Figure 1.1. Components of a simple hypermedia system
[ Back ]
Michael N. Louka, March 1994
HTML conversion, August 1994
Note that there is a glossary of frequently used terms close to the end of this report.
The term 'hypertext' was coined by one of the technology's pioneers and evangelists, Ted Nelson, in 1967. Hypertext systems are "a computer-based medium for thinking and communication" (Conklin, 1987:32). Nelson is developing a system called Xanadu, after the "magic place of literary memory" in Coleridge's 'Kubla Khan'. Nelson's vision is of the creation of a single repository into which the entire world's literature can be stored and accessed. He created the hypertext model of links and nodes as a means of efficiently storing such a vast amount of information. Unlike the Memex, which was never built, Xanadu has been under development for a number of years (Nelson, 1988). However no software has been delivered yet, despite numerous announcements of alpha-test versions (Adie, 1993).
In 1994, the largest distributed hypermedia system, in terms of the number of hypermedia servers in use, is the Internet-based World Wide Web (WWW) system, which originated at CERN's high energy physics laboratories in Switzerland. It is a large-scale distributed hypermedia system which anyone with access to the Internet can use. Hypermedia techniques are employed to join server computers around the world into a single inter-linked 'web'. Usage of the WWW is currently doubling every four months, with at least 100 servers on the Internet available to approximately 20 million people in more than 50 countries. The American NCSA (National Centre for Supercomputing Applications) has played a significant role in the growth of the WWW by, amongst other things, developing a free graphical browser called Mosaic, which is available for UNIX, Macintosh and Windows platforms.
Hypermedia is multimedia hypertext. However, the terms 'hypertext' and 'hypermedia' are often used interchangeably. The nodes are not restricted to text and may contain sound, graphics, video, animation or other media. Hypermedia hyperdocuments (collections of related nodes in a network) may be self contained or may link to other hyperdocuments, which may be distributed over a local or wide-area computer network.
There is little consensus regarding what, exactly, is different about hypermedia compared to other ways of structuring large amounts of information, but the hallmark of a hypermedia system is a "single coherent interface to the database" (Conklin, 1987:18). What the user sees in a window on a computer display corresponds to a node in the database on a one-to-one basis. Hypermedia offers a level of integration of information and interaction that other information retrieval systems do not.
The most essential feature of hypermedia is the link, since it is the ability to link information together in a non-sequential manner that is the primary feature of hypermedia. Links can take many forms. Those that are explicitly defined by the author of the document in the authoring process are called static links because they are fixed. Some systems also allow links to be created while the system is running, using one of a number of processes. These are called dynamic links. They can be used, for example, to incorporate security procedures (a link is only available to a particular type of user) or to allow the user to create personal links between nodes that are not usually linked, or for links to be generated automatically between a new nodes and existing nodes. Graz University's Hyper-G (Kappe et al., 1992) and Southampton University's Microcosm (Fountain et al., 1990) are both systems that support dynamic links.
In its simplest form, the links between nodes in a hyperdocument have no structure. They are associated with anchors, which are smaller parts of the nodes. Typical anchors are words, buttons and pictures. When an anchor is activated, usually by the user clicking on the anchor with a mouse, an associated link is followed to a destination node, which is then displayed for the user to see. The new node might replace the old one on the computer screen or it may appear in its own window (we are assuming a graphical interface here, but hypertext can be presented using a simple, entirely text-based, command-line interface too).
The window that the user interacts with is called a browser . It is the user's interface to the information in the hypermedia network. Mechanisms are usually provided for tracing back along a path through a network and accessing maps, indices, and other devices that offer the user an overview, and which might facilitate navigation. Browsers often offer facilities for searching for words or phrases, and some can be customised to suit the personal requirements of the user.
Figure 1.1. Components of a simple hypermedia system
Figure 1.1 shows a simple hyperdocument, with seven nodes and seven links between them.The browser shows a representation of the contents of the current node, which is 'Node A' in this example. The links are shown by the browser as anchors. If the user were to activate the anchor '1' by clicking on it, then 'Node B' would become the current node, replacing 'Node A' in the browser window or appearing in its own window. If 'Node A' were a description of a work by a certain composer, then 'Node B' might be a piece of music written by the composer, 'Node C' might be a video clip from an opera and 'Node D' might be a link to information about the town in which the composer was born. From 'Node C', the user is allowed to go to 'Node F', and thus has a route to 'Node F' that bypasses 'Node D'. By activating anchors in the node. the user can navigate through the information, following links defined by the author, in whatever way that is available to the user. A user might simply want to explore the hyperdocument, or go to a known location in it to find a certain piece of information.
Hypermedia techniques have been used for many different applications including electronic books, on-line help systems, CAL (Computer-Aided Learning) systems and collaborative writing. In fact many applications which require rapid interactive storage and retrieval of information from a large data repository can benefit from a hypermedia-based approach. The WWW system mentioned earlier is an example of the hypermedia concept being used to link information stored in a distributed network so that it can be easily retrieved from anywhere in the world. Another common use of hypermedia systems is as a framework for organising technical reports and articles, with links providing cross-references, in a manner similar to that proposed by Bush.
It is often desirable to convert existing linear texts into hyperdocuments, but the success of this kind of transformation depends on the kind of material involved. Reference books lend themselves well to conversion, and in many cases links can be generated automatically using by conversion software. Converting a traditional novel into a non-sequential hyperdocument is a more taxing task because novels are usually intended to be read from beginning to end and, therefore, links to anything other than footnotes and pictures are often meaningless and do little to enhance the reading experience. Hypermedia as a creative medium is an exciting, but largely unexplored, area, mainly because we are still trying to apply our experiences of an old medium (linear text) to a new medium (n-dimensional hypertext).
One of the greatest problems with hypermedia is usability, and little empirical research has been done in this area despite the fact that the problem is well documented in the literature. Users of hypermedia often risk becoming disorientated in the information space. A number of techniques exist to help orientate the user but it is not always clear which techniques are most appropriate to particular situations. The use of metaphors to provide the user with an interface that is easily understood and the use of overview diagrams such as maps and indices to help the user to navigate are often, but not always, appropriate. Some systems provide mechanisms for users to trace their way back through a network or to follow 'guided tours' along paths through a network to help them to locate the information that they are seeking. Some systems also allow users to add their own links between nodes and to add new nodes or annotations. In many cases this is a very useful feature, but it is not always either desirable or appropriate.
It is unlikely that hypermedia will become universally accepted as a practical medium until the overall usability of large systems improves. Both the design of the information structure and the design of the user interface reflect on the ability of a system to ultimately meet the user's requirements, as well as to convey the ideas of the author. Despite this, hypermedia systems do have great potential for meeting the requirements of people in the information age. The rapid growth of the WWW demonstrates that users are willing to make use of the medium and that hypermedia is not all hype. The social implications of global hypermedia systems could be tremendous. More research into how to create usable, successful, hypermedia is necessary if the potential of this new medium is to be exploited to the full.
There are a number of different kinds of nodes. Simple nodes are monomedia objects. In other words, they are retrieved and displayed as single units. Examples are a piece of text or a single picture. Some systems support 'composite nodes' that contain data of multiple media types, such as text with pictures. With the exception of 'leaf nodes', nodes have source anchors that are used to link them to other nodes. These terms for nodes are used by Adie(1993) and are based on the Dexter hypertext model, which is a formal reference model that describes the structures needed for links (Hardman, et al., 1993).
