

Initial analysis of the dominant link type used by each design team resulted in two sequential style documents and six relational style documents. Further examination of the structural representations in the six relational style documents revealed two types of knowledge construction process: relational-unstructured and relational-structured. Figure 6 provides a complete analysis of types of knowledge construction processes.

Sequential construction. Two design teams (2, 3) exemplified this linear approach of knowledge construction process. This process involved the accumulation of factual information and the forming of a single linkage between each category of information. It can be predicted that this process would produce an equal number of nodes and links when all nodes were linked together. The project created by Team 3 was a representative case of this process. This team organized information into a circle with an identical number of nodes and links in the final project (Table 1).

Unlike Team 3's parallel construction of nodes and links, the construction of nodes in Team 2 preceded the forming of links between each node. Related concepts were grouped into a series of boxes and then successively linked together. There was no connection made between different series of boxes. This approach made the integration of concepts difficult. As a result, this team was able to connect some conceptual elements, but failed to put together all parts of the project. This is evident in their construction of fewer links than nodes (Table 1).

Relational-unstructured (R-U) construction. This process of knowledge construction resulted in a network style of information organization, in which conceptual elements were connected by relational links without being framed into a particular structure. Two design teams (5, 6) illustrated this style of knowledge construction, but differed in some aspects of design. For Team 6, the initial design involved the establishment of a foundation (in the form of introduction to the knowledge base), and then the construction of new nodes and links proceeded simultaneously. As the project expanded, the foundation was used for mapping new information. This unstructured and parallel construction process resulted in rapid and concurrent expansion of words, nodes, and links. Although the established foundation allowed this team to maintain a manageable approach in the construction of its knowledge web, the rapidly expanding nodes and links made it difficult for them to organize their knowledge base.

Similar to Team 2 described earlier, Team 5 engaged in accumulation of factual information initially and postponed the task of linking until the second half of the project. Unlike Team 2 's arrangement of unlinked nodes with a fixed sequence, however, Team 5 constantly reorganized unlinked nodes with relational links. Without a conceptual model or a foundation for accommodating new information, this team became increasingly frustrated by the tasks of linking as their knowledge base expanded.

Relational-structured (R-S) construction. The projects of four design teams (1, 4, 7, 8) were identified as structured hypermedia documents. Looking closely at the entire process of design revealed two opposite ways of knowledge construction: top-down process (Teams 4 & 7) and bottom-up process (Teams 1 & 8). Despite their different processes, it is interesting to note that Team 8 eventually developed a similar conceptual model to that of Team 7.

In their top-down approach, both teams 4 and 7 used the concept mapping technique to plan their structural representations. The basic structure of Team 7's design was created during the first week of the project, and detailed substructures were subsequently added. This team described their project as "everything is in a good order." On the other hand, Team 4 experimented with structural design during the early phase of the project. Unsatisfied with its first attempt of a web style structuring, this team started over with a strictly hierarchical format. As the structures of their projects became more explicit, both teams in the top-down, R-S construction not only created more links but also represented concepts with additional abstract structures. For example, the number of hierarchical levels in Team 4's conceptual structure increased from two to four throughout the project.

The bottom-up process in R-S construction involved the creation of a new conceptual structure after the build-up of nodes and links. Prior to developing an overall conceptual model, Team 8 focused on constructing substructures and experimenting with alternative organizations. Toward the end of the project, a broader scheme emerged to relate all of the substructures. This emergent style of construction involved the creation of a comprehensive structure, which encompassed established substructures. Compared to Team 8's emergent style, the restructuring process in Team 1 was somewhat revolutionary. Near the end of the project, Team 1 created a table of contents to integrate their knowledge base. The resulting structure was completely different from all of its previous forms of organization.

Hypermedia Authoring and Knowledge Representation

Constructing hypertext requires "a capability to act: to create, to change, and to recover particular encounters within the developing body of knowledge. These encounters.. .are maintained as versions, i.e., trails, paths, webs..." (Joyce, 1988, p. 11). To understand knowledge representation within the hypermedia environment, the characteristics of these versions in different types of student projects were analyzed.

