Link Types: a second look
One of a scientist's key responsibilities is to identify and explore the limitations of his or her work, providing a clearer path for those who follow. Randall Trigg, a hypertext pioneer whose work has always epitomized the finest qualities of the field, may have been too good a scientist in his early work. His 1983 doctoral dissertation, describing a hypertext system for scientific writing, addressed a topic of great importance with such authority (and pointed its own weaknesses so candidly) that it deterred many designers from building upon its foundations.
The key section of Trigg's dissertation -- long inaccessible but now, available on the Web -- proposes a catalog of link types. Simple links, like those familiar to all of us, simply point to a destination without explicitly indicating what the destination contains or why the link exists. Trigg sought to remedy this state of affairs by listing the varieties of links that might be expected to appear in scientific writing. A vocabulary of link types might prove useful in a variety of ways:
- Readers could use link types to understand the structure of an argument.
- Computers, by interpreting the link types, could provide behind-the-scenes assistance to readers. For example, if the computer knows whether a link leads to a citation or a counter-argument, it might be able to make better choices of window size and layout.
- By providing a rich but finite vocabulary of link types, the system could encourage writers to create richer and better-structured arguments.
After finishing his dissertation, Trigg joined Frank Halasz at Xerox PARC to create a new hypertext system, NoteCards, which replaced Trigg's fixed vocabulary of link types with a flexible, extensible mechanism. Halasz, Trigg and their colleagues, however, were not satisfied with creating an exciting hypertext system that greatly advanced the state of the art, but also explored in detail how people used the system to do actual work in their daily jobs. They found, in particular, that many users tended not to assign types to links, lest they choose the wrong type. Indeed, Trigg and Halasz did such an exemplary job of demonstrating the limitations of their approach that many hypertext system designers abandoned typed links completely.
A closer look at Trigg's taxonomy of links can be extremely rewarding. Trigg begins by dividing links into Normal links -- those that connect parts of a single scientific work -- and Commentary links that offer comments and criticisms to that work. Trigg's Normal links alone make up a substantial and powerful array; the catalog of Commentary links is even more intriguing. (I have changed a few of Trigg's names for clarity)
|Normal Links||Commentary Links|
This vocabulary of link types lets us situate each link's rhetorical role clearly. For example, an update link directs the reader to information that has been discovered since the work was originally written, but a correction link identifies an error in the work as it originally appeared.
Because Trigg was specifically addressing writers of scientific communications, he could ignore multivalence and assume that each link ought to play a single specified role. One of the key surprises in the development of hypertext since Trigg's work has been the complex and subtle role of rhetoric. Although we may claim to admire "artless" writing -- what Richard Lanham in The Electronic Word calls the style of "Clarity, Brevity, and Sincerity" -- real communication (even between scientists) is often full of multivalent messages. A simple "citation/source" link, for example, may declare an alliance or offer an apology for a past oversight in addition to playing its overt role. Similarly, a ProblemPosing/unimportant comment may represent broad institutional or ideological frictions while it addresses a highly specific and narrow issue. Multivalence, like rhetoric itself, will always be part of human communication; writing, after all, is not merely knowledge representation. Link types may not capture or express this multivalence, but that is not a unique failing: book titles, chapter headings, dedications, indexes, and indeed writing itself fails to convey all of the multivalence of human feeling and communication. By increasing our range of hypertextual expression, link types might well give all hypertext writers -- not just scientists -- a denser and richer medium.
For fourteen years the hypertext community, led by Trigg himself, has focused on the shortcomings and limitation of Trigg's link taxonomy. Storyspace has no link types, nor do HyperCard and its descendants, nor does the Web. Perhaps it's time for hypertext system designers, and hypertext writers, to take a second look.