The size of a node is determined by the author and, as Conklin (1987) points out, the process of determining how to modularise a document into nodes is an art "because its impact on the user is not understood". Conklin examines nodes in some detail and his work, which is a classic text in this field, is one of the most complete introductions to hypertext to date, even though it was written over six years ago. In his discussion of nodes, he examines the concepts of the modularisation of ideas and ideas as objects, pointing out that hypertext imposes a new kind of awareness on both the author and the user. He notes that they must both be aware of the way the flow of the information branches.This is an important observation which is relevant to both the design of the hyperdocument structure and to the choice of content of individual nodes.
How the information contained in each node is stored varies from system to system but there is an increasing trend towards the use of markup languages for systems that link a large number of (particularly) textual documents, especially when individual nodes are distributed across a network. The most significant advantage of using a markup language is that documents written for one system using a 'standard' markup language can be understood by other systems that understand the markup language used. Fountain et al. (1990) refer to what they perceive as the problem of proprietary document formats when listing the problems with some hypermedia systems that their system, called Microcosm, overcomes. Since nodes in a hyperdocument can hold multimedia information, hypermedia systems often need to be able to recognise many different information formats in addition to markup language formats. Graphical images are a particular problem because there are so many different formats in use (GIF, TIFF, JPEG, etc.).
The most common document markup languages are LaTeX and SGML (Standard Generalised Markup Language) which are used by publishers, and are occasionally used to define the format of nodes in a hypermedia system. Documents formatted using a markup language are commonly used as initial documents for converting linear documents into hypertext. Standard formats that build on SGML and are used by some hypermedia systems include HyTime (Hypermedia/Time-Based Structuring Language) and MHEG (Multimedia and Hypermedia information coding Expert Group -- a standard that is currently under development).These are surveyed in some detail by Adie (1993).
CERN's WWW system uses a markup language called HTML (HyperText Markup Language) which is based on SGML. SGML can be considered as a device independent programming language for style sheets and so HTML, in practical terms, is a collection of styles used to define the various components of a WWW document. Documents that are to contain links to other documents on the WWW must be structured in HTML format. Non-HTML documents can be accessed via the system as leaf nodes. HTML is currently being extended to include multimedia facilities (Berners-Lee & Conolly, 1993).
Hypermedia applications developed using HyperCard or SuperCard rarely make use of markup languages, although they have been used for converting documents to and from their own proprietary formats. Instead they store the information for each node in a 'card' in a 'stack' or 'project'. Therefore, markup languages have not been used to format the electronic book that has been developed as part of this project.
Conklin (1987) identifies two types of links: referential and organisational. Referential links are non-hierarchical and are the kind of links that most clearly distinguish hypermedia from other forms of information storage. Organisational links differ in that they correspond roughly to the is-a links of semantic networks. They are used to organise nodes hierarchically in a strict structure. For example, an electronic book might use organisational links to link to the next page or the previous page (the previous and next sibling in the hierarchy) or to the contents list (which may be the parent) which in turn may be linked at a higher level to another contents list in the form of a library of electronic books. Referential links can be used to link to footnotes, pictures, or any other pieces of information, in any node.
Eliëns (1993:10) calls organisational links structural links and makes a couple of useful points about such links. The first is that structural links are essential to producing a linearised version of a hyperdocument, which may occasionally be desirable for some purposes, such as printing out a hypertext version of a linear book. The other point made is that structural links allow the author to define semantic checks whereas referential links "are by nature arbitrary".
McKnight et al. (1988) rightly complain that too little attention has been given to the problems faced by the author. Authors such as Conklin (1987) and Marchionini and Shneiderman (1988) concentrate on the users' problems and to a great extent ignore the problems of the author. McKnight et al . state that the three most common problems with documents in a hypertext system are all related to the links defined by the author. The problems are that links are made to nonexistent nodes, that node are not accessible (no links to them) and that nodes are inadvertently linked to themselves. These problems usually arise as a result of errors made by the author.
McKnight et al. only consider links that are explicitly created by the author, but the problems that they list can also occur in texts that have been converted to hypertext by a computer, or are linked dynamically at run-time. Fountain et al.'s Microcosm is an open hypermedia architecture with dynamic linking, but it is a system that is also prone to problems as Adie (1993) observes. The problem that is relevant to this discussion of incorrect links is that documents may be moved or changed, invalidating the links. Microcosm date-stamps links in order to help to detect, but not correct, this problem.
The question of dynamism in hypermedia is being addressed in several research systems. Deja-vu (Eliëns, 1993), Hyper-G (Kappe et al., 1993), and Microcosm (Fountain et al., 1990) are all systems where dynamic links are supported and play an important role in making it easier to add nodes to a hyperdocument with little effort. Waterworth (1992:97-99) gives a good introduction to the topic of static and dynamic links. Static links are explicitly defined by the author whereas dynamic links are created in a context-sensitive fashion (are inferred by the system by some process) at run-time.
Systems that have support for dynamic links also support explicitly defined links, since static links are often needed to define the underlying structure of a hyperdocument. Figure 2.1 (after Waterworth, 1992:97) illustrates the difference between static and dynamic links. It shows links defined by the author, links selected by the reader and links inferred by the system.
Figure 2.1. Static and dynamic hypermedia
Kappe et al. (1993:13) explain that that Hyper-G (an ambitious distributed hypermedia research project at the University of Graz in Austria) has 'user modes' that allow only certain kinds of users to perform authoring functions. One of the points that Kappe et al. make (and one that is also relevant to the WWW system) is that in order to avoid the 'Big Brother' scenario often associated with large information systems, it should be possible for users to access the system anonymously. One of the purposes of user levels is, therefore, to prevent completely anonymous users from authoring, while still allowing them access to the information stored in the system. Whether this kind of security will be incorporated into Microcosm is unclear, but advanced security features will be important if systems such as these are to be used in security conscious organisations to store sensitive information.
In large distributed systems, where the amount of information stored may be counted in terabytes, it is impossible to maintain static links to every document accessible via the system. Access to documents on remote databases is only possible using dynamic links (Kappe et al., 1993:10) because the information in the remote database may change at any moment. It is also clearly an advantage to be able to automatically link new documents into a large system as this reduces the amount of authoring work required and improves flexibility.
Although much of this review of links has concentrated on systems designed to store a very large number of nodes, the techniques involved can be used for small-scale hypermedia projects too. A particular advantage of using dynamic links is that the number of links that are explicitly authored is reduced. If an electronic book has over 100 chapters then linking each one individually to a contents list could be very time-consuming. One could save time by using an algorithm that examines the text that the user clicks on in the contents list and then automatically links to the chapter with a title that corresponds to the selected text.
Although the actual structure of a hyperdocument, when all links available to the user are taken into consideration is almost always a network, it is in some cases appropriate to use a hierarchic or even a linear structure as an underlying structure for all, or part of, a hyperdocument or collection of hyperdocuments. Benest (1989) writes strongly in favour of the book metaphor for the presentation and structuring of hypertext information.
Figure 2.2. A linear structure with referential links.
Benest's ideas are sound for traversing formatted linear text on a hypermedia system, but a linear structure would provide inadequate support for searching for linear hyperdocuments on a large hypermedia system. However, linear structures are useful for organising the parts of an information structure where single paths through the information are desirable. Another use for linear structures is for 'tours' that guide a user around a system. Tours can be prepared in advance by experts in order to guide a novice through a network (Kappe et al., 1993:10). In the case of a tour, it is likely that context-sensitive links would be used, so that only the links that lead along the path of the tour would be available to the user while the tour facility is in use.