Pattern-generating. One important characteristic of students' hypermedia design was the continual process of forming and breaking down patterns or constellations, which often evolved from the visual clustering of unlinked concepts. It is interesting to note that the spatial proximity of these unlinked concepts usually suggested an implicit structure. Subsequently, modules emerged from the linkages among a set of nodes (Kommers, 1996). During this process, conceptual elements were rearranged for more specific representations. The pattern-generating process was evident in different types of student project, but was most extensively observed in the bottom-up, relational structured (R-S) type of construction. The capability for visual representation in Storyspace (1994) appeared to enhance students' pattern-generating process in conceptual representation. One example of this is the use of color-coding techniques by several design teams to visually represent and define various sections of information in their proj ects. The type of Story-space views (map or chart) employed by students also appeared to affect the nature of their pattern-generating and structural representation process.

Structural representations. There were several distinct conceptual models observed in students' projects, including hierarchical, network, and a hybrid of the two. The freedom provided by Storyspace (1994) to move and change links and nodes appeared to encourage variations (and even competition) in students' design of structural representations. Bolter (1991) used the phrase "topographic writing" to describe hypertext authoring. Topographic writing encompasses verbal, visual, and spatial elements. He noted, "It is not the writing of a place, but rather a writing with places, spatially realized topics" (p. 25). In many ways, this distinct feature of hypertext was reflected in students' use of hypermedia for knowledge representation. Analysis of structural representations in students' hypermedia documents revealed four features of knowledge representation: typography, associativity, nonlinearity, and abstraction (Table 2). These features distinguished different types and subtypes of student projects.

As shown in Table 2, all types of student projects encompassed various characteristics of typographical elements and were associative to various extents due to the linking process. The relational types of projects included one essential feature of hypermedia, nonlinearity, which was lacking in the sequential type of project. The feature of abstraction further distinguished R-S projects from R-U projects. In the four R-S projects, the depth of structures was observed with three forms of typographical representation: (a) linear branching out of nodes in a space of two dimensions, (b) hierarchical nesting of nodes in a three-dimensional space, and (c) a combination of two and three-dimensional representations.

Both R-S projects created by top-down techniques (Teams 4 & 7) clearly showed the first two forms of depth representation from the onset of the projects. While Team 4 represented depth with hierarchically nesting of nodes (see Figure 5 for an example), Team 7 provided a representative example of two-dimensional, spatial representation. As shown in Figure 7, the spatially arranged hierarchies represent depth and depict the structure of information in a space of two dimensions. The final format of their representation was hierarchically structured with embedded linear loops.

The third type of depth representation combined nested hierarchies with a web style structure. This was a complex design, especially with respect to linking between nested concepts. It should be noted that most students planned their projects on the designer's notebook using the technique of concept mapping. The nature of their planning was very different from the nesting concept enabled by the map view, in which concepts were hidden beneath other concepts (see Figure 5). The nesting of concepts was confusing for students not only in visual and conceptual representation, but also in navigating between different levels of representations. The transfer of concept maps into Storyspace (1994) was more effortless among design teams that preferred the use of chart view for knowledge representation.

Student Perceptions and Learning

Most students perceived the hypermedia environment as fun and easy to work with, and they preferred constructing hypertext to conventional writing. They viewed hypermedia as ideal for their individual learning styles, by allowing them "taking different paths to get there." Example of student comments included:

I think it's easier to understand when it's spread out like in the different links...it shows the relationship between the links.

With Storyspace you have different places you can go. It's like exploring.

There are separate boxes you can make little links and make it look kind of like an adventure.

As authors of hypermedia, most students recognized the advantages of the hypermedia tool for organizing the learning materials and creating alternative knowledge representations. Even in the sequential style projects, student designers expressed: "Although the links relate to each other they could be organized in a different way;" and "I have seen a lot of different ways the concepts could be linked."

The visual and spatial capabilities of the representational tools (views) were especially appealing to the students:

I like the way you can make your charts and webs on there, and link them all together and make them look like a big spider.

It's good for organization. The visual aids are great because of the visual effects of it. You can see everything lay out and you can see the dimensions it goes into. That's what I think is a plus of it, because it shows you're on this level of information and it goes into depth.

Nevertheless, it should be noted that there were individual differences in students' ability to use hypermedia for knowledge representation. As indicated in the preceding excerpt, some students were able to organize information into higher levels of abstraction and appreciated multiple and varied ways to view the knowledge content. One student described it further:

It's in depth, and there are different things you can do like you touching on highlighted words and they will go into different levels of this project. That's another thing that's great about the computer, is the book is one level, it's one dimensional, you learn one way and one way only, and this, you can see this as many different ways as you want.

