Background Literature

Hypertext systems are intended to help users more effectively retrieve information stored in database as Chen, Ekberg and Thompson (1990) describe. In traditional text, presentation of information assumes single location for each piece of information within a collection and an implicit linear order. Hypertext technology, on the other hand, offers new possibilities of document organization to the author and navigation to the user. However, with these new possibilities come new design challenges. The hypertext designer must create the structure flexible enough to express the flowing nature of hypertext system, while still clear enough to guide the user's navigation.

Navigation in Hypertext

How a user navigates through hypermedia space is affected by the linkages made by the hypermedia author (possibly better: the linkages made by the hypermedia author affects how a user navigates through hypermedia space) (Conklin, 1987). Similar in conception to Jonassen's (1986) levels of hypertext, Conklin (1987) distinguishes between two methods of linking two points in hypertext: referential links and organizational links. Referential links are non-hierarchical, while organizational links connect hierarchical information having both parent and child nodes. Having many referential links can make the hypertext document seem multi-dimensional (Anderson-Inman, 1989).

Although hierarchical and nonhierarchical links may appear to serve a purely organizational function, they may possess meaning that can influence the navigation of the user. As the complexity of a hypertext document increases, the potential for navigation problems also increase (McKnight, Dillon & Richardson, 1989). The most common navigational problem in hypermedia documents is disorientation or "getting lost" (Anderson-Inman, 1989; Botafogo & Shneiderman, 1991; Conklin, 1987; Nielson, 1990).

Disorientation

Conklin (1987) defines disorientation as the tendency to lose one's sense of location and direction in a nonlinear document, using the expression "lost in hyperspace" to describe it. Edwards and Hardman (1989) describe "getting lost in hyperspace" in hypertext applications as the result of users being unable to find desired information, or unable to gain an overview of information space. The user becomes "lost" because he does not know what actions he can perform. The problem of disorientation must be resolved. Bieber (1997) claims that proper implementations of advanced hypermedia features will alleviate these problems, as well as provide readers with a rich information environment.

Many researchers have investigated the navigation problem in hypermedia systems. Some led to techniques and design strategies for finding proper structure of a hyperdocument (Acksyn, McCracken & Yoder, 1988), while others led to visual tools such as overview diagrams or maps that help users navigate through complex hypermedia structures (Halasz, Moran & Trigg, 1987; Yankelovich, Meyrowitz & Drucker, 1988).

Navigational Tools

Graphical maps and browsers are quite useful in providing structural overviews of hypertext. Intermedia (Garett, Smith & Merowitz, 1986), Neptune (Delisle & Schwartz, 1986), Notecards (Halasz, 1988), Planetext (Conklin, 1987), gIBIS (Conklin & Begeman, 1988), and others use a graphical browser to help the users develop a mental model of the information in the system and maintain a sense of orientation as they navigate through it. This overview presentation provides a global view of a whole hyperdocument, as well as, local views of parts of the hyperdocument. These graphical representations are particularly good at representing the structure of small hyperdocuments with many cross-reference links, but are poor at representing hyperdocuments containing more than about 30 nodes because of the screen limitation. As systems grow beyond a browser's ability to depict concisely the information contained there, however, it becomes increasingly difficult to maintain a sense of orientation and context by looking at the browser. Techniques such as fisheye view (Furnas, 1986), and pruning help manage this problem.

There have been a number of studies investigating the use of maps, but very substantive evidence cannot be drawn from them. Hammond and Allinson (1989) compared the use of maps with no maps on performance of a database inquiry task. They found, although the performance was superior for subject using the maps, this difference was not statistically significant.

Stanton, Taylor and Tweedie (1992) conducted a study to determine if maps were an appropriate navigation aid for searching in hypertext documents. The researchers found that the high scores in the "no map" condition significantly outperformed the high scores in the "map" condition on the sentence completion tasks. They also found that those in the "no map" condition significantly outper-formed those in the "map" condition on the cognitive mapping task. In addition, those in the "no map" condition felt as if they had more control over the lesson than those in the "map" condition. Stanton et al. (1992) concluded that in an information search and retrieval task, maps actually impaired performance.