Apple, in their 'HyperCard Stack Design Guidelines' (1989:25), describe a jump-linear structure (see Figure 2.3 below), which is technically a form of tree structure, where the user can traverse in either direction along a linear path, but can also jump out of the path from any node to a 'home-base', such as an index or contents list. The home-base could be a library containing several linearly structured electronic books, for example. Apple also comment that linear structures are not useful if a user is to be required to select different paths or branches within a hyperdocument, depending on her own decisions.
Figure 2.3. A jump-linear structure
Authors such as Jonassen (1989) assert that hypermedia actually mimics human memory and that the nodes should be structured as a web (network) that reflects human memory. Whether this makes learning any easier has not been conclusively proven although Jonassen (1993) has studied the effects on learning of semantically structured hypertext. There is an underlying relationship between hypermedia and the artificial intelligence (AI) concept of semantic networks which has led to some research into frame-based (semi-structured nodes) hypermedia (Koh et al., 1989), as well as Jonassen's own research into using knowledge acquisition methods from AI to create a concept map that can be used as the basis of a hypermedia structure.
Figure 2.4. Coexistence of three structuring principles
The idea of integrating different information structures into one hyperdocument is an attractive one and would appear to be sensible, since it means that some structure is applied but that the structure can be sufficiently loose that the user is not constrained by it. This is particularly relevant to large systems where design issues regarding navigation are usually complex. Navigation can be made easier for the user if the entire system is split up into inter-linked clusters of structured hyperdocuments (at least at a virtual level) in order to provide hierarchic indices and other navigational aids (Kappe et al., 1993).
Clearly, hypermedia is, by definition, structured as a network, but there would appear to be organisational and navigational advantages in using organisational links to create an underlying structure that is appropriate to the information stored, which may not be a network, and use referential links to enhance the underlying structures.
If the database is fast, dynamic linking can be a very effective way of automatically adding new documents to a system, but the situation is more complex if an entire hyperdocument is to be generated, with sensible links, from a document or collection of related documents. Simple string matching is inadequate and AI methods are being investigated, in particular for the automation of the conversion of books to hypertext.
The success of a text to hypertext conversion is often dependent on the material chosen. Automatic conversion can be performed easily for reference works (Rada, 1992), such as dictionaries, encyclopaedia, technical manuals, and textbooks, but would not work very well for a novel or an autobiography, because the markup of the book should contain logical divisions. Any novel can be converted into an electronic book that mimics a printed book, but most would not be enhanced by the conversion unless other information can be appropriately linked to the text in a manner that does not detract from the original message of the author. This process is difficult, if not impossible, to automate. Automatic conversion of text to hypertext will not, therefore, be attempted for the application that will be developed as part of this project. It is interesting, however, to review the current state-of-the-art in this area as it illustrates the complexity of authoring hypermedia applications.
Knopik and Ryser (1989) have written a text analyser that parses a textual document, looking for nouns that are used as candidate 'concepts'. Potential candidates are compared with existing concepts in a hypertext network after converting them into a standard form using heuristic methods. The authors' motivation for creating this tool was that it is difficult for an author to build a hypermedia network, maintaining the correctness and completeness of the nodes in the net. The goal for the tool is to build up part of a network depending on the content and type of the nodes linked, and to create new nodes depending on the semantics of the given type of links and the contents of existing nodes. The system is limited to three kinds of links, which are 'super-concept-of', 'is-defined-by', and 'uses-concept' links which are used to assess the need for a link. It also has a number of rules for processing potential candidates. The tool is successful for texts that don't challenge it's limitations, and is important because it demonstrates that such methods can be made to work. The authors believe that more sophisticated natural language processing is required to make the tool more accurate.
HEFTI (Hypertext Extraction From Text Incrementally, Chignell et al.,1990) is another tool that has been developed to assist the author in the authoring process. It has seven modules through which a text is passed sequentially in order to generate a hypertext document. In common with most other text to hypertext conversion tools, the task of linking other media to the text is left to the author to carry out by hand. The modules perform text preparation, node preparation, indexing, link creation, organisation and link refinement before the result is outputted in a proprietary hypertext markup format. One of the advantages of a modular system is that the modules can be altered individually in order to evaluate different techniques for each process. An interesting feature of the system is that it attempts to generate 'landmark' nodes (nodes of particular cognitive interest) which may improve the usability of the the hypertext generated.
A more recent paper on converting a textbook to hypertext has been written by Rada (1992). Rada's textbook to hypertext conversion system parses a text in a markup language format to create a semantic net of link objects pointing to text blocks. The text blocks are sequentially numbered, because a textbook can usually be viewed as a linear sequence of text blocks separated by section headings. The system then locates bibliographic citations and generates links to the bibliography, as well as generating links to cross-references and footnotes that are explicit in the markup of the original document. The index defined in the original document is then placed in an index node and links between items in the index and their corresponding text blocks are established. Rada calls the generated hyperdocument, that directly reflects the markup of the original document, 'first-order hypertext'. 'Second-order hypertext' also contains links that are not explicit in the markup of the original document. Second-order hypertext is created after first-order hypertext has been created by examining word patterns to identify new links. The system also tries to create alternative outlines (or 'views') of the hypertext structure.
The first-order hypertext was found to be more popular with users than the second-order hypertext. This is probably because the first-order hypertext has the basic structure of a textbook, which is relatively familiar with users. The quality of the second-order hypertext links is dependent on the computer's ability to generate links that are relevant, a task that probably needs more advanced AI techniques than those Rada uses. The conversion process here is less ambitious than, say, HEFTI, which tries to be a general text to hypertext conversion system. Rada's system relies heavily on knowledge of the format of textbooks and so it is unlikely to be equally successful at converting other kinds of documents into hypertext.
Many of the studies of hypermedia usability have focussed on areas such as the ability of users to find information easily, to recall facts, or to comprehend material stored in systems. However, Waterworth (1992:159-173) has carried out an empirical study to assess what users think of the quality of links between text nodes. In two different experiments, Waterworth tried to assess the quality of automatically generated links compared with links created by a human author. If automatic linking was acceptable then it should rate as highly as manual linking. As a control condition, random links were added. The results from the first experiment were rather unexpected. The automatically generated links were rated more highly than the manually generated links for both 'ease of use' and 'usefulness'. For ease of use, the results were highly significant, and for 'usefulness', the results were also found to be significant. Later statistical tests revealed that manual linking was significantly worse than both random and automatic linking, and that automatic linking and random linking were statistically not different from each other in terms of rated ease of use. In the second experiment, manual links were clearly preferred by the users but, again, no significant difference was found between random and automatic linking.
Interpreting these results combined is not too easy, and Waterworth, bravely, after finding little wrong with the experimental methods, suggests that "we would be forced to conclude from both sets of results that we may as well use random linking, since this is rapidly and easily achieved"! It is difficult to find any faults in his experiment that would suggest why the results were so unexpected, but nevertheless it is unlikely that random linking is as successful as these results would indicate, so the results of this experiment should be treated with some caution.
Although further research into more powerful AI methods for converting text to hypertext is clearly necessary, the results so far are promising. The tools reviewed here do work quite well. Advances in natural language processing technology should improve the capabilities of such tools so that they can become practical to use for the complete automation of the process of linking together documents in dynamic hypertext systems. If Ted Nelson's dream of putting all the world's literature into a gigantic hypertext system (Nelson, 1988) is to ever become reality then some very powerful text to hypertext tools will be necessary.