Regarding the hypermedia tool as "something you use to learn with," students described the nature of learning as "constantly dealing with everything." Deeper levels of learning appeared to take place, as reflected in the following student remarks:

We really had to work at knowing what goes where and how it connects to make it work.

You had to link them the way that would best describe how the concepts were related to each other.

We have to review it more often...we have to think about how it relates to the other topics.

It's pretty much showed me that there's a lot more to learn than just reading what you see in the book.

Regardless of various types and subtypes of knowledge construction process, all design teams faced a challenge at some points of their projects, although to varying extents. They had to come up with a coherent system to integrate the existing, and often distinct, parts of their projects. A one student put it: "you want to make it the way you're comfortable with--it just flows." The strategy of suspending linking mentioned in the patterns of link construction should be noted here. To achieve coherence, it appears necessary to allow some time and space for certain elements to fit into place. This is reflected in the following student comment:

I was looking at one sub-unit and the subject was really long, and I was going to skip a lot of it, but I went into the next sub-unit and you had to have a lot of the sub-unit before to know what was going on in the next sub-unit. So you've just kind of got to go with all the subunits and look at all of them and decide what you need to have it all run together, because if you're missing something, it would be like well, this doesn't make any sense.

This need for coherence was important in helping students to realize their knowledge lacks. One student described this process as follows:

When the project first started, we had the journals and I had ideas. I had ideas of what I wanted to do with the information I already knew, so I got all this information and I'm going to go straight to the computer. I got to the computer and I got a few things and it looked blank. I was looking at the computer and I needed more information. I was like I can do all this. I can do everything. There's so much I can do, but what's stopping me? And I realize I didn't have the knowledge for it. That's the only way you can do it is to go back and get the book, and that made me see that that book right there, I may not be interested in it because I know what's in that book, but it's going to take me to the next level. There's more depth. I went from a book to a computer and I can combine the two, and now I'm two times more powerful than I was because I know both of them. I know how to use the book, I know how to use the computer, I know how to make them work together which is even bigger key, because without that bo ok, that computer's worthless because the computer only does what I tell it to do.

Knowledge restructuring can be described as "a 'click of comprehension,' a reasonably strong feeling of insight or understanding of a topic that makes a large body of previously acquired (but ill-structured) information fit into place" (Rumelhart & Norman, 1978, p. 38). This experience is illustrated by the following excerpt, in which the student was comparing his project to conventional writing assignments:

Writing assignments and do stuff like that, it does help you learn. But I think doing this project, it just helps you, makes you focus more on the project because you're doing. For instance, the water cycle, how the water cycle is--if you use it and move things around in the computer and put it together, you just kind of figure out how it works because that goes first, that goes second. And then you read it, when you're putting the definitions in, you say, "oh, okay," and that clicks in your head and you know it by heart. But if you're just reading a book and you just read over it once and that's it, you're not going to really care about it next year.

Even though they enjoyed using hypermedia to organize knowledge content, students often perceived this task as the hardest part of the project. As the following excerpt suggested, this difficulty was related to their lack of experience in being authors of knowledge:

I guess maybe we didn't really sit down and figure out how we wanted to organize it. We just had the boxes with the information in it.. .It seems like always before we'd have the information already laid out for us and we just had to learn it and put it onto paper when we had to take a test or whatever...We've just always [gone] from one place to another. We never had to see how they fit together, in between or...

DISCUSSION

Complex Learning in Knowledge Construction

Thee modes of learning, accretion, tuning, and restructuring (Rumelhart & Norman, 1978) were observed in students' design of hypermedia documents. According to Rumelhart and Norman, learning by accretion is a prerequisite of more complex levels of learning. As some "critical mass" of information is accumulated, complex learning involving categorical or structural changes (i.e., tuning and restructuring) may then take place in order to represent information more accurately.

Accretion. Learning by accretion was observed in (a) the continuous increase in the patterns of word production shared by all design teams and (b) the sequential style of knowledge construction process in the two student projects. This learning process demands little effort and represents the simplest mode of learning (Rumelhart & Norman, 1978).

Tuning. Various forms of tuning were observed in the patterns of node and link construction. As a tool to construct and link together nodes, Storyspace (1994) and visualization tools (views) were helpful for students to manipulate, organize, and reorganize related topics. This higher order manipulation, or topical writing in Bolter's (1991) terms, was continually observed in students' fine-tuning of topics into different levels of representations (see Figure 4 for an example).