Chen, Lin and Sun (2000) have investigated the effects of navigation maps on search performance, browsing behavior, and cognitive map development within hierarchical structured hypertext courseware. The results of the study showed that map type — the global map in particular — significantly affected students' search performance, and their browsing behavior. However, students with no map had the highest scores on the development of the cognitive map. Providing maps led not only to poor performance, but also to less perceived control and inferior development of cognitive maps when compared to a no-map group. This suggests that maps may hamper efficient use of a hypertext system.

Dee-Lucas (1996) studied the effects of overview structure on text review and the resulting internal representation when readers had a general and a specific learning goal. The results of the study showed that the hierarchical overview was easier to use in deciding what to review, as well as, for initial reading. The result of this experiment indicates that providing structure in a hypertext overview can enhance its usability by facilitating selection of a unit. Besides, regardless of the learning goal, readers spent less time selecting units from the hierarchical overview than from the list overview.

Providing a hierarchical skeleton, or tree structure, for hypertext organization can help users to develop a mental model of a system and maintain orientation (Akscyn, McCracken & Yoder, 1988).

Mental Models

Rovie and Morris (1986) have defined mental models as "… mechanisms whereby humans are able to generate descriptions of system purpose and form, explanations of the system functioning and observed system states and predictions of future system states" (p. 351). A user who is interacting with a system constructs a subjective causal model of this system in order to give reasons for the invisible structure of the system or for the process running in the system (Seel, 1999). However, mental models are often incomplete and do not necessarily reflect a technical correct image of the target system (Norman, 1983). In fact, sometimes users are not even aware of having such a model. Their mental model is only partially represented in conceptual structures. Mental models may comprise structural knowledge about the system, rules and procedural knowledge (e.g., on how to perform operations). They capture only aspects relevant to the user.

Users acquire and expand their mental models by observing the system's behavior in the course of its use, by training, instruction and documentation. The mental model is also expanded by deduction based on the existing model and is influenced by experience with other systems (Rupietta, 1990). During the design and development of a software system, Norman (1983) has suggested that designers need to create a conceptual model, which is a representation of the system on a conceptual level depending on the specific view of the designer, of the mental model. He suggests that designers should actively attempt to elicit mental models in users for the system being designed. The conceptual model should take into account the beliefs users are likely to bring to the system, the need for the concepts that make up the mental models to be observable in the system, and the need for the mental model to have predictive power.

Designers must not only consider how to structure knowledge from a system advantage, but also what views and corresponding navigational tools are provided for the users. The views and navigational tools will be assimilated easily into a mental model for a system if they are familiar. This reasoning lies behind the desktop and other metaphors (Marchionini & Shneiderman, 1988). Designers must know how a user seeks information in traditional print systems, if they are to provide effective interface for a new system.

Metaphors are well-recognized mechanisms for providing easy to use, interactive access to the complex functionalities of information systems. Metaphors are also employed for facilitating the forming and development of mental models. Examples are a storyteller, a navigator, a guide, a book, a map (Cartwright, 1999) and a travel metaphor.

Table of Contents

In the paper world, a table of contents is usually given at the beginning to provide a systematic list of headings identifying the items discussed in the documents; "it is an aid to both ways-finding and sense-making" (Carey, Hunt & Lopez-Suarez, 1990, p. 58). A table of contents orients the reader to the scope of the publication as the authors intend it to be viewed. In hypertext systems, a table of contents offers an overview of the document content, and overviews have often been recommended as useful navigation aids (Utting & Yankelovich, 1987).

A number of hypertext systems use a table of contents to aid navigation and represent the structure of a hyperdocument (Jacques, Nonnecke, McKerlie & Preece, 1993; Perrott & Smith, 1992). WebTOC (Heflin, Komlodi, Pasricha & Tan, 1998) is designed to diminish the difficulties for users to get an overview of the contents and the structure of the Web sites. A WebTOC provides a graphical description of the hierarchy of a Web through a hierarchical table of contents that employs a fisheye view method. The study comparing WebTOC, Textual TOC (a basic tables of contents interface), and ordinary Netscape found that the differences in task performance times were not statistically significant. However, users prefer having a table of contents of a Web site regardless of performances.

