May 18, 2004

What Hypertext Is

by Noah Wardrip-Fruin · , 12:55 am

I’m working on a short paper for ACM Hypertext 2004, which will be at UC Santa Cruz this August (and where Matt Webb and I will be offering a tutorial on blogging). The short papers deadline is the 28th of this month. The working title of my paper is “What Hypertext Is” and my goal is to provide a 2-page answer to the old chestnut “What is Hypertext?” I want to give a much, much better answer than you find in many places — such as the current everything2 entry, which begins: “Hypertext is nothing more than the inclusion of links within a body of text.”

I’m including a draft below, and would definitely appreciate comments. I can’t make it any longer, but I could substitute, clarify, reconsider, etc. Here’s a preview:

We can now, based on our examination of Nelson’s texts, provide a relatively concise definition of hypertext appropriate for a world familiar with the Web: “Hypertext is a term coined by Ted Nelson for textually-focused forms of hypermedia (new media that branch or perform on request). Examples include the link-based ‘discrete hypertext’ (of which the Web is one example) and the level-of-detail-based ‘stretchtext.'”

What Hypertext Is

Noah Wardrip-Fruin


Over the past couple decades, as the term “hypertext” has gained a certain popular currency, a question has been raised repeatedly: “What is hypertext?” This paper serves to provide a relatively-concise answer to this question. Hypertext is a term coined by a 20th Century thinker — and it is in this way similar to “natural selection” or “communism.” In this case the thinker was Theodor Holm (“Ted”) Nelson, and in this paper two of his early publications of “hypertext” are used to determine the term’s meaning: the 1965 "A File Structure for the Complex, the Changing, and the Indeterminate" and the 1970 "No More Teachers’ Dirty Looks." It is concluded that hypertext is a term for textually-focused forms of hypermedia — which are media that “branch or perform on request.” Finally, the understanding of the term hypertext presented here is considered in light of the work of Vannevar Bush and Douglas Engelbart, recent critiques of hypertext, and possible future directions for the hypertext conference and community.


When we ask the question, “What is hypertext?” we are asking a question fundamentally similar to questions such as “What is psychoanalysis?” “What is natural selection?” and “What is communism?” We are asking about the meaning of a term coined by a particular 20th Century thinker. While such terms are often taken in a variety of different directions after they are coined (much as Stalin took usage of the term “communism” in a direction that Marx is unlikely to have imagined) it is generally agreed that serious discussion of the meanings of such terms must begin with the work of thinker who coined them.

In the case of “hypertext” the term was coined by Theodor Holm (“Ted”) Nelson. This paper examines two of Nelson’s early publications of the term in order to provide a definition. It is found that definitions of hypertext that focus on “the link” or “structured knowledge work” are not supported by an examination of Nelson’s work, whereas definitions that focus on media are supported. Implications for recent hypertext critiques and possible futures for of the hypertext conference and community are then presented.


While Nelson may have presented the terms earlier, his first significant publications of the terms “hypertext,” “hyperfilm,” and “hypermedia” occur simultaneously — in the 1965 paper “A File Structure for the Complex, the Changing, and the Indeterminate” [1]. As the title suggests, this paper was primarily concerned with outlining a file structure (the Evolutionary List File, or “ELF”) inspired by Vannevar Bush and intended for personal use by knowledge workers. However, a final section (titled “Philosophy”) makes a departure, noting that file structures like the ELF “make possible the creation of complex and significant new media, the hypertext and hyperfilm.” Two paragraphs later, Nelson expands the term “hypertext” in this famous sentence: “Let me introduce the word ‘hypertext’ to mean a body of written or pictoral material interconnected in such a complex way that it could not conveniently be presented or represented on paper.” In the following paragraph “hyperfilm” is briefly mentioned again: “The hyperfilm — a browsable or vari-sequenced movie — is only one of the possible hypermedia that require our attention.”

It is worthwhile to note the following: (1) “hypertext” and “hyperfilm” are coined within the same sentence; (2) both hypertext and hyperfilm are characterized as “new media”; (3) the larger category in which at least the hyperfilm is included is “hypermedia”; and (4) while the material Nelson offers in this brief section does not explicitly contradict definitions of hypertext that focus on the link, links are not mentioned.


Just as Nelson’s 1965 paper was initially concerned with the presentation of a file structure (rather than an explication of “hypertext”), his 1970 “No More Teachers’ Dirty Looks” [2] begins with a critique of concepts of “computer-assisted instruction” founded on a drill-and-practice model. Nelson’s proposed alternative to such systems is “responding resources.” He writes, “Responding resources are of two types: facilities and hyper-media.” On-screen calculators and graph plotters are given as examples of facilities. Nelson then writes of hyper-media:

Hyper-media are branching or performing presentations which respond to user actions, systems of prearranged words and pictures (for example) which may be explored freely or queried in stylized ways. They will not be “programmed,” but rather designed, written, drawn and edited, by authors, artists, designers and editors. (To call them “programmed” would suggest spurious technicality. Computer systems to present them will be “programmed.”) Like ordinary prose and pictures, they will be media; and because they are in some sense “multi-dimensional,” we may call them hyper-media, following mathematical use of the term “hyper-.”