As noted, the patterns of link construction illustrated the greatest inter-group diversity. This finding indicated the crucial function of tuning in the process of linking. Referring to the structure of hypertext as soft structure, Bolter (1991) suggested that the unstable and ephemeral characteristic of hypertext can foster new ways of organizing and visualizing content matenals. Through linking and necessary tuning, the soft structures of students' hypermedia documents evolved into different conceptual models. During this process, the interrelationships between nodes and links changed constantly, and thus entailed an ongoing tuning process to accommodate new information.

Restructuring. In the four structured hypermedia documents, both top-down and bottom-up construction processes showed some extent of knowledge restructuring. Vosniadon and Brewer's (1987) distinction between weak restructuring and radical restructuring can be used to explain the differences. While weak restructuring involves both forming new relations between concepts and increasing in abstract representation, radical restructuring is typified by the creation of a new conceptual structure. In the top-down, R-S construction, both teams established a framework (2) and then represented concepts with new links and additional abstract structures. These changes met the criteria of weak restructuring as described by Vosniadou and Brewer (1987). On the other hand, the two bottom-up, R-S projects showed the creation of a new conceptual model and illustrated two types (emergent vs. revolutionary) of radical restructuring.

For restructuring to occur, it is essential to recognize the discrepancy between one's knowledge structure and the incoming information (Rumelhart & Norman, 1978). Joyce (1988) believed that "When the author is oneself, apprehending authorial intention becomes a discovery of one's own distinctive structures of thought" (p. 13). This discovery requires the author to focus on perceiving and generating coherent patterns and linkages, and to "use a full range of cognitive skills, especially visual ones, to discover new structures and linkages" (p. 13). Results of this study provided initial evidence of a coherence-seeking tendency during the process of linking and visual/spatial representations. This tendency might have forced students to become aware of the inconsistencies in their understanding. We would argue that the need for coherence is important for complex learning to occur by allowing students to recognize their own knowledge lacks.

Knowledge Representation in Hypermedia Design

Previous studies have suggested that student designers often consider their hypermedia artifacts as incomplete. This is probably due to the soft structure of hypertext. Joyce (1988) described the developing elements of hypertext as "versions of what they are becoming, a structure for what does not yet exist" (p. 11). Such "versions of becoming" provided an opportunity to document the emerging features of students' hypermedia projects.

Typographical representation. The visual and spatial arrangement of nodes and links was probably the most prominent feature in students' knowledge representation, especially within the Storyspace (1994) map view. In all types of student projects, the typographical elements not only assisted student designers in manipulating and generating visual patterns and structures, but also provided variations, interests, and aesthetic considerations in the design process. Thus, it seems reasonable to suggest that typographical representation affords creative expression and permits thought discoveries (Joyce, 1988) in constructing hypermedia knowledge artifacts.

Associativity and nonlinearity. Associativity and nonlinearity received most student attention, as a result of the teacher's emphasis on the essential features of hypermedia in this project. Among the six relational projects, the two relational-unstructured (R-U) projects can best represent the associative and nonlinear features of knowledge representation. The difficulty students had with this type of knowledge representation indicated the vital role of a foundation during knowledge construction (3). With an established foundation, for example, Team 6 appeared to enjoy the associative and nonlinear process of linking. Without a conceptual model or a foundation for mapping new information, Team 5 perceived linking as the most difficult task. Throughout their project, conceptual elements were in a "perpetual state of reorganization" (Bolter, 1991, p. 9) and "the changing patterns or constellations, were in constant danger of breaking down and combining into new patterns" (p. 9). One important implication shoul d be considered here: although associativity and nonlinearity appear likely to afford freedom in conceptual exploration, these features may not be helpful for conceptual integration. Thus, when the purpose of learning is to achieve knowledge integralion, it is important to help students to move beyond associative and nonlinear ways of thinking. The feature of abstraction should be emphasized.