Unfortunately there are few methodologies available for designing hypermedia applications. Systems such as g-IBIS (Conklin & Begeman, 1988) are system rather than application oriented models. The Dexter model referred to by Hardman et al. (1993), is a standard for comparing the static structures and abilities of different static hypertext systems (Garzotto et al., 1993) and so is not suitable for application design purposes. The Trellis model (Stotts et al., 1989) is basically a model of the 'behaviour' of hypertext where networks are modelled as Petri nets, and is not, therefore, suitable for designing an application such as a hypermedia book.
There are a number of informal guidelines as to how to build applications, such as those provided by Apple (1989), Hardman and Sharratt (1989) and Waterworth (1992:31-71). However, these only provide guidance rather than a structured design methodology. Also these guidelines tend to be biased towards human-computer interface issues and to the design of browsing facilities, which will be discussed later in this review. Often the hypermedia system used defines the underlying structure. Both HyperCard and Guide (a hypertext system from OWL Ltd.) make use of linear structures to clarify the organisation of hyperdocuments. KMS (Akscyn et al., 1988) encourages the use of top-down design by forcing a hierarchic structure (although referential links that cross the hierarchy are permitted so that a complex network of nodes is still possible).
Jonassen (1989 and 1993) proposes that knowledge acquisition methods from the field of AI should be applied to produce a semantic network of concepts that can serve as a blue-print for a hypermedia network. This idea is interesting but it is difficult to see how it can be applied as a general method for designing hyperdocuments. Jonassen is interested in designing systems for teaching purposes, and so is interested in systems where concepts are associated in such a manner as to facilitate learning.
The most interesting, general purpose, model-based approach to hypermedia application design is HDM (Hypertext Design Model Garzotto et al., 1993). The authors claim that it can be used as both a modelling and an implementation device (using specially developed tools). For this project, its use as a modelling device is of interest, since the software tools are not generally available. Since the model is system independent, it can be used to simplify the conversion of applications from one environment to another (for example, from Claris' HyperCard to Xerox's NoteCards system). Another advantage that the authors claim is that it makes organising team projects easier, in that it makes it easier to split the application into tasks that can be carried out by different members of a team.
The general philosophy behind HDM is that there are two kinds of authoring, called 'authoring-in-the-large' and 'authoring-in-the-small.' Authoring-in-the-small is the implementation part of the authoring process (which is dependent on the development environment and tools used), whereas authoring-in-the-large is concerned with descriptions of overall classes of information elements and navigational structures for complex applications, without much concern for the implementation details. Weight is laid on the fact that the model is system independent.
As a modelling device, HDM supports the definition of high-level specifications. One interesting feature is that an author can define 'perspectives', which allow for alternative presentation of information. This can be useful, for example, if a system is to be bilingual or if it is desirable to be able to switch between textual and graphical representations of the same information. There is some resemblance to the Entity-Relationship models that are often used for designing databases, but there are a number of significant differences. In an E-R diagram, links between entities are shown for representational reasons, but in HDM, links are used to define navigational paths and 'entities' also have complex internal structures with organisational links.
Guidelines and principles from the field of human-computer interaction can be applied to hypermedia in order to improve the interface and overall usability of the medium. Research into the use of metaphor in the design of hypermedia is a topic that is particularly interesting as is research into new ways of visualising and interacting with information.
In sections 2.7 and 2.8, usability, navigation, and browsing will be discussed in detail, but in this section the focus will be on topics from the field of human-computer interaction (HCI) that can be applied to hypermedia rather than techniques that are specific to hypermedia.
Hardman and Sharratt (1989) carry out a comprehensive analysis of HCI design principles and guidelines in order to identify those that are relevant to the field of hypermedia. The results of the analysis are sound and can be used both when designing the hypermedia application and afterwards as a checklist to review the result. However, Hardman and Sharratt stress that the principles and guidelines that they present apply to hypermedia that is not being continuously updated. This is a limitation if the results are to be applied to large systems where documents are continuously being added, moved, and removed, but the results can be applied usefully to an application such as the hypermedia book that has been produced as part of this project, since the contents of the book will not change while the application is running.
The five design principles given are "consistency" (in presentation and navigation), "mental processing" (that the mental load on the user should be minimised), "ease of learning how to use" (to find a balance between being easy for a beginner to use without restraining an experienced user), "flexibility" (that a hypermedia system should be able to adapt to the needs of the user), and "task compatibility" (that user's expectations and requirements are foreseen and met). These are useful general principles.
Sixteen guidelines are given by Hardman and Sharratt, which are classified as "user action guidelines", "information display guidelines", "dialogue design guidelines" and "on-line assistance". These cover the layout and presentation of information and also address navigation and the ways users combine or compare information from different parts of the hypertext. They cover mainly practical aspects of the interface, such as minimising cursor movement, effective use of graphics, highlighting of information, and so on.
Thomas and Norman (1989) criticise current hypermedia systems for lack of feedback. If a user clicks on what is perceived to be an anchor and nothing obvious happens then the user will probably ignore it. This would be unfortunate if it was indeed an anchor and that the user clicked just outside the active area, or if the anchor caused something to happen in the background (such as start a clock). What seems like a simple interface may not be easy to use because of a lack of "conversational competence". The user should be 'told' that an attempt to do something has failed or that something is actually being done. NCSA's WWW browser, Mosaic, has an animated icon that shows that the browser is retrieving information. If a large file is being downloaded from across the world then the browser might otherwise have appeared to be doing nothing for several minutes. The user needs to know that something is actually happening so that the process is not unintentionally aborted. Thomas and Norman aim their criticism mainly at authors of HyperCard stacks, but it is interesting to note that Apple (1989:99) advise authors that users need feedback to reassure them they have interacted with the interface.
Waterworth (1992:57) warns that although the use of guidelines is generally positive, there is a danger of constraining the system if excessive reliance is placed on standardisation but Hardman and Sharratt's guidelines are sufficiently general and flexible that they allow the HCI designer to be creative. They state commonly accepted HCI guidelines rather than provide a strict prescription of how an interface should be. However, different applications perform different functions and require interfaces that are suitable to their purpose, so it is important for the designer to decide which guidelines can be applied appropriately. In the field of HCI, the trend is currently moving away from general principles towards layout style guides, which ensure consistency of 'look and feel' but can also be restrictive (Waterworth, 1992:59).
Laing et al. (1984:9) define graphic design as
"the translation of ideas and concepts into some sort of structural order and visual form. It is the art of books and magazines, advertising, packaging and propaganda."Graphic design is particularly important in the context of hypermedia systems because it affects users' attitude towards a system:
"The way in which information is presented, and the circumstances in which it is presented, are extremely important. If the context or the surroundings are unfamiliar, the results can be obscure, misleading or comical" (Laing et al., 1984:17).The designer needs to ask what is to be achieved, who the information is aimed at, what the message is, and how it is to be conveyed. Rough sketches are useful to determine what is important in a message and where the visual emphasis should be focussed. Sketches can be done on paper or on a computer. Using a computer is probably preferable as it is easier to experiment with how text and graphics are positioned, how scale and colour can be applied, and how making alterations can affect the the overall message.