Nelson then presents examples of types of hyper-media that could be made available to students. The first of these is under the heading “Discrete Hypertexts.” Nelson writes: “‘Hypertext’ means forms of writing which branch or perform on request; they are best presented on computer display screens… Discrete, or chunk style, hypertexts consist of separate pieces of text connected by links.” This is the first appearance of the term “link” in the essay.

Nelson next presents a type of hypermedia called the “hypergram” (“a performing or branching picture”) followed by another form of hypertext — “stretchtext.” Nelson writes: “This form of hypertext is easy to use without getting lost… There are a screen and two throttles. The first throttle moves the text forward and backward, up and down on the screen. The second throttle causes changes in the writing itself: throttling toward you causes the text to become longer by minute degrees.” Note that Nelson referred to hypertext as “forms of writing which branch or perform on request.” Discrete hypertext uses links to branch on request. Stretchtext uses no links — instead making a non-branching performance on request. Definitions of hypertext that treat the link or branching as fundamental are therefore clearly insufficient.

Nelson’s essay continues, outlining further types of hypermedia such as the “hypermap” (a smooth zooming interface) and “queriable illustrations” (a type of hypergram). The fact that both “discrete hypertext” and “stretchtext” are situated within this list of examples of types of hypermedia leaves little doubt that hypertext is a subcategory of hypermedia (which, again, are “branching or performing presentations which respond to user actions”). While the behaviors of discrete hypertext and stretchtext are quite different, what unites them is that they are “forms of writing.” While the examples shown of hypergrams, hypermaps, and queriable illustrations all include text, discrete hypertext and stretchtext are primarily textual. Hypermedia, also, are differentiated from “facilities.” They are not tools, but media — “designed, written, drawn and edited, by authors, artists, designers and editors.” From this we can conclude that tools such as spreadsheets and word processors are, along with calculators and graph plotters, not hypertext.

We can now, based on our examination of Nelson’s texts, provide a relatively concise definition of hypertext appropriate for a world familiar with the Web: “Hypertext is a term coined by Ted Nelson for textually-focused forms of hypermedia (new media that branch or perform on request). Examples include the link-based ‘discrete hypertext’ (of which the Web is one example) and the level-of-detail-based ‘stretchtext.'”


If we provisionally accept a definition of hypertext such as the above, what are the implications?

First, we can understand the relationship that the work of figures such as Vannevar Bush and Douglas Engelbart has to the term “hypertext.” Both, rather than producing work that provides a definition of hypertext, envisioned particular forms of hypertext. Bush’s Memex, with its use of links and trails, can be seen as a form of discrete hypertext when employed as media, and as a hypermedia authoring system when used to record the user’s material. Engelbart’s NLS — with its links and powerful hierarchical organization — can be usefully described as combining elements of the discrete and stretchtext approaches, and again as both a presentation and authoring system for hypermedia. Like Nelson’s proposed Xanadu, the Memex and NLS can also function as what Nelson calls facilities, as tools for knowledge work rather than media. This is similar to the distinction drawn by Nelson in 1965 between the use of his ELF concept for media (enabling the hypertext and hyperfilm) or knowledge work.

Second, a number of “hypertext critiques” collapse. In the new media theory community a number of writers have constructed a false opposition between hypertext and new media. In fact, Nelson’s writing from 1965 makes clear that hypertext has been positioned among “new media” for four decades. Similarly, those who critique hypertext as simplistic, who usually define hypertext using a simple link-based model, are shown to be operating in error. (For example, take Espen Aarseth’s statement, in his 1994 “Nonlinearity and Literary Theory,” that “[h]ypertext, for all its packaging and theories, is an amazingly simple concept. It is merely a direct connection from one position in a text to another.”)

Third, we can imagine certain future directions for the community that calls its work hypertext (from literary work to system building to theorizing related to this range of activities) and for the ACM Hypertext conference. Peter J. Nürnberg’s closing keynote for Hypertext 04 conference (“What is Hypertext?”) suggested that the hypertext community is at a moment that calls for self-definition. At times of crisis within intellectual communities it is common to urge a return to the work of the founding figure. I follow in this tradition, by suggesting that this view of hypertext opens a direction for the hypertext community to focus on types of new media for which text is central. This places us in close proximity with the digital media community, the AI community (particularly NLP), and the electronic publishing community (especially in relation to document engineering and the Web). This is to say, such a definition embraces what we currently create and study, but broadens our self-conception to include, for example, the most interesting language-oriented innovations in computer games, electronic art, educational systems, and other forms of digital media. It takes away the cause of saying, “this isn’t really a hypertext paper” for any textually-focused type of hypermedia.


[1]    Nelson, Theodor H. A File Structure for the Complex, the Changing, and the Indeterminate. Association for Computing Machinery: Proceedings of the 20th National Conference (Cleveland, Ohio, U.S.A.) 1965, 84–100.

[2]     Nelson, Theodor H. No More Teachers’ Dirty Looks. Computer Decisions 9, 8 (Sep. 1970) 16–23.