Abstraction. Bolter (1991) pointed out that abstraction allows the author of hypertext to see the structure of the text more clearly, and thus encourages a more global and abstract way of thinking about the text. Joyce (1988) also noted that this ability allows students to recognize "the existence of information below the surface of a writing, and to use that awareness of structure in commonplace fashion to empower themselves" (p. 38). The various types and subtypes of student projects indicated that there were considerable variations in students' ability to perceive and represent abstract concepts with depth. The four R-S projects represented depth with typographical elements, and, notably, the two top-down, R-S projects were especially highly regarded by the other design teams. As noted, this feature of design is closely related to the extent of students' manipulation of higher order representational objects. The authors suggest that the ability to perceive and express abstract concepts with depth is crucia l for developing sophisticated knowledge representation skills within the hypermedia environment. This ability not only freed students from the constraints of linear thinking, but also empowered them with a multidimensional conceptual space for effective construction and communication of knowledge.

Individual differences. It is not clear what factors influenced students' use of the aforementioned features in hypermedia design. There were considerable individual differences in students' ability to perceive and generate patterns in using hypermedia for knowledge representation. These differences were even more evident in using hypermedia to organize information into higher levels of abstraction. The ready adaptability of hypermedia tool for linear uses was found in the two sequential style projects. Comparable to novice learners, these student designers used only "impoverished strategies for synthesizing data into patterns" (Carlson, 1990, p. 18). On the other hand, some design teams demonstrated skills used by expert learners. They were able to recognize "associational patterns in data and, at the same time, organize the data into progressively higher levels of abstraction" (Carlson, 1990, p. 18). Notably, student designers in the two sequential style projects felt that their projects were too limited bo th visually and conceptually following the initial stage of design. Further research is needed to investigate whether novice designers develop more effective strategies for organizing information with repeated hypermedia experiences.

CONCLUSION

Designing hypermedia documents for content area learning allowed students to be authors of knowledge and thus provided opportunities to develop effective cognitive and knowledge representational skills. Students widely perceived hypermedia as "a rich environment to be traversed, manipulated, and experienced" (Carlson, 1990, p. 31) and used a number of features in experimenting and building the structures of their learning experience. Four features of hypermedia design characterized different types and subtypes of knowledge construction processes identified in this study: typography, associativity, nonlinearity, and abstraction. Nevertheless, there were considerable variations in students' incorporation of these features into their design.

"Linking is the heart of hypertext" (Carlson, 1990, P. 1.6). Different approaches to linking appeared to affect the nature of students' knowledge construction and knowledge representation. Further research is needed to account for the different cognitive processes used in linking, and to investigate the factors that affect students' learning and development of knowledge representational skills based on hypermedia. To provide an effective hypermedia design environment, special attention needs to be paid to students' ability to perceive and use various features of hypermedia for knowledge construction and knowledge representation. It is important to define specific cognitive goals in relation to various features of hypermedia design. To assist students in knowledge construction, for example, we need to focus on the feature of abstraction by helping students to recognize and develop coherent patterns in knowledge content, and more importantly, to organize knowledge content into higher levels of abstraction.

Much of previous discussion on hypermedia authoring has emphasized the associative and nonlinear features of hypermedia. It is important to realize that the strength of hypermedia goes beyond its capability to represent human cognition as associative and nonlinear nodes and links. Results of this study indicated that the difficulty in constructive use of hypermedia for conceptual integration was related to the students' lack of experience in intentionally building up a structural representation. Thus, this study supported previous studies concluding that students need new or alternative models for effectively constructing knowledge within hypermedia environments. When the purpose of learning is to achieve complex learning, it appears necessary to empower students with an awareness of structure and allow them the opportunity to practice building it.

In the discussions of the soft structure of hypertext, the authors noted the perpetual process of forming and breaking down of patterns or constellations. Bolter (1991) noted, "the unity or coherence of an electronic text derives from the perpetually shifting relationship among all its verbal elements" (p. 9). Research is needed to investigate (a) how student designers achieve (or not achieve) this unity, (b) what instructional methods would be needed to help students achieve this unity, and (c) the cognitive consequences of the perpetual process of forming and breaking down of patterns and linkages. Is this process likely to equip students with mental skills essential for cognitive flexibility, as discussed by Spiro, Feltovich, Jacobson, and Coulson (1992), or is it detrimental to students' learning, or does it perhaps depend on whether the learner is a novice or more advanced? Joyce (1988) pointed out that transformation of knowledge is the critical test that we should use to judge constructive uses of hype rtext. This study provided initial evidence of some processes involved in the transformation of knowledge during the design of hypermedia documents. However, further research is needed to provide full understanding of the nature of this transformation.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

[FIGURE 3A OMITTED]

[FIGURE 3B OMITTED]

