Hypertext is textual information organised into nodes within a web. Nodes are sometimes called pages. Associative Links are provided between nodes to move through the material. Typically a link is anchored at a word or phrase in one node, and points or refers to the entirety of another node. The best known example of a hypertext system is the World Wide Web, the main innovation of which was the ability to link between nodes on different servers on the Internet.

There are several approaches to the design of hypertext. The inventors of the concept and the term "Hypertext" themselves provided minimal guidelines on how associative links should be drawn, arguing that the information itself must determine the shape of the hypertext web [4][5]. Later, as interest in hypertext was stimulated by Apple Computer’s decision to include the Hypercard system free with the early Macintosh (1987), researchers began to study ways to help readers/users in their struggle to avoid getting "lost in hyperspace". The classic books on hypertext design date from the late eighties [6][7], and provide principles that help situate the current information within the web as a whole. For example, structuring information in a tree or a chain is often appropriate, and providing overviews of either the whole or of parts of the web helps in orientation and way-finding through the information. With the development of the World Wide Web in the 1990s, style guides have been published, providing rules not just for hypertext, but for graphical design, information latency and so on [8].

While there have been and continue to be experiments in hypertext literature (see [9] for references), little has been written about the design of such material. Most authors agree that the standard guidelines for hypertext do not apply in a creative work. For example, the author may wish the reader to traverse the text in one of a large but limited number of ways. If so, random access to the material is not desirable, and overview pages, with links to all the information, should be avoided. Furthermore, most hypertext literature (particularly that published in print form such as "choose-you-own-adventure") has a tree format structure where progress down one branch ends at a leaf and the reader must then backtrack to a branching point to read more. This constraint makes the literature easier to design but is hardly exploiting the power of hypertext.

I have found no examples of hypertext literature deliberated designed as a richly-connected web. This is unsurprising: in a novel, for example, there are likely to be many constraints on the order of presentation. On the other hand, creative works offer the opportunity to manipulate form and content together, so that it becomes possible to specify a structure and then (to attempt) to pour the writing into it. This is what I tried to do in designing my hypertext autobiography. Because events in my life are linked in my own mind in various non-chronological ways, it seemed that I could provide branching streams of narrative using those links. However, I wanted to avoid simply placing links where they naturally occurred because that would rapidly lead to tight looping (for example, where page A links to page B, which links to C, which links back to A).

Since I found no precedents for the kind of hypertext web I was seeking, I will recount my design process as it actually happened. In retrospect I think things could be done in a more systematic way.

1. I originally wrote autobiographical vignettes for illustrative purposes in a paper that I didn’t complete. So I began with various pieces from different stages in my life.
2. Having decided to attempt a hypertext autobiography, and found little in the way of precedents and design advice, I decided to structure the writing of further vignettes fairly tightly, so that fitting them into a web would be easier.
3. I prepared a table with column headings "Family", "Music", "Relationships", "Talking and writing", "Reading and watching", "Religion", "Working" and "Travel" and with rows from 1968 to 1998, which I then fit my existing vignettes into.
4. I tried to populate the table with more vignettes. It quickly became clear that I would have to limit the table both along the columns and the rows, so I left the "Family" and "Religion" columns blank as well as all the rows after 1977. In this way I created about 20 vignettes.
5. To avoid either extreme of items linking in a long chain or in a massively connected web, I decided to design a network structure into which I would place the vignettes. The network should allow branching at each node (to avoid mere serialization), but should not allow you to return to the same node too quickly. I therefore drew out a full-availability, non-blocking network [10]. This is one of many networks that has the property that you have no chance of returning to the same node in three moves, and only 1/8 probability of returning to the start node on the fourth move. Therefore tight loops are unlikely to happen. My particular network has two other properties that I thought desirable: (a) There are serial paths through the information, if you happen to choose the right links at each stage. This allowed me to provide a serial presentation in text form at the start of this document. (b) Many pairs of nodes point to the same two destination nodes. This allows for similar material to share forward links, simplifying their placement.
6. I moved vignettes from the table into nodes of the network. Many items linked themselves naturally, and I spent considerable time placing items so that almost all the natural links would be fulfilled.
7. I was then left with 12 empty spaces in the network as well as dangling links. It was necessary to write further vignettes to fill in the holes while providing appropriate connections.

I have some views on how well this design attempt succeeded.

1. The table structure was an intriguing way of stimulating memories. Although I tried to make my categories fairly loose, they definitely conditioned the kinds of memories I ended up writing down. In some ways the table was limiting, so I broke away from it at the end.
2. The need for a critical mass of items to make a viable web (32) was a problem for finding interesting vignettes. Many of the events are relatively trivial, but that may be a reflection on my life rather than my hypertext writing skills!
3. The internal content of the various pages was constrained by the formal structure. Each page has to be the destination of two inward links and the source for two outward links. So far as possible I tried to write the page initially without considering this, then, if natural links couldn’t be used, added a paragraph to provide the outward links. In some cases it is clear that a linkage is tenuous, but I preferred to do this than to add inappropriate content. I think that perhaps a quarter of my links could be improved in terms of providing a smooth flow from one page to another.
4. Hypertext allows (and perhaps promotes) variations in style between nodes. Although style variation in literature is fairly common (e.g. Tristram Shandy, Ulysses), the mental distance between hypertext pages seems to be wider than between pages in a book. I have not been able to exploit this as fully as I would have liked.
5. Once the network structure is imposed it becomes harder to add additional material without violating the design constraints. A few scattered nodes can be inserted without unbalancing the web significantly, but to weave in a complete new strand (for example, extending the time window by a few years, or adding entries on "Religion") might necessitate a redesign. Of course, if the principles of avoiding tight loops and serialization are dropped, then it is possible to add material ad lib.

There are many other experiments I would like to attempt with the form, involving style, structure and the inclusion of visual material (photos, graphics and movies). However the key question is whether the hypertext structure actually adds any value to the material. On this, I remain undecided. When I read what I have written it seems very fragmented, but, as I discuss below, this may be inherent to autobiography and has simply been made more obvious by my structure.

[4] Vannevar Bush, "As We May Think", The Atlantic Monthly, July 1945.

[5] Ted Nelson, "Computer Lib/Dream Machines", Revised ed. Micro soft Press, 1987.

[6] Ben Shneiderman, Greg Kearsley, "Hypertext hands-on!: an introduction to a new way of organizing and accessing information", Addison-Wesley, 1989.

[7] Jakob Nielson, "Hypertext and Hypermedia", Academic Press, 1 990.

[8] Patrick Lynch, "Yale C/AIM Web Style Guide" http://info.med. yale.edu/caim/manual/contents.html

[9] Terrell Neuage, "The Influence of the World Wide Web on Literature: Bibliography" http://www.geocities.com/SoHo/Lofts/5289/THESIS2.html

[10] M T Hills, "Telecommunications Switching Principles", George Allen & Unwin, 1979.