The choice of typeface is crucial. Serif typefaces, such as Times, are more comfortable to read as the serifs assist the eye in following the letters in the text. Sans serif typefaces are very '20th Century' and may look too clinical for some applications. Decorative typefaces should not be used for large blocks of text but can be effective for titles. Script typefaces are often used to convey an air of importance or quality, and can also be used to emulate handwriting. Although script typefaces are attractive, they should not be used for more than a few lines of text as they are tiring to read. It is important that the chosen typeface is both appropriate and legible. If a handwritten letter is to be reproduced on a computer screen, it might be tempting to use a script typeface, but if the letter is more than a few lines long, it will be tedious to read.
The designer must also consider the size, weight, and colour of the text. Many lines of large text in a relatively small area can look clumsy and lots of small text is difficult to read. If the text is to convey urgency, or a warning, then making it red and bold might be appropriate to give the right kind of impact. The position and orientation of the text on the working surface is also important.
Rules and border can be used to good effect, but should be used sparingly. Laing et al. (1984:55) recommend that rules be used to:
They also recommend that if a rule is to appear in a text, then it should preferably be the same length as the line width of the text.
Borders can also be used to good effect, particularly if a traditional or nostalgic air is required, but should, like rulers, be used sparingly.
Pictures are important as they can be used to evoke recognition and reaction as well as to illustrate. Laing et al. (1984:56-59) suggest that the function of a picture can be:
Pictures can also be used to break up text in order to provide relief to the reader from continuous reading. Photographs can be cropped to remove unnecessary information, and should be checked to make sure that they convey the right mood and expression (by looking at clothes, facial expressions, age, etc.).
The colour of a background can be used to express different moods. White is neutral, greys and blues are usually cool, and creams and pinks are warm. It can often be useful to make a checklist of the qualities that are required which can be referred to when considering whether changes can be made to better fulfil the requirements.
Designing symbols or icons can be one of the most difficult tasks. Apple's 'HyperCard Stack Design Guidelines' (1989) provides a number of recommended symbols that can be used as anchors in hypermedia applications. It is better to use symbols that are well established (and therefore easily recognised and understood by users) than to attempt to design new ones. Should it be necessary to design a completely new symbol then it is important to remember that the simplest ideas are best. Laing et al. (1984:71) recommend that the designer writes down all of the words that he or she can think of related to a subject (or a function) and then take each word and try to make a simple picture from it. Sometimes elements can be combined to produce a good symbolic representation.
Texts written primarily for animators and film makers can also be a useful source of ideas for designing the time-based dimension of a hypermedia system. Storyboards can be used to design animations and for planning digitised videos, and sound effects can be used to create atmosphere as well as provide feedback.
Multiple metaphors are recommended by Waterworth if it is likely to be necessary to present different views of an information structure, using devices such as indices (book metaphor) and maps (travel metaphor). The metaphors chosen should match the system as closely as possible, but this can be restrictive and so Waterworth recommends the use of 'magic' features to provide extra flexibility. A magic feature is one that deviates from a real-world metaphor by giving the user extra powers and thus improving on the metaphor. For example, in a system that builds on a travel metaphor, a user might also be offered a time-travel option. If applied carefully, magic features will not cause confusion but will be understood and accepted by users.
Future systems will need a "sophisticated open-ended conceptual model" to support their features and "metaphorical elements are likely to contribute significantly" to that model (Waterworth, 1992:94). Later, however, he appears to almost contradict himself when he suggests that there may be advantages in having a non-metaphorical model at the overall conceptual level, but with a mix of metaphorical and magical features incorporated where appropriate. Waterworth believes quite strongly, on the whole, that the use of metaphor is necessary for hypermedia applications because of the difficulty that users have creating mental models of hypermedia structures. They need metaphors to help them navigate.
Nielsen (1990:300) does not agree with this viewpoint and states that there is empirical evidence that users behave differently when reading hypertext than when reading printed text. In particular, he criticises overemphasis on the book metaphor and notes that experienced hypertext users leave their "book habits" behind. His main criticism is that the book metaphor often limits the potential of hypertext and non-linear navigation. Preece (1993a) also comments on the use of metaphors, saying that they must, by definition, provide imperfect mappings to their target domains.
The majority of researchers, and in particular those with backgrounds in cognitive science, support the use of metaphor. Baird and Percival (1988:68-69) quote other works that show that the book metaphor, and the travel metaphor are particularly appropriate for presenting hypertext information. Dillon et al. (1993:171) comment that even those who dislike the use of metaphor use a travel metaphor when they write about hypermedia as a 'space' through which users 'navigate'. Allinson and Hammond (1988:53-63) use a travel holiday metaphor for their 'learning support environment', with guided tours, and found that the system is popular with users, who consider it to be easy use. They include features that are not part of the travel metaphor, such as quizzes, where the user can choose to be taken to the part of the system where the answer to a question can be found. This is an attempt to avoid being restricted by the travel metaphor.
Although those that support the use of metaphors recommend that they should not restrict systems but enhance them, there are others who believe that metaphors should be used to restrict systems, believing that this will make them easier to use. This is particularly true of those who believe that the book metaphor is perfect for hypertext information. Benest (1989) and Catenazzi et al. (1992) deliberately enforce the book metaphor quite rigidly because they believe that electronic information should be presented in a way that is familiar to users. Benest attacks the "new medium demands new methods" approaches as being more disruptive and less visually appealing. Users value the book metaphor (Rada, 1992) and it is probably the most appropriate metaphor for presenting linear books that have been converted to hypertext, if the hypertext version of the book only makes limited use of the new medium. However, it is difficult to see how the book metaphor can be the only appropriate metaphor for presenting hypertext information.
The Hyper-G system (Kappe et al., 1993) totally separates the 'hypermedia engine' from the user interface, so that any presentation metaphor can be implemented on top of the structure realised by the hypermedia engine. Kappe et al. hope that this will allow the system to be used to experiment with different presentation metaphors.
Benest (1989), Catenazzi et al. (1992), Waterworth (1992), Allinson and Hammond (1988), Baird and Percival (1988), and others believe that metaphors are necessary to help new and occasional users to understand hypermedia systems, but if, as Nielsen (1990) claims, experienced users leave behind their old ideas and adapt to the new medium, what is the point of adapting hypermedia to fit old concepts? Finding a 'metaphor' that describes the hypermedia medium in its own right would be a major step forward. If everyone were familiar with the hypermedia 'metaphor' then using restrictive analogies would no longer be necessary.
Conklin describes William Gibson's 'Cyberspace', as described in the novel 'Neuromancer' (1984), as the ultimate hypermedia system. Gibson's Cyberspace is a virtual reality in which users are totally immersed in the system to the extent that they literally connect their brains up to the computer and perceive information directly. Hypermedia is ultimately about communication and thought, of manipulating ideas and relationships.
With a 'grand opening' in February 1994, Brown University, in the United States, set up a text-based virtual world for real-time networked reading, writing and annotation of a fictional hypertext called 'Hypertext Hotel' (Meyer, 1994). The system is intended for the teaching of hypertext writing to students, for research into collaborative writing, and for experimenting with hypertext publication of research journals. The users of the system can create new nodes (rooms in the hotel) and link them to other rooms. Although there are no graphics, this is a system in which the user becomes immersed in a textual cyberspace and can communicate with other real people who are also in that space.