Table 1

Observations of Words, Nodes and Links

Design
Teems
        Week 1  Week 2  Week 3  Week 4  Week 5  Week 6  Week 7  Week 8

Teem1
 Words     153     163   483       648    1063    1063    1357    1602
 Nodes       6      19    22        23      29      29      31      36
 Links       0       0     1         1       6       0      13      79

Teem2
 Words     448     601   807      1040    1040    1040    1369    1804
 Nodes       9      12    14        17      17      17      25      43
 Links       0       0    11        11      11      11      20      25

Teem3
 Words     206     308   344       595     738     846    1052    1076
 Nodes       5       5     6         7       8       9      11      12
 Links       5       5     6         7       8      10      13      12

Teem4
 Words      36      43   150       394     747     775    1238    1409
 Nodes       2      13     9        12      17      34      32      32
 Links       1      11     9         0      34      34      59      75

Teem5
 Words     606     606   929      1085    1655    1686    1789    1670
 Nodes       9       9    13        14      18      17      17      16
 Links       0       0     0         0      10      17      14      17

Teem6
 Words     514     802   802      1645    2096    2096    2130    2159
 Nodes       8      13    13        30      44      44      45      45
 Links       5      12    12        45      65      65      69      72

Teem7
 Words      84     200   329       571     711     828    1196    1199
 Nodes       6       6    15        15      19      19      27      27
 Links       5       5    15        24      36      38      56      56

Teem8
 Words     305     305   826       969    1612    1612    1953    1959
 Nodes       7       7    13        15      20      20      24      24
 Links       4       4    13        13      20      20      25      27

Table 2

Types of Project and Features of Knowledge Representation

                                  Features of Knowledge Representation

Major Types of Project         Typography               Associativity

Sequential                         X                          X
Relational-unstructured (R-U)      X                          X
Relational-structured (R-S)        X                          X

                                  Features of Knowledge
                                     Representation

Major Types of Project         Nonlinearity  Abstraction

Sequential
Relational-unstructured (R-U)       X
Relational-structured (R-S)         X             X

Notes

(1.) See Chen (1999) for a more complete description of the procedures, the class, and the setting.

(2.) The primary function of this framework was for structure building, as opposed to the function of information accumulation found in the R-U type of knowledge construction.

(3.) In general, the linking process was reportedly easier for design teams that constructed their links from a central theme or foundation.

References

Bernstein, M. (1994). Storyspace for Windows (Version 1.2) [Computer software]. Watertown, MA: Eastgate Systems.

Bolter, J.D. (1991). Writing space: The computer, hypertext, and the history of writing. Hillsdale, NJ: Lawrence Erlbaum.

Carlson, P.A. (1990). Square books and round books: Cognitive implications of hypertext. Academic Computing, 4(7), 16-31.

Carver, S.M., Lehrer, R., Connell, T., & Erickson, 1. (1992). Learning by hypermedia design: Issues of assessment and implementation. Educational Psychologist, 27(3), 385-404.

Chen, H. (1999). A case study of knowledge construction and knowledge representation in high school students' design of hypermedia documents (Doctoral dissertation, Kansas State University, 1999). Dissertation Abstracts International, 60, 1884.

Dede, C. (1988, June). The role of hypertext in transforming information into knowledge. Paper presented at the National Educational Computing Conference, Dallas, TX.

Dede, C.J. (1992). The future of multimedia: Bridging to virtual worlds. Educational Technology, 32(5), 54-60.

Ehrmann, S.C., & Balestri, D.P. (1992). Learning to design, designing to learn: A more creative role for technology. In D.P. Balestri, S.C. Ehrmann, & D.L. Ferguson (Eds.), Learning to design, designing to learn: Using technology to transform the curriculum (pp. 1-20). Washington, DC: Taylor & Francis.

Erickson, J., & Lehrer, R. (1998). The evolution of critical standards as students design hypermedia documents. The Journal of Learning Sciences, 7(3&4), 35 1-386.

Erickson, J., & Lehrer, R. (2000). What's in a link? Student conceptions of the rhetoric association in hypermedia composition. In S. Lajoie (Ed.), Computers as cognitive tools: Vol. IL No more walls (pp. 197-226). Mahwah, NJ: Lawrence Erlbaum.

Farrow, M., (1993). Knowledge-engineering using HyperCard: A learning strategy for tertiary education. Journal of Computer-Based Instruction, 20(1), 9-14.