SuperBook (Landauer, Egan, Remde, Lesk, Lochbaum & Ketchum, 1993) is designed for accessing static documents, and its major concern is the improvement of retrieval efficiencies for each individual. The SuperBook browser provides computer-based enhancements to texts. It makes heavy use of the paper document's table of contents for hierarchical searching. SuperBook's dynamic expanded table of contents is not presented as a linear structure, but rather as an appropriation of a fisheye display (Furnas, 1986). When a document first comes up, only the highest hierarchical level of headings is displayed. Next to each heading, under which there are subheadings, is an asterisk, indicating the user can expand that section to its next lowest level by a mouse click. The user can open up as many different parts of the table of contents at the same time as desired.

Frames

Multiple-window environments have been used extensively for a decade in typical hypermedia documents as navigational aids. Frames, one may call windows, also have emerged as a feature of Netscape Navigator. The frames feature provides Netscape Navigator with sophisticated page-presentation facilities to display multiple, independently scrollable frames on a single screen.

A frame is a [rectangular] region in the window. It can have a name, so it can be addressed or targeted by being linked to documents. If a screen is divided into regions, each one is a frame, and a particular combination of frames is called the frameset. To use frames, one document must define the frameset, so that other documents can be displayed in each frame (Engelfriet, 1997).

According to Nielsen (1996), the fundamental design of the Web is based on having the page the atomic unit of information, and the notion of page permeates all aspects of the Web. The simplicity of the original Web contributed to its ease of use and its rapid update. Frames break the unified model of the Web and introduce a new way of looking at data that has not been well integrated into other aspects of the Web. With frames, the user's view of information on screen is now determined by a sequence of navigation actions rather than by a single navigation action. However, Nielsen (1996) claims that these problems with frames will go away over the next few years.

Currently, numerous home pages on the World Wide Web (WWW) use frames to present a table of contents. Unfortunately, no actual research studies are found either to support or to refute the use of these tools as generalized guidelines for hypermedia or Webpage designers. There are, however, no general guidelines for structuring and for the interface since their optimal design depends upon the task for which the information system is used (Vries & Jong, 1997). Thus, the hypermedia designers need to consider the types of tasks users will be undertaking in their design of hypermedia navigation system (Wright & Lickorish, 1990).

Information Seeking

Information seeking is a fundamental learning activity, a precursor to many others (Jonassen & Grabinger, 1990). It is also a pervasive human activity (Nickerson, 1986) and a special case of problem solving (Machionini, 1989). As problems can be classified into two broad categories — well-defined and ill-defined — in terms of the clarity of the starting point and goal, search targets are different in their complexity and specificity. Findings from studies that focus only on the retrieval of facts cannot be generalized to the wide range of information-seeking activities in which learners may be engaged in real information-searching situations.

Research has shown that information searchers prefer search facilities such as key word search or index mechanism (Joseph, Steingberg & Jones, 1989). But when search questions are vague, people tend to resort to the browsing or exploring strategies (Marchionini & Shneiderman, 1988). Marchionini (1992) proposed a framework of five functions involved in information seeking tasks. This framework reflects the iterative, nonlinear, and opportunistic characteristics typical of end-user information seeking patterns. Tonta (1991) has argued "…hypertext systems are not designed for fast and efficient fact retrieval. Rather they support unhurried and informal information searching" (p. 22). Therefore, when the search task is simple and specific, comparing different browsing systems is actually a comparison of search facilities and indexing schemes provided in each system rather than a comparison of the structure or interface design (Lai & Waugh, 1995). This is consistent with Vies and Jong (1997), who argue that information systems, and especially hypertext, can be used for tasks other than searching for information, such as information exploration. Information systems are thought to facilitate both problem structuring and problem solving (Begoray, 1990). In fact, on the task side, the development of hypertext systems coincides with a shift in interest from searching and locating a specific fact to tasks involving browsing and the exploring of subject areas.

Even though there is no actual research on the use of frames and the expandable table of contents on WWW applications, on the basis of previous research studies, the following hypotheses were generated for the present study.

H1:	Users' performance in information searching and browsing with expandable table of contents will be more accurate than with traditional table of contents.
H2:	Users' speed in searching and browsing with expandable table of contents will be higher than with traditional table of contents.
H3:	The navigation will be more satisfy using expandable table of contents as opposed to traditional table of contents.
H4:	Users' performance in information searching and browsing with multiple-frame will be more accurate than with single-frame.
H5:	Users' speed in searching and browsing with multiple-frame will be higher than with single-frame.
H6:	The navigation will be more satisfy using multiple-frame as opposed to single-frame.