Having given a science fiction view of the ultimate hypermedia system and having given an example of a text-based hypertext system in which the user can become mentally immersed, it is worth pointing out that research into virtual reality (VR) systems is beginning to produce results that can be applied to hypermedia systems. Two of the major limitation are the raw computer power necessary and the cost of adequate computer hardware, but those limitations will be eroded with time. However, navigating a graphically rendered virtual world may not be suitable for all kinds of interaction with information. Waterworth (1992:184) points out that VR technology is highly suitable for practical skills but not so well suited to more abstract topics.
VR and 3D modelling techniques are being used to attempt to create new navigational models of hypermedia structures. Dillon et al. (1993) refer to a navigational technique offered by Lai and Manber (1991) which allows the user to 'fly' through hypertext. They suggest that this should be used as an additional tool rather than as a replacement for other navigational methods.
Hypermedia is a multi-dimensional medium and using two or even one dimensional metaphors to navigate through complex information structures may be too limiting.
Monk (1989) distinguishes between directed and exploratory navigation. In the first case the user tries to reach a know location to retrieve information and in the second case the user does not know where the information is and looks for it by exploring the information space. Addressing the problem that different researchers are using the terms differently, McAleese (1988) offers definitions of 'navigating' and 'browsing' for clarification. He defines navigating as going to a location using an overview representation of nodes and links provided by a browser, and browsing as going to a location using the links between individual nodes in a hyperdocument. Browsing is also defined as "an active information seeking activity". What differentiates these definitions is that McAleese is describing the method while Monk is describing the mode of navigation.
A study of the ways in which database users approach navigation carried out by Canter et al. (1985:100) reveals that five strategies are used:
Supporting these strategies in the user interface of a hypermedia system may improve its usability. McAleese (1988) suggests that each of these strategies, except the last one, are relevant to research into making navigation in hypermedia information structures easier.
In addition to navigating or browsing to find information, Marchionini and Shneiderman (1988:78) suggest that providing a search facility, to look for a particular word or phrase in a hyperdocument, can improve the user's ability to locate information. They warn. however, that such facilities can cause additional cognitive load for the user, and substantial processing on the part of the computer which must search the network. Clearly, in a large network, the search needs to be limited in some way. For small applications where the contents of the system change little, such as an electronic book with linearly arranged nodes, search features can be useful although deciding what kinds of query should be allowed may be difficult (Landauer et al.,1993).
Both the structure of a hyperdocument and the facilities provided by a hypermedia system for browsing must be designed to enhance the user's ability to locate information.
The basic mode of navigation in a hyperdocument is via links, from node to node, often moving through a hierarchical or linear structure to locate information. Nielsen (1990) makes many recommendations in his paper on navigation, which is one of the most comprehensive papers in the field on the subject. He recommends that animation should be used where appropriate to emphasise where the user is going. In an electronic book, for example, where the user is often traversing in a linear fashion, the pages can be animated to flip in the direction of movement.
Nielsen also stresses that users should be aware of their present location, and that the use of dramatically different graphic designs for different parts of a hyperdocument can help them to recognise where they are. However, in a large system it is unlikely to be possible to produce many different graphical designs, especially if it contains documents that are linked dynamically, but variations in fonts, background colours, and text colours can have a similar effect. However, a study carried out by Simpson and McKnight (1989) found that typographic cues had no influence on navigation. That result is not what would be expected, but may have arisen because the subjects were not experienced users of the system that they used. It is likely that typographical cues will be useful for those who have used a system long enough to recognise them. Preece (1993a:143-144) recommends the use of landmarks which the users will recognise and use to remember the relative location of nodes. In an electronic book, it might be useful to have pictures intermingled with the text. Users might then remember more easily the location of pages within the book relative to pages with pictures than if every page looked the same, improving overall navigation by the users.
Homogeneity in hypermedia is also important. Nielsen stresses that the use of graphical cues, such as the provision of landmarks, and the use of colour and fonts, must be consistent, otherwise users might become confused. If the main text of an electronic book is presented as an open book with two pages shown at once, then it should not suddenly change to show just one page in the next section or chapter.
Another technique that Nielsen discusses is 'time-stamping'. Users might find it easier to orientate themselves and perceive the importance of a node if they can see when it was last visited a node. In a large, multi-user system, it might also be useful for users to be able to see when someone else last visited a node, or when it was last updated.
'Footprints' are a more commonly used device than time-stamping but serve a similar purpose. Colour or special icons can be used to emphasise which anchors a user has activated before, or to indicate that a node has been visited previously, so that the user does not visit the same place more often than necessary. NCSA Mosaic has an option that users can select where it changes the colour of anchors that link to nodes that have been visited previously. This kind of feature is particularly useful for navigating large information spaces, especially where the user is likely to want to explore to locate information.
Nielsen also recommends the provision of a feature that allow users to trace their way back to a previously visited node. A history facility that shows recently visited nodes can be useful to speed up backtracking by allowing users to jump directly back to a particular node rather than trace backwards node by node. HyperCard has a built-in mechanism for this and an example is shown in Figure 2.5. A user can click on any thumb-nail view of a card shown to go directly to the corresponding card. Unfortunately HyperCard shows views of reduced black and white cards, so it is often difficult to distinguish between cards that are similar in appearance.
Figure 2.5. A history of the most recently visited cards in a HyperCard session.
Edwards and Hardman (1988) found that the subjects of their study of navigational aids appeared to attempt to create cognitive representations of hypermedia structures in the form of survey-type maps. Mixed-structure hypermedia can therefore create difficulties for users if the navigational devices provided do not help them to understand the structure.
Since users apparently try to create cognitive maps of hyperdocuments, an overview map should be of assistance. Dillon et al. (1993) comment that maps show users what the overall information space is like, how it is linked together, and how they can move from one node to another. Maps can also be used to provide users with short-cuts, so that clicking on an icon representing a particular node in a map will take them to that node directly. This kind of feature can be very useful. Nielsen (1990) recommends that maps should be structured, so that users can have a more satisfactory view of a hyperdocument. Structured maps attempt to shield users from the full complexity of a hyperdocument by showing only the areas that are relevant to the current context. At one level, one could have a top-level overview of the hyperdocument (or collection of hyperdocuments), while at another level, one could have an overview map relevant to the user's present location. Nielsen adds that two levels would be insufficient in a large system and that navigating between the maps could become as complex as navigating through the actual hypermedia that they represent. It is probably useful to only show organisational links if the inclusion of referential links would make the view too complicated to be useful.
Hyper-G (Kappe et al., 1993) uses collections of related nodes called 'clusters' (a cluster is a hyperdocument in the terminology used in this report). Each node in the Hyper-G information space belongs to a cluster and so, although the amount of information in the system may be enormous, a graphical overview of a user's position can be produced by showing how the current cluster is related to others and how the nodes within a cluster are related. In this way a restricted view of the system can be presented to the user. Although clusters make it easier for an overview map to be produced at all, the problem of navigating between levels of maps remains unsolved, as the maps can clearly be complex structures in their own right.
Indices and tables of contents can also be useful and there have been a number of studies into their use. Simpson and McKnight (1989) found that the user's structural knowledge is more accurate and navigation more efficient if hierarchically arranged contents lists are used rather than alphabetical ones. From this result, one can draw the conclusion that the structure of overview diagrams such as indices, contents pages, and maps should reflect the navigational needs of the user. Another study, by Wright and Lickorish (1989), found that for book-like hypertext the use of an index was preferred for navigation, whereas in a hierarchically arranged hypertext, users preferred to use links to navigate directly between nodes, utilising the hypertext's structure directly.