Grabinger, S., & Dunlap, J.C. (1996). Links. In P.A.M. Kommers, S. Grabinger, & J.C. Dunlap (Eds.), Hypermedia learning environments: Instructional design and integration (pp. 89-114). Mahwah, NJ: Lawrence Erlbaum.

Hay, K.E., Guzdial, M. Jackson S., Boyle, R.A., & Soloway, E. (1994). Students as multimedia composers. Computers & Education, 23(4), 301- 17.

Jonassen, D.H. (1988). Designing structured hypertext and structuring access to hypertext. Educational Technology, 28(11), 13-16.

Jonassen D.H. (1989). Hypertext! Hypermedia. Englewood Cliffs, NJ: Educational Technology Publications.

Jonassen, D.H., & Can, C.S. (2000). Mindtools: Affording multiple knowledge representations for learning. In S. Lajoie (Ed.), Computers as cognitive tools: Vol. II. No more walls (pp. 165-196). Mahwah, NJ: Lawrence Erlbaum.

Jonassen D.H., Myers J.M., & McKillop, A.M. (1996). From constructivism to constructionism: Learning with hypermedia/multimedia rather than from it. In B.G. Wilson (Ed.), Constructivist learning environments: Case studies in instructional design (pp. 93-106). Englewood Cliffs, NJ: Educational Technology Publications.

Joyce, M. (1988). Siren shapes: Exploratory and constructive hypertexts. Academic Computing, 3(4), 10-14, 37-42.

Kommers, P.A.M. (1996). Hypermedia and multimedia concepts. In P.A.M. Kommers, S. Grabinger, & J.C. Dunlap (Eds.), Hypermedia learning environments: Instructional design and integration (pp. 1-68). Mahwah, NJ: Lawrence Erlbaum.

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: Lawrence Erlbaum.

Lehrer, R., Erickson, J., & Connell, T. (1994). Learning by designing hypermedia documents. Computers in Schools, 10(1-2), 227-254.

Liu, M., & Rutledge, K. (1997). The effect of a "learner as multimedia designer" environment on at-risk high school students' motivation and learning of design knowledge. Journal of Educational Computing Research, 16, 145-177.

Marshall, C., & Rossman, G.B. (1995). Designing qualitative research (2nd ed.). Thousand Oaks, CA: Sage.

McAleese, R. (1985). Some problems of knowledge representation in an authoring environment: Exteriorization, anomalous state meta-cognition and self confrontation. Programmed Learning and Educational Technology, 22(4), 299-306.

McFarlane, A.E., Williams, J.M., & Bonnett, M. (2000). Assessment and multimedia authoring - A tool for externalising understanding. Journal of Computer Assisted Learning, 16(3), 201-212.

McGrath, D., Cumaranatunge, C., Ji, M., Chen, H., Broce, W., & Wright, K. (1996/1997). Multimedia science projects: Seven case studies. Journal of Research in Computing in Education, 30(1), 18-37. Retrieved June 1, 2002, from http://coe.ksu.edu/JRCE/v28-5/

McGrath, D., Sylvester, A. & Chen, P. (1999, June). At risk: Traditional teaching. Bring in the learners and the Web! Presentation at the National Educational Computing Conference, Atlantic City, NJ.

Nicol, A (1989-90). Children's hypermedia compositions. Journal of Computing and Childhood Education, 1(2), 3-17.

Park, O. (1991). Hypermedia: Functional features and research issues. Educational Technology, 31(8), 24-31.

Perkins, D.N. (1986). Knowledge as design. Hillsdale, NJ: Lawrence Erlbaum.

Perkins, D.N. (1991). What constructivism demands of the learner. Educational Technology, 31(9), 19-21.

Rumelhart, D.E., & Norman, D.A. (1978). Accretion, tuning, and restructuring: Three modes of learning. In J.W. Cotton & R.L. Klatzsky (Eds.), Semantic factors in cognition (pp. 37-53). Hillsdale, NJ: Lawrence 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. 121-128). Hillsdale, NJ: Lawrence Erlbaum.

Vosniadou, S., & Brewer, W.F. (1987). Theories of knowledge restructuring in development. Review of Educational Research, 57(1) 51-67.

Wilhelm, J.D., Friedemann, P., & Erickson, J. (1998). HyperLearning: Students inquiring, learning and sharing with technology. Columbus, OH: Stenhouse Publishers.