Cheaper than printing it out: buy the paperback book.

None of these questions is new. They have been asked before in different contexts by others. If the web of knowledge were completely wired then I could tag on the appropriate historical citations at this point, and pull out the historical context for all these musings.

Researchers dream of such a heavily connected network of data and ideas. Science today is at the other end of a connectivity limit; the nodes in the distributed network of science need to be much more connected before they reach maximum evolvability.

The first step toward a highly linked web of knowledge was made by U.S. Army medical librarians trying to unify the indexing of medical journals. In 1955, Eugene Garfield, a librarian on that project who was interested in machine indexing, developed a computer system to automatically track the bibliographic citations of every scientific paper published in medicine. Eventually he founded a commercial company in his garage in Philadelphia -- the Institute of Science Information (ISI) -- that would track on a computer every scientific paper published, period. Today ISI -- a company with many employees and supercomputers -- cross-links millions of scholarly papers with their bibliographic references.

For instance, let's take one of the papers I refer to in my bibliography: Rodney Brooks's 1990 article "Elephants Don't Play Chess." I can go to the ISI system to find "Elephants Don't Play Chess" listed under its author and read off the list of all other published scientific papers, in addition to my Out of Control, that have cited "Elephants" in their bibliographies or footnotes. On the premise that other researchers and authors who find "Elephants" useful may also be useful to me, I have a way to backtrack the influence of ideas. (However, books are not at the moment indexed for citations, so in reality this example would only work if Out of Control were an article. But the principle holds.)

This citation index allows me to track the future dissemination of my own ideas. Again, assume Out of Control was indexed as a paper. Every year I could consult the ISI Citation Index and get a list of all those authors who cited my work in their work. This web would bring me to many people's ideas -- many of them very germane since they quote me -- that I might never find otherwise.

Citation indexing is currently employed to map the breaking "hot" areas of science. Clusters of a few extremely highly cited papers can indicate a rapidly moving area of research. An unintended corollary of this system is that government fund-givers use the Citation Index to assist them in determining whose research to fund. They count the total number of citations -- adjusted for the "weight" or stature of the journal publishing the paper -- of an individual scientist's work in order to indicate the importance of that scientist. But like any network, citation evaluation breeds the opportunity for a positive feedback loop: the more funding, the more papers produced, the more citations garnered, the more funding secured, and so on. And it engenders the identical reverse loop of no funding, no papers, no citations, no funding.

The Citation Index can also be thought of as a footnote tracking system. If you think of each bibliographic reference as a footnote in a text, then a citation index brings you to the footnote and then permits you to chase down the footnote to the footnote. A more elegant description of that system was coined "Hypertext" by Ted Nelson in 1974. In essence, hypertext is a large distributed document. A hypertext document is a vague network of live links between its words and ideas and sources. The document has no center, no end. You read hypertext by navigating through it, taking side tours to footnotes, and to footnotes to the footnotes, following parenthetical thoughts as long and complex as the "main" text. Any other document can be linked to and become part of another text. Computerized hypertext incorporates marginalia and commentaries to the text by other writers, updates, revisions, abstracts, digests, misinterpretations, and as in citation indexing, all bibliographic references to the work.

The extent of the distributed document is thus unknowable because it is without boundaries and often multiauthored. It's a swarm text. But a single author can compile a simple hypertext document which can be read in many different directions and along many paths. Thus, the reader of hypertext creates a different work of the author's web depending on how she goes through the material. Therefore in hypertext, as in other distributed creations, the creator must give up some control of his creation.

Hypertext documents of various depths have existed for ten years. In 1988, I was involved in developing one of the first commercial hypertext works -- an electronic version of the Whole Earth Catalog, rendered in HyperCard on the Macintosh computer. Even in this relatively small network of texts (there were 10,000 microdocuments; and millions of ways to travel through them), I got a sense of this new space of interlinked ideas.

For one thing, it was easy to get lost. Without the centering hold of a narrative, everything in a hypertext network seems to have equal weight and appears to be the same wherever you go, as if the space were a suburban sprawl. The problem of locating items in a network is substantial. It harks back to the days of early writing when texts in a 14th-century scriptorium were difficult to locate since they lacked cataloguing, indexes, or tables of contents. The advantages which the hypertext model offers over the web of oral tradition is that the former can be indexed and catalogued. An index is an alternative way to read a printed text, but it is only one of many ways to read a hypertext. In a sufficiently large library of information without physical form -- as future electronic libraries promise to be -- the lack of simple but psychologically vital clues, such as knowing how much of the total you've read or roughly how many ways it can be read, is debilitating.

Hypertext creates it own possibility space. As Jay David Bolter writes in his outstanding, but little known book, Writing Spaces:

In this late age of print, writers and readers still conceive of all texts, of text itself, as located in the space of a printed book. The conceptual space of a printed book is one in which writing is stable, monumental, and controlled exclusively by the author. It is the space defined by perfect printed volumes that exist in thousands of identical copies. The conceptual space of electronic writing, on the other hand, is characterized by fluidity and an interactive relationship between writer and reader.

Technology, particularly the technology of knowledge, shapes our thought. The possibility space created by each technology permits certain kinds of thinking and discourages others. A blackboard encourages repeated modification, erasure, casual thinking, spontaneity. A quill pen on writing paper demands care, attention to grammar, tidiness, controlled thinking. A printed page solicits rewritten drafts, proofing, introspection, editing. Hypertext, on the other hand, stimulates yet another way of thinking: telegraphic, modular, nonlinear, malleable, cooperative. As Brian Eno, the musician, wrote of Bolter's work, "[Bolter's thesis] is that the way we organize our writing space is the way we come to organize our thoughts, and in time becomes the way which we think the world itself must be organized."

The space of knowledge in ancient times was a dynamic oral tradition. By the grammar of rhetoric, knowledge was structured as poetry and dialogue -- subject to interruption, questioning, and parenthetical diversions. The space of early writing was likewise flexible. Texts were ongoing affairs, amended by readers, revised by disciples; a forum for discussions. When scripts moved to the printed page, the ideas they represented became monumental and fixed. Gone was the role of the reader in forming the text. The unalterable progression of ideas across pages in a book gave the work an impressive authority -- "authority" and "author" deriving from a common root. As Bolter notes, "When ancient, medieval, or even Renaissance texts are prepared for modern readers, it is not only the words that are translated: the text itself is translated into the space of the modern printed book."

A few authors in the printed past tried to explore expanded writing and thinking spaces, attempting to move away from the closed linearity of print and into the nonsequential experience of hypertext. James Joyce wrote Ulysses and Finnegan's Wake as a network of ideas colliding, cross-referencing, and shifting upon each reading. Borges wrote in a traditional linear fashion, but he wrote of writing spaces: books about books, texts with endlessly branching plots, strangely looping self-referential books, texts of infinite permutations, and the libraries of possibilities. Bolter writes: "Borges can imagine such a fiction, but he cannot produce it....Borges himself never had available to him an electronic space, in which the text can comprise a network of diverging, converging, and parallel times."