Both Edwards and Hardman (1988) and Wright and Lickorish (1989) suggest that it is probably helpful to offer users different kinds of overview presentations to help them to develop a cognitive map of the information structure. Edwards and Hardman make an interesting point, though, when they suggest that a user should be allowed to develop a cognitive map using one representation of the information structure before being given the opportunity to navigate through the same information in another manner. The author of a hypermedia application should be careful not to confuse the user by providing too many navigational devices that are not mutually supportive.
Navigational aids can be very useful if created with the user's requirements in mind. They should present appropriate views of the information that support rather than distort the cognitive representation that the user tries to construct of the information space. The techniques that are available now will undoubtedly be improved on when new ways of visualising and navigating n-dimensional information spaces have been found. The techniques described here are useful (especially for small applications where the information space is a manageable size) but none of them produce entirely satisfactory solutions for navigating through large complex hypermedia structures.
Edwards and Hardman (1988) define being lost in a hyperdocument as:
They also define good navigational skills as:
Research into hypermedia usability has attempted to address these problems. In some cases solutions have been offered and a number of techniques have been discussed in the previous section on navigational aids. Nielsen (1990) is a particularly good source of information regarding navigation in hypermedia. In a study with Lyngbæk (1989), he found that of two populations (a group of adult professionals and a group of kindergarten children), 56% of users of the two systems that they tested, agreed fully or partially with the statement "I was often confused about where I was". Unless Nielsen's hypermedia applications were particularly poorly designed, this would indicate that the problem of user disorientation is significant. Waterworth and Chignell (1989:207) criticise this study, however, for being too artificial. They refer to the fact that this study was not performed on regular users of hypermedia. The subjects had no prior experience of the systems used. This is a valid criticism, but the result is important nevertheless because this is one of the few empirical studies into hypermedia usability that have been carried out.
Another interesting study, and one which is of particular relevance to the application developed as part of this project, is a comparison of linear text and hypertext for information retrieval by McKnight et al. (1989). They found that subjects reading hypertext spent considerably more time viewing the index and contents list than did subjects reading linear text. Subjects had to answer questions on the contents of the two texts and it was found that those who read the linear text answered more accurately, although there was no difference in the time needed to complete the task between readers of the two texts. It seems strange that the subjects spent more time scanning through the index and contents list than they did using links to find information. This might suggest that the provision of good navigational aids, such as indices, will improve the ability of users to locate information in hyperdocuments. Alternatively, it might suggest that the provision of some navigational aids might distract the user from using the links in the hyperdocument, when following links might in fact direct the user more quickly to the information that is required.
The additional mental load commonly experienced by users of hypermedia is frequently a result of concentrating on following a trail through a network. The user may want to explore an idea and then return to a known location and this means that the user must remember how to get back to that location, as well as try to come to terms with potentially large amounts of associated information in the system. In addition to that, the user may be presented with many choices each time a new node is reached. The dilemma of having to decide which links should be followed and which should be ignored, and which deviations are worth exploring, is called 'informational myopia' by Conklin (1987:40).
Stark (1989) studied the way in which the use of pop-up and replacement windows can affect a user's cognitive load. The studies suggest that replacement windows create an extra memory demand compared with new pop-up windows. Stark found that people using pop-up windows commented more on the layout of information in a window rather than the interface. Wright (1993:148) suggests that the difference may be that a pop-up window does not represent a jump but is rather the creation of a subgoal. This would suggest that where a user is likely to want to deviate from a path, which may later be returned to, a pop-up window may be more comfortable mentally for the user. This issue is one that should be considered when designing the browsing facilities for a hypermedia system. If a user is reading the text of an electronic book and wants to go to a reference or a photograph, then the cognitive load of the user might be reduced significantly by having the photograph or reference appear in its own window rather than replacing the text that the user was reading and to which the user will probably want to return.
The problem of getting to known locations in hyperdocuments has been addressed by Monk (1989) who suggests that a personal browser would assist the user in getting to a location that has been visited previously and is known to be a good source of information or a good starting point for browsing. NCSA's Mosaic has a 'hot list' to which users can add interesting nodes that they visit. This list can be saved at the end of a session, so the user can to go directly back to those locations another time. This is extremely useful for getting to a frequently used locations quickly. Although Monk does not address distributed hypermedia systems directly, it is arguably vital that a browser for a large system can be customised in this manner because it might otherwise take a long time for a user to navigate, link by link, to some known locations.
More research into the usability of hypermedia, to improve the ease of use and efficiency of information retrieval, is clearly necessary. This is particularly important in the area of distributed hypermedia where the information structures can be enormous and the user needs to be protected from being overwhelmed by the volume of information available. For relatively small applications, the techniques described earlier in this chapter can often produce very satisfactory applications. To a certain extent much of the usability problem can be viewed as the result of users not having enough experience of thinking in an associative manner when reading. It is possible that the perceived problem of usability may eventually be seen as a far less significant problem than it is today when users have had more experience of reading hypertexts.
The chapter began with an examination of the basic components of hypermedia, looking at both research and theory. The issues regarding nodes and links and how they can be appropriately structured as hypermedia networks have been examined. The notion of dynamic and static linking has been introduced and the benefits of dynamic linking have been discussed along with the additional problems relevant to large dynamic hypermedia systems.
Different ways of structuring hypermedia information and methods and guidelines for designing hypermedia applications have been reviewed. The current state of the art in methods and tools for automating the conversion of sequential text to hypertext has been surveyed.
Issues relevant to the user interface, navigation and the overall usability of hypermedia have been discussed in detail, and a number of studies have been reviewed that aim to improve our understanding of how to make hypermedia an intuitive medium to use.
Research into issues such as versioning and support for collaborative work has not been discussed, although these are important issues for certain applications of hypermedia systems.
Adobe (1992) Adobe Premiere Limited Edition User Guide (Version 2 for Macintosh). Mountain View, California, USA: Adobe Systems Inc.
Akscyn, R., McCracken, D., and Yoder, E. (1988) KMS: A distributed hypertext system for managing knowledge in organisations. Communications of the ACM 31(7). pp.820-835.
Allinson, L. and Hammond, N. (1988) A Learning Support Environment: The Hitch-Hiker's Guide. In R. McAleese (ed.) Hypertext: Theory and Practice. Oxford: Intellect (1993). pp.43-63.
Apple Computer, Inc. (1989) HyperCard Stack Design Guidelines. Reading, Mass., USA: Addison-Wesley Publishing Company.
Baird, P. and Percival, M. (1988) Glasgow Online: database development using Apple's HyperCard. In R. McAleese (ed.) Hypertext: Theory and Practice. Oxford: Intellect (1993). pp.64-79.
Benest, I.D. (1989) A hypertext system with controlled hype. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Berners-Lee, T. and Conolly, D. (1993) Hypertext Markup Language (HTML). Available from the WWW, CERN, Switzerland (URL is ftp://ftp.info.cern.ch/pub/www/doc/html-spec.ps).
Bush, V. (1945) As we may think. The Atlantic Monthly 176(1). pp.101-108.
Canter, D., Rivers, R., and Storrs, D. (1985) Characterizing user navigation through complex data structures. Behaviour and Information Technology 4(2). pp. 93-102.
Catenazzi, N., Landoni, M., and Gibb, F. (1992) Design issues in the production of hyper-books and visual-books. pp.40-55.
Chignell, M.H., Nordhausen, B., Valdez, J.F., and Waterworth, J.A. (1990) Project HEFTI: Hypertext Extraction From Text Incrementally. Singapore: National University of Singapore.
Conklin, J., and Begeman. M.L. (1988) gIBIS: A hypertext tool for exploratory policy discussion. ACM Transactions on Information Systems 6(4). pp. 303-331.
Conklin, J. (1987) Hypertext: A Survey and Introduction. IEEE Computer, 20 (9). pp. 17-41.
Dillon, A., McKnight, C., and Richardson, J. (1993) Space - The Final Chapter or Why Physical Representations are not Semantic Intentions. In C. McKnight, A. Dillon, and J. Richardson (eds.) Hypertext: A Psychological Perspective. Chichester, England: Ellis Horwood Limited. pp.169-191.
Edwards, D.M., and Hardman, L. (1988) 'Lost in hyperspace': cognitive mapping and navigation in a hypertext environment. In R. McAleese (ed.) Hypertext: Theory and Practice. Oxford: Intellect (1993). pp. 90-106. Eliëns, A. (1993) Deja-Vu Distributed Hypermedia Application Framework. Amsterdam: Vrije Universiteit, Dept. of Mathematics and Computer Science, Amsterdam, The Netherlands.
Fountain, A.M., Hall, W., Heath, I., Davis, H.C. (1990) MICROCOSM: An Open Model for Hypermedia with Dynamic Linking. In A. Rizk, N. Streitz, and J. André (eds.) Hypertext: Concepts, Systems and Application (Proceedings of ECHT '90). Cambridge, England: Cambridge University Press.
Garzotto, F., Paolini, P., and Schwabe, D. (1993) HDMA Model-Based Approach to Hypertext Application Design. ACM Transactions on Information Systems (11) 1. pp.1-26.
Gibson, W. (1984) Neuromancer. New York: Ace Science Fiction.
Goodman, D. (1990:183) The Complete HyperCard 2.0 Handbook. New York: Bantam Books.
Hardman L., and Sharratt, B.S. (1989) User-centred hypertext design: The application of HCI design principles and guidelines. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Hardman, L., Bulterman, D.C.A., and van Rossum, G. (1993) The Amsterdam Hypermedia Model: extending hypertext to support real multimedia (Report CS-R9306). Amsterdam: CWI, P.O. Box 4079, 1009 AB Amsterdam, The Netherlands.
Jonassen, D.H. (1989) Semantic network elicitation: tools for structuring hypertext. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Jonassen, D.H. (1993) Effects of Semantically Structured Hypertext Knowledge Bases on Users' Knowledge Structures. In C. McKnight, A. Dillon, and J. Richardson (eds.) Hypertext: A Psychological Perspective. Chichester, England: Ellis Horwood Limited. pp. 153-168.
Kappe, F., Maurer, H., and Sherbakov, N. (1993) Hyper-G A Universal Hypermedia System. Austria: Institute for Foundations of Information Processing and Computer Supported New Media (IICM), Graz University of Technology, Graz, Austria.
Knopik, T. and Ryser, S. (1989) AI methods for structuring hypertext information. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect. pp.224-230.
Koh, T-T., Loo, P.L., and Chua, T-S. (1989) On the design of a frame-based hypermedia system. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Lai, P. and Manber, U. (1991) Flying through hypertext. In Hypertext '91, Proceedings of the Third ACM Conference on Hypertext. New York: The Association for Computing Machinery. pp. 123-132.
Laing, J., Walton, R., Gillies, K., and Heppell, R. (1984) Do-it-Yourself Graphic Design (J. Laing, ed.). London: Swallow Publishing Ltd.
Landauer, T., Egan, D., Remde, J. Lesk, M., Lochbaum, C., and Ketchum, D. (1993) Enhancing the Usability of Text Through Computer Delivery and Formative Evaluation: The SuperBook Project. In C. McKnight, A. Dillon, and J. Richardson (eds.) Hypertext: A Psychological Perspective. Chichester, England: Ellis Horwood Limited. pp. 71-136. Lund, S. (1944) Sol Gjennem Skyer (vol.1). Oslo: Gyldendal Norsk Forlag.
Lund, S. (1944/5) Sol Gjennem Skyer (vol.2). Unpublished manuscript. The text is available in electronic form, in the application developed as part of this project as well as an earlier electronic book (Spring 1993). Halden, Norway: Østfold District College.
Marchionini, G., and Shneiderman, B. (1988) Finding Facts vs. Browsing Knowledge in Hypertext Systems. IEEE Computer, January. pp.70-90.
McAleese, R. (1988) Navigation and browsing in hypertext. In R. McAleese (ed.) Hypertext: Theory and Practice. Oxford: Intellect (1993). pp. 5-38.
McKnight, C., Richardson, J., and Dillon, A. (1988) The authoring of hypertext documents. In R. McAleese (ed.) Hypertext: Theory and Practice. Oxford: Intellect (1993). pp. 116-124.
McKnight, C., Dillon, A., and Richardson, J. (1989) A comparison of linear and hypertext formats of information retrieval. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Meyer, T. (1994) Notes for the Hypertext Hotel Moo. Available from Brown University, U.S.A. (URL is ftp://ftp count.cs.brown.edu/pub/hypertextdocs.txt). To visit the Hypertext Hotel, use telnet to log onto count.cs.brown.edu 8888.
Monk, A.F. (1989) Getting to known locations in hypertext. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Nelson, T.H. (1988) Managing Immense Storage. Byte, January. pp. 225-238.
Nielsen, J. (1990) The Art of Navigating in Hypertext. Communications the ACM 33 (3). pp. 296-310.
Nielsen, J. and Lyngbæk, U. (1989) Two field studies of hypermedia usability. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Preece, J. (1993a) Hypermedia, multimedia and human factors. In C. Latchem, J. Williamson, and L. Hederson-Lancett (eds.) Interactive Multimedia: Practice and Promise. London: Kogan Page Limited. pp. 135-150.
Preece, J. (1993b) A Guide To Usability: Human Factors in Computing. Wokingham, England: Addison-Wesley (The Open University).
Rada, R. (1992) Converting a Textbook to Hypertext. ACM Transactions on Information Systems (10) 3. pp. 294-315.
Simpson, A., and McKnight, C. (1989) Navigation in hypertext: Structural cues and mental maps.In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Stark, H.A. (1989) What do readers do to pop-ups and pop-ups to readers? In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Stotts, P.D., and Furuta, R. (1989) Petri-net based hypertext: Document structure with browsing semantics. ACM Transactions on Information Systems (7) 1. pp. 3-29. Thomas, P.J., and Norman, M.A. (1989) Interacting with hypertext: functional simplicity without conversational competence. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.
Waterworth, J.A., and Chignell, M.H. (1989) A Manifesto for Hypermedia Usability Research. Hypermedia (1) 3. pp.205-234
Waterworth, J.A., and Chignell, M.H. (1990) The Human Interface. IEEE Computer. pp.137-145.
Waterworth, J.A. (1992) Multimedia Interaction with Computers. Chichester, England: Ellis Horwood Ltd.
Wright, P. (1993) To Jump or Not to Jump: Strategy Selection While Reading Electronic Texts. In C. McKnight, A. Dillon, and J. Richardson (eds.) Hypertext: A Psychological Perspective. Chichester, England: Ellis Horwood Limited. pp. 137-152.
Wright, P., and Lickorish, (1989) An empirical comparison of two navigation systems for two hypertexts. In R. McAleese and C. Green (eds.) Hypertext: State of the Art. Oxford: